Cloudflare Stock

All right. Fantastic. I will go ahead and kick us off here on stage. Thomas Seifert, CFO of Cloudflare. Really grateful to you joining us multiple years in a row. Thank you for taking the time.

I'm aging, yes.

So I comment, I love that. I get to ask you this question because you're going to give me the most conservative possible response.

The most thoughtful.

Most thoughtful -- most balanced. We have a debate with investors on this concept of the inference economy which is the idea that, look, we're so early in agents and agentic capabilities that the best part of whatever product cycle is tied to agentic inference is still ahead of us. Now you're already putting up 36% growth at the very beginning of this product cycle. So if we would extrapolate out and say, look, the best part of Cloudflare's growth opportunity tied to workers, tied to SASE, tied to Act 4, it's still ahead of you. That would suggest directionally that the best part of your growth story is also still ahead of you. So talk to us a little bit about the framing there. Why would that not be true or what would we need to see to feel even better than it is going to be true?

Here, we -- one of our founders, Michelle, will make fun of her. Because every year, when we have kickoff for the year and talk about the targets for what is to come and how we approach the year, she says, we are only getting started. Now this is our eighth year as a public company. So the beginning of the year, she said at the eighth time, and I truly think that what she just said is true. When we think about our life in Act, Act 1, Act 2 at Act 3, Workers, Workers AI, we'll talk about it, Act 4, content protection, how we monetize it. We still -- we think that, that opportunity is dwarf -- is going to dwarf what is already behind this.

Last year, we were sitting here, and agentic AI was still far away. The step-up of the LLM only came in November, and we were primarily concerned about convincing you all that our sales execution was on track. That productivity was coming that we felt comfortable investing in sales capacity, and that was driving the early steps of acceleration in growth in a quite significant way. Now I think we are what you described in the -- in this paradigm shift of what an agentic AI world is going to do to us. And we find ourselves at a point where the network, the architecture, the biggest competitive moat of Cloudflare, the network itself seems to be designed for what is going to come and support what agentic AI workflows are looking for, super low latency, secure sandboxes to execute code and so on.

So what you look for are the signs that signal this early -- the early innings sign because of the architecture of the network and the benefits, developers are flocking to Cloudflare. So that is an early sign that I think is signaling this dramatic shift. Developer account was up 2 million developers quarter-over-quarter alone. You see that the acceleration of revenue is now happening in what we call variable revenue, right? So consumption-based revenue. Another big shift, I think, in the agentic world is that we move and disrupt SaaS from a seat-based metric into highly variable consumption-driven revenue. That is where we see acceleration.

We're at a point now where pretty much all the large names of AI native companies, whether it's Lovable, Sigma, Anthropic, OpenAI, building on Workers on top of the network. And in one of the stats we gave at Investor Day was that we think about 80% of the significant AI enterprise infrastructure is a customer of Cloudflare. So all the signs you want to see.

And there's one trend that I think is signaling what is going to come when we talk about Act 4. There was this vertical that we always were highly underrepresented, it was content creators. Media was always a vertical that was revenue weak at Cloudflare because there was not a lot of value for us, just moving pipes from one place to another, and they are now shifting towards Cloudflare to start to protect and make their content that they create scarce in order to generate leverage.

So I think there are so many structural signs that signal what is coming. And that is, to be very honest, what gave us the confidence at Investor Day to raise our North Star, right? We said for many years now, 40% was our North Star, we shifted that to 50%. We'll exit this year with significant momentum in that direction. And I think there's clearly a world, and that is what we also said beyond '27 where we will live in a world beyond the Rule of 50. And that will be the result of acceleration Act 3 and Act 4.

Let me pick up on the network architecture comment that you just made. Many investors in the room will spend time at the hyperscalers. Hyperscalers will invest hundreds of billions in the CapEx guidance to be able to support training clusters, but also inference clusters and there's some debate around mix between training and inference that we'll leave aside for now. The interesting thing about Cloudflare is you were not investing hundreds of billions of dollars and yet you have this product cycle and compute tied to inference in the -- at the edge. So maybe square that cycle for us.

Well, a big part, I think, is already what you mentioned. When the CapEx was started, we said training is not our world, at least not at this point in time, not for the super large models, things might change as models get smaller and maybe run even more on-premise, on enterprises, on open models. Our world is inference. Inference is an interesting, interesting world, right? When we talk about AI, it's in the larger extent the GPU-driven discussion. There was a very interesting part at Investor Day in Rita's presentation on how important CPU capacity is in order to orchestrate agentic AI workflows. Where does the data come from? Instruction, what needs to get computed? What API needs to get access? That is our world. That is what Cloudflare's network is today already.

A reason why our gross margins used to be so high and still are high, because we really understand well how workloads get scheduled. That allows us really to take that strength into this AI world of inference where we think GPU is important, but the demand curve is really bending when it comes to CPU and memory capacity, and that's where we excel.

In this world of inference, I think an important advantage for us is that we do not have to provision ahead of demand, right? Two years ago, when we started work on Workers AI, I mean, there were so many questions. When will we see revenue? What is the pricing? And we always said, it's not what's important. What is important is to get as many developers on the platform. We need -- we want to understand the variety of inference workloads because that is what we need to optimize, right? We need to understand what hardware stack we have to put in place and learn how to schedule AI workloads in a way that gets us to utilization that is similar to our Act 1 and Act 2 products.

That work is now paying off, right? We have made a tremendous progress on how we schedule inference workloads on CPU capacity, it doesn't mean, we don't need GPU capacity, we deployed GPU capacity pretty much in all of our PoPs in all of our data centers. But that is the strength we built from based on the fact that this investment can follow a demand signal, and we do not need to provision ahead of it. It also means, that doesn't mean that our CapEx will stay low. CapEx can go up but it will go up in conjunction with me telling you that revenue is going to come up, too. So we are quite comfortable that the CapEx efficiency is going to hold for a while.

You said something interesting there, which is the training opportunity may change as models get smaller or not quite the training opportunity, but there is a scenario where the industry evolves and if we're training smaller models, maybe Cloudflare has a role in that. Maybe just elaborate on that on what you mean...

I don't want to go too far there and get ahead of my skis. But what I think is that how you optimize -- smaller models and open models will put pressure or pricing pressure on tokens. And how you cope with that from a revenue perspective, right? In order to get throughput up, so you can have -- you optimize revenue in a different way. I think it's going to be really interesting, and I think this is our playground. So it will play -- it will be another vector that will play to our strengths, whether that is inference or early training adoption of models to enterprise-specific use cases. On enterprise-specific data, maybe even on-premise in order to make sure that the data leakage is contained.

We're actually hearing more about this idea that enterprises could bring more compute on-premise. So maybe pick up the thread there. What do you think the opportunity is for Cloudflare, whether it's on the sovereign SASE side or more on the on-premise compute side?

Well, we were just debating on centralized large data centers versus a highly decentralized network. That is what we are, right? We drove data sovereignty and data privacy products from a very different perspective before AI was even a topic. Those products are now are the basis of making sure that either critical infrastructure customers -- sovereign nation customers, federal business in parts of -- in friendly parts of the world can contain their data privacy and data sovereignty risk. We are built for that. You literally can wrap something around a byte of data. You're a citizen of Zurich and your credit card can only move between those 2 cantons, and can only be on a data center of that security level. That is what the network was built for, and that's why we think this will be another growth vector that will play well for us.

Absolutely. Okay. I want to ask you about Act 4. And if I think about some of the products that you've announced around identity payments gateway, one of the comments that Matthew has made is, look, one of the reasons this has become important to Cloudflare is because the existing payments network don't give you the throughput that you would need to support agentic transactions at scale. I would otherwise have been skeptical about Cloudflare's ability to build all of these pieces of infrastructure, but for the fact that you've done bold things with your architecture before and you have Stephanie Cohen driving some of this for you. So talk to us a little bit about what the building blocks are to be able to actually realize the vision on agentic payments, agentic transactions, Pay Per Crawl, Act 4.

Let's start with Pay Per Crawl, right? This is how it all started when we saw that crawl rates and links were completely diverging and content creators lost an opportunity to monetize the content that they created. These were the early innings of Act 4. And we started with established products like bot mitigation of our tool in order to give content creators a tool to make their content scare and put them in a position to negotiate for people paying for crawling their content. This is how it all gets started.

There are 2 sides of the situation. Meanwhile, we are working on both sides. You probably saw our announcement -- research announcement with OpenAI to index what we have -- today, LLMs crawl literally every page that is important every minute because they don't know what changed, right? And you don't want to miss that an important football player dropped out of the game 2 minutes before it starts, but this costs significant of money, bandwidth that ultimately stayed by the content creators. So more than 20% of the [ bet ] is behind us, we can signal what changes. We know what changes. We know what has refreshed and that is valuable knowledge for the LLMs that they are ready to pay for because it contains the complexity of what they need to do. It reduces the cost and it makes our friendly ecosystem. So significant progress.

Now we talk about how we monetize transactions. And when 20% of the [ bet is ] or more is behind on our network when all these agentic AI agents and workflows go to our network, we are in the flow of what commercial transactions are going to come and you want to be a part of it. So we announced a couple of products that our first step in that direction. We announced what we call monetization gateways, so content providers can literally sell what they have. We introduced Cloudflare Wallets, which allows you and your agents to have autonomous transactions. We announced identity tools, and there's more to come. The big lift to be honest, is still in front of us because Matthew, I think, gave some stats on the last earnings call, we had of how many, 10,000 transactions per second, a significant payment provider today is able to handle that is literally in the 10,000s of transactions. We think we need to be in the millions or tens of millions of transactions in small value, high volume on our network, another -- and we are quite excited about that. There's significant effort going in that direction together with partners. Literally every credit card, every large fintech company is part of this effort. So we think that is why it all accumulates to Act 4 being probably the biggest opportunity that is still in front of us and it's going to dwarf whatever we have done so far.

Say more about the heavy lifting to realize the vision. You mentioned -- do you already have the ability to do the throughput at the hundreds to thousands, millions?

We are not at the millions yet. This is work in progress. There's engineering -- significant engineering effort going in that direction, yes.

What about from a partnership and an ecosystem standpoint?

All the partnerships announced. There's Visa and there's Mastercard, there's American Express, there's PayPal, there's Coinbase. So it's a large set of partners that need to come together, but will provide the infrastructure in order to enable that, yes.

When is the right time for us to get excited about seeing it in the numbers?

Well, remember, it's about building a durable business model, right? So we pace our time in light of the opportunity. We cannot take this lightly, processing millions of transactions per second that will be a responsibility and obligation. We want to make sure that the tech is ready for it. And there's still so much earlier stuff that needs to happen that gets us excited when we talk about Act 3. So I think that acceleration is coming, and we're in a good track.

This is a little bit of a framework monetization question for you. So when I think payments, I think, 275 basis points as a take rate. This is a little bit interesting because I guess you could do it as a take rate model. You could do it as a per transaction model. What are some of the things going through your head?

This is where you pay people to give you good advice. We are working with specialists in this arena, even down to academics, figuring out what the best model is. I think it will be -- there's no one size fits all. So there will be a differentiated approach. But I think we -- like so many things, when we start, we look at it from a first principle perspective. And we really need to start from scratch in order to come up with a model that is fit for purpose.

Very fair. Okay. Let me ask you a little bit about the growing piece of your business that is tied to pool of funds and consumption revenue. One of the comments that you made this quarter was customers consumed more intra-quarter, which is where some of the upside came from relative to the guide. How do you think about -- there's a lot of ease of use in pool of funds that reduces friction. At the same time, if more of your revenue is coming from consumption revenue, it creates some variability from quarter-to-quarter. How do you adopt that into your forecasting?

That's a really good topic. Literally, the team, a large cross-functional team is meeting tomorrow for 2 days to really rethink most of our processes from scratch. Forecasting is one. We have to tie way deeper into the infrastructure to understand what drives consumption. A SaaS model was a demand-limited model, right? If you have the time, you figure out how much you get.

In this world, we are not demand-restricted it seems, right? It's what products and what supply do we have at the right point in time, how you forecast that? What are the early indicators. You have to go far below the billing metrics that indicate or could signal consumption. And then there are -- in the SaaS model, you can ask your customer and say, how many people do you intend to hire next quarter? How many more seats are going to occupy on the SaaS product in the agentic, our customers don't know where demand is going. So you have to really rethink what indicators you take. We have -- for this year, at least, we talked openly at our earnings call about it. In this world, guidance becomes more difficult if you want to keep the range, you have to shift the midpoint, right? Because we want to surprise you all to the upside and not to the downside. This approach only goes so far, right? Because the numbers become so big and the growth rates becomes so steep, so we have to figure that out. But we have an excellent data science and finance team, we'll get this under control.

What do you think was unique about 2Q? It sounded like something changed in 2Q in terms of the way customers were consuming, where they started consuming more relative to whatever your baseline forecast was. What do you think was unique about 2Q? And why wouldn't the same thing happen in 3Q?

That it goes faster or...

That customers consume more than your baseline forecast.

That would be a good thing, right? I'm not going to exclude that.

Why do you think it was 2Q...

Well, if you think about -- so if the upside comes from variable consumption, it is more agentic AI traffic. It's more requests. It's faster adoption of products across pool of funds. If it's a pool of funds contract, people burn through their commitments way faster. We renew mid-quarter. That is the dynamic around that. But you ask infrastructural signs of us participating. I think this is one of the structural signs where we see that the flywheel is starting to spin and it's probably going to spin faster and not slower moving forward.

Is there a more heavy lifting to do on the go-to-market for enterprise?

Heavier lifting, I don't think it's the right word, but we see that serving agentic AI customers is different. Let me put it this way, in terms of the personas, in terms of the sales cycles, they are way shorter. The deal sizes are way bigger, but it needs a significant support in terms of forward-deployed engineers that can nerd out on a super technical level. And I don't think this is a major lift, but it's an evolution of the go-to-market model, but it caters to us, right, having forward-deployed engineers, people that are able to talk technology with the customers that has been our strength. But yes, that evolution is happening, and it's happening pretty fast.

Let me ask you a little bit about the core businesses with CDN and SASE. One of the hypotheses that we're pulling at is if agentic traffic goes up, core CDN has historically been -- or network security and CDN inversely, forward proxy, reverse proxy, fundamentally, have been a function of network traffic. So I guess the question for you is, have you started to see a change in network activity? You have the Cloudflare Radar where you show percentage of agentic traffic versus human traffic. But it's more a question on, does it create an upgrade cycle in CDN? And then does it actually -- there's so much more to talk about on the security side, too, is there going to be an underlying upgrade cycle in the core Act 1, Act 2 businesses?

Yes. But for more differentiated reasons. I think we see an uptake for sure in Act 1 because verticals that have not been on the network are moving to the network, right? Our share in the media vertical has and publisher vertical has gone up quite a bit, so for sure about this. I think you'll see an uptick in SASE-driven products. All enterprises start to deploy AI like there's no tomorrow, and then we wake up and figure out and we need to control what data they access, what data shouldn't leak. So there's a whole slate of products that comes with us, that's what I meant. And agentic AI request might be lower in gross margin, high growth rate, but it comes with a significant opportunity to wrap other value-added products and services around it. We see that in Act 2. So there is there's an acceleration from the baseline, not to the level of Act 3, but you can clearly see that.

Maybe stay on this point, why would agentic traffic be lower in gross margin? And why -- what's the attach rate on the product side that's higher?

Well, it is -- we are quite good at making progress on how we schedule, but the utilization of the throughput on a given CPU or GPU in terms of how many tokens you can generate within a given set of compute capacity is different, right? And we are getting that up. So the gross margin is improving. But that is the reason why agentic AI works, gross margin is lower.

But it comes with really low cost to book. And so we make it up on the operating margin side. And then the workload needs -- it needs to be secure. It needs sandboxes, AI gateway products that all come along with it that make for an upsell or a cross-sell opportunity. It all happens on a platform. If we are lucky, the customer is on a pool of funds contract. So the friction to adopt is really, really low. So that's why every request. I remember, I will say that when I teach an incoming class at Cloudflare, I said one of the things that really stood out in my interview many, many years ago with Matthew was, and there was no Act 2, there was no Act 3. Thomas, when we think about pricing and customers is the most important aim is to never discourage a byte of data from moving through the network. This stood for me, and this drives a lot of the decisions, how we price, how we discount, how we think about products, is never discourage a byte of data from moving through the network because we will find opportunity to wrap value around that byte of data. And in a world of agentic AI where they request the AI -- agentic AI or agent initiated requests are going through the roof, a multiple of human-initiated requests, that opportunity, of course, multiplies too.

Let me ask you a question on competition. At various points through Cloudflare history, I think back to the early CDN days where Akamai was part of the conversation. And then I think through some of the wins that you've announced recently on the security side, whereas the Zscaler has been the dominant SASE player, all the way through Act 3 where you've won hyperscaler displacements more recently. Maybe just take a step back from all of this, how do you see competition evolving in the spaces you compete holistically, rather than us creating a map of saying, well, here are each of the products and here's who compete in each lane.

We have formidable competitors out there. So you have to acknowledge this. But I think there's a strength to the business model that is really hard to attack. And that makes it easy -- it's not easy but easier to defend. There is this architecture of the network that is just so superior and our principle and permits to keep every product on every server, in every region is just a flywheel in itself, right? We started with 5 products, I think, and when we went public, we are now getting to 70. The products live off each other. The reason why SASE products have high gross margin is because they live off the infrastructure that Act 1 has built. Act 4, it will live off the infrastructure that Act 3 has built and therefore has the significant margins.

And then you are now in a world where you have so many independent growth vectors driving the business that there are so many ways you can go in order to evolve the business model. So I think this makes this very unique, and that's why over the years now, the competitive position and strength has increased in some areas, dramatically, increased compared to the competition. And that flywheel of Act 4 will make that mode even higher.

I want to end here with a conversation that we started with Matthew right after the RIF on how do you operate a company in 2026 such that it's different from perhaps the SASE era or the 2015 era? And so I'd love to hear how have things changed reflecting inward on Cloudflare, you've always been an innovative company, I think, of Birthday Week, for example. But share with us a little bit about what's changed behind the scenes to the extent that some of that innovation has been amplified.

When we announced our employee reduction in the second quarter. There are many big question marks of why we would do that, how could inefficiencies have built up that much. And it took us until Investor Day, and people got a glimpse into what we call Cloudflare OS product that, again, we developed ourselves that we are now starting to sell and the harness and AI infrastructure we build ourselves, so people do understand how we think about productivity and how the company would evolve. And we spend with 80 people tomorrow and the day after tomorrow to really rethink every process in the finance stack from an AI-first perspective, right? So we deeply believe in that.

And there were question marks when Matthew, last year, at the end of last year went out and said, everybody stop hiring fresh outs and grads, but we will hire 1,111, that's a product that's why this is a strange number. And that is what we did. I was at the Austin office the other day and there's 350 interns working on the craziest projects, nonbiased by any legacy or any process. They never worked in a billing process that works from completely free to reimagine how that process could look like in an AI-first world, or how demand generation could look like in that world. A lot of these interns will get job offers. So there is an approach to this where you just give credit Matthew and Michelle for rethinking things like that from a very different perspective. There were many. Our executive team would have, really? Now we hire 1,111 interns. How do we do that? And now we look back and said, we'll do that next year again. It was exactly the right thing to do, yes.

Hey, fantastic. Thomas, thank you for your time. Please join me in thanking Thomas.

Thank you.

Ladies and gentlemen, thank you for standing by, and welcome to Cloudflare's Second Quarter 2026 Earnings Conference Call. [Operator Instructions]

I would now like to turn the conference over to Phil Winslow. Phil, please go ahead.

Thank you for joining us today to discuss Cloudflare's financial results for the second quarter of 2026. With me on the call, we have Matthew Prince, Co-Founder and CEO; Michelle Zatlyn, Co-Founder and President; and Thomas Seifert, CFO. By now, everyone should have access to our earnings announcement. This announcement as well as our supplemental financial information may be found on our Investor Relations website.

As a reminder, we will be making forward-looking statements during today's discussion including, but not limited to, our customers, vendors and partners operations and future financial performance, are anticipating product launches and the timing and market potential of those products, our anticipated future financial and operating performance, and our expectations regarding future macroeconomic conditions. These statements and other comments are not guarantees of future performance and are subject to risks and uncertainty, much of which is beyond our control. Our actual results may differ significantly from those projected or suggested in any of our forward-looking statements. These forward-looking statements apply as of today, and you should not rely on them as representing our views in the future. We undertake no obligation to update these statements after this call.

For a more complete discussion of the risks and uncertainties that could impact our future operating results and financial condition, please see our filings with the SEC as well as in today's earnings press release. Unless otherwise noted, all financial numbers we talk about today other than revenue will be on an adjusted non-GAAP basis. You may find a reconciliation of GAAP to non-GAAP financial measures that are included in our earnings release on our Investor Relations website. For historical periods, a GAAP to non-GAAP reconciliation can be found in the supplemental financial information referenced a few moments ago. We would also like to inform you that we will be participating in Stifel's Tech Executive Summit on August 24 and Goldman Sachs Communacopia and Technology Conference on September 9.

Now with that, I'd like to turn the call over to Matthew.

Thank you, Phil. We had an extremely strong second quarter. We achieved revenue of $696.1 million, up 36% year-over-year. We now have 4,698 customers paying us more than $100,000 per year, a 27% increase year-over-year. We added a record number of these large customers, 986 net additions year-over-year, the most we have ever added over 12 months in our history. Our dollar-based net retention was 120%, up 2% quarter-over-quarter and up 6% year-over-year. Our gross margin was 73.1%, improving sequentially for the first time in 8 quarters. We delivered operating profit of $96.1 million, representing a non-GAAP operating margin of 13.8% and we generated free cash flow of $56.4 million during the quarter, up 69% year-over-year.

The strong momentum we've been seeing for some time in our business continue to build in the second quarter, some highlights. Sales productivity increased year-over-year for the tenth consecutive quarter. New customer bookings increased at the fastest rate in more than 5 years. New pipeline generation continues to accelerate, again, growing sequentially at its fastest pace in 5 years. We added more than 80,000 paying customers this quarter, driving 74% year-over-year paying customer growth.

For some context, we added as many paying customers this quarter as we had in total when we went public in Q3 of 2019. We had nearly 1,000 large customers year-over-year. We also set a record for year-over-year net additions across every large customer tier from $100,000 to over $5 million in annualized revenue. And we ended the quarter with more than 7.4 million developers on our platform. That's nearly 2 million developers in Q2 alone, surpassing the 1.5 million we added during all of 2025.

It's clear that the agentic future needs a new kind of cloud. Developers are flocking to Cloudflare because our workers' developer platform gives them what they need to build that agentic future. We're the fastest or the most secure and the most cost-effective place to build, deploy and scale agents and the code they generate.

At Cloudflare, what we continue to prove is that when you build a company the right way, you can grow and innovate while executing and profiting. We have the network. We have the platform. We have the culture of innovation. We have the business momentum. We have the disciplined execution. We're in the right moment in time, and we are keeping our foot firmly on the gas. This is the winning formula that we know works. It's the same one that has propelled Cloudflare into the pole position to lead the next phase of the Internet in the age of agentic AI. That's a good segue to discuss some of our wins in the quarter.

A leading digital native media company expanded their relationship with Cloudflare signing a 5-year, $31.8 million contract for Application Services and Zero Trust. To combat aggressive scraping and accelerate global performance, this customer chose Cloudflare for our best-in-breed edge capabilities and operational velocity. Despite competitive pressure to consolidate spend with their incumbent hyperscaler, this customer's long-term commitment is a proof point that when performance and security are nonnegotiable, enterprises choose Cloudflare's unified platform.

A global 2,000 European technology company expanded their relationship with Cloudflare signing a 3-year, $11 million contract for Application Services and Zero Trust with our developer platform seeded for future AI workloads. After years of acquisitions resulted in a fragmented IT footprint, this customer chose Cloudflare to eliminate a stack of 5 incumbent legacy point solutions with up to 7 targeted on their long-term road map in favor of our single unified platform as the foundation for their entire organization to build on.

A large U.S. federal agency expanded their relationship with Cloudflare signing a 5-year, $7.7 million contract for Magic Transit and network firewall. After our legacy providers outage locked over 100,000 users out of a mission-critical system for days, the agency urgently sought greater resilience and real-time control. Cloudflare slashed the customer's global rule change time from a 1-week SLA down to just 30 seconds. They've also cut hardware costs by blocking unwanted traffic at the edge and have already shut down an entire data center. As an existing application security customer already, this agency can now run their network and application security on one unified platform, cementing Cloudflare as the front door for all their Internet traffic.

A rapidly growing generative AI company signed a 1-year $7.5 million pool of funds contracts for our developer platform. This customer's workloads pull an enormous volume of images and video. At that scale, a hyperscaler's egress tax would break the economics and create vendor lock-in, limiting their choice of inference tools and GPUs. Their engineering team evaluated multiple providers and chose Cloudflare as the only 1 that pairs a 0 egress model with the reliability, scale and comprehensive capabilities of an enterprise-grade platform. By structuring this as a pool of funds deal, the customer can solve their immediate storage needs while retaining the flexibility to expand across our entire developer platform.

A rapidly growing technology company in APAC expanded their relationship with Cloudflare, signing a 1-year $4 million pool of funds contract for our workers' developer platform. This deal accelerates a powerful partnership, building on an $8.7 million application services contract signed just last quarter. In only 1 year, this customer has standardized on Cloudflare end-to-end from application security and Zero Trust to now our developer platform, directing every request through a Cloudflare worker and using KV and durable objects as the routing and tenant configuration layer for their entire platform. They chose Cloudflare over their incumbent hyperscaler to avoid added latency, proving the flywheel of our unified offering.

Once performance and security run on Cloudflare, our developer platform becomes a natural foundation for the next layer of any company's stack.

A Fortune 100 technology company expanded their relationship with Cloudflare signing a 3-year $5.2 million contract for our full SASE portfolio. This customer is replacing legacy VPNs and virtual desktops to move their entire global workforce onto a single Zero Trust platform. In a competitive evaluation, Cloudflare beat 2 first-generation Zero Trust vendors winning due to our faster network performance and our single pane of glass management that is so easy the customer expects to run our services with roughly 1/3 of the stats. This is exactly the type of security consolidation we see accelerating. Enterprise is retiring fragmented point solutions in favor of Cloudflare's blazing fast, easy-to-use and unified platform.

A Fortune 1000 technology company expanded their relationship with Cloudflare, signing an 18-month $15.9 million contract for application services in our workers' developer platform. This customer serves hundreds of thousands of businesses, which requires an architecture that can act as their global front door for security and performance without adding latency. By standardizing on Cloudflare over legacy alternatives, they eliminated multiproduct complexity and secured long-term operational predictability as they build an AI-first customer platform.

A leading technology company expanded their relationship with Cloudflare signing a 1-year $6 million pool of funds contract for our Workers' Developer platform. As this customer scales their new AI agent capabilities, they needed an elastic secure container infrastructure that could scale with their rapid growth and ship new capabilities in weeks, not quarters. They chose to build on Cloudflare over legacy hyperscalers and point solution competitors because of our built-in threat intelligence that passively prevents compute abuse, rapid pace of innovation and the ability to deliver FedRAMP compliance. This win also shows how the most sophisticated AI builders are increasingly selecting Cloudflare as the agent cloud of the future.

These customers and others like them are increasingly choosing Cloudflare because we don't just keep up with change, we drive it. We ship fast, we solve hard problems, and we do it in a way that meaningfully raises the bar for the entire Internet. Nowhere is this relentless focus on innovation more exciting than the work we are doing in AI. For the first time in human history, in Q2, more than 50% of the traffic flowing across Cloudflare's network was not human. The number of requests on our network from AI agents continues to grow unabated. With the web shifting from human-driven browsing to AI answer engines and agent-driven commerce, we are witnessing a fundamental rewrite of the Internet for machine-to-machine traffic. Cloudflare is positioned at the center of this paradigm shift building the scalable infrastructure that controls the developer tools and the payment rails to power the agentic Internet.

Just since the start of Q3, we've celebrated innovation in this area from content Independence Day to agent week, which is happening right now. During these, we unveiled the key building blocks for a 2-sided agentic marketplace. Monetization gateway allows our customers to sell any resource behind Cloudflare, whether it's a web page, an API, a data set or an MCP tool. This will empower new business models that will define the next generation of the Internet. In addition, we announced wallets, which will offer a way for buyers to pay autonomously through their agents and cloudflare.pay, which will provide merchants and buyers an agent-friendly means to identify themselves and establish trust.

Not only are we building a foundational elements for agentic commerce to succeed, we also believe AI companies and content owners should thrive together. That's why we recently announced a first-of-its-kind research pilot with open AI that we believe may help pave the way to a sustainable ecosystem of content creators and AI companies. Over the coming months, we'll announce more ways that AI companies, content creators and businesses large and small, can thrive together. The business model of the Internet is changing, and there is no company better positioned to define its future and cloud flare. It's an incredibly exciting time.

That seems like a good start to turn it over to Thomas to talk about the financials. Thomas, take it away.

Thank you, Matthew, and thank you to everyone for joining us. We delivered a stellar second quarter with strength across all major metrics we track, driven in particular by another quarter of rapid growth in our Workers' Developer platform and agentic workloads across our network, continued momentum with our largest customer cohorts and a robust go-to-market execution.

Turning to revenue. Total revenue for the second quarter increased 36% year-over-year to $696.1 million. From a geographic perspective, the U.S. represented 51% of revenue and increased 41% year-over-year. EMEA represented 27% of revenue and increased 30% year-over-year. APAC represented 14% of revenue and increased 32% year-over-year.

Turning to our customer metrics. We ended the quarter with 4,698 large customers, those spending more than $100,000 annually with us, representing an increase of 27% year-over-year and an acceleration from 25% growth last quarter. We added 282 large customers in the quarter and a record 986 large customers year-over-year, the most we have ever added year-over-year in our history. In fact, every one of our large customer cohorts from $100,000 to $500,000 to $1 million to $5 million in annualized revenue added a record number of net new customers year-over-year in the second quarter.

Revenue contribution from large customers was 73% of revenue during the quarter, up from 71% in the second quarter last year. The significant expansion with our largest customers drove an acceleration in our dollar-based net retention rate to 120% in the second quarter, up 2% sequentially and up 6% year-over-year.

Moving to gross margin. Second quarter gross margin was 73.1%, representing an increase of 30 basis points sequentially and a decrease of 320 basis points year-over-year. Paid versus free traffic on our network continues to grow year-over-year again, driving additional allocation of network costs from sales and marketing into cost of revenue. However, as we discussed at Investor Day, this trend is showing signs of beginning to stabilize.

Network CapEx represented 7% of revenue in the second quarter. As a reminder, there can be some variability in this metric quarter-to-quarter, and we expect network CapEx to be 14% to 15% of revenue for full year 2026.

Turning to operating expenses. Second quarter operating expenses as a percentage of revenue decreased by 3% year-over-year to 59%. Our total head count ended the quarter at 4,700.

Sales and marketing expenses were $232.5 million for the quarter. Sales and marketing as a percentage of revenue decreased to 33% from 36% in the same quarter last year. Research and development expenses were $104.1 million in the quarter. R&D as a percentage of revenue decreased to 15% from 16% in the same quarter last year. General and administrative expenses were $76.3 million for the quarter. G&A as a percentage of revenue increased to 11% from 10% in the same quarter last year.

Operating income was $96.1 million, an increase of 33% year-over-year compared to $72.3 million in the same period last year. Second quarter operating margin was 13.8%, an increase of 240 basis points sequentially and a decrease of 30 basis points year-over-year.

Turning to net income and the balance sheet. Our net income in the quarter was $107.8 million or diluted net income per share of $0.29. Excluded from these non-GAAP results for severance and other restructuring charges of $151 million for the second quarter, of which $99 million was paid in the second quarter. For full year 2026, we now expect severance and other restructuring charges of up to $165 million with up to $130 million expected to be cash related. While higher than initially anticipated, we prioritized speed of resolution, particularly internationally to ensure a continued focus on execution.

Free cash flow was $56.4 million in the quarter or 8% of revenue, compared to $33.3 million or 6% of revenue in the same period last year. Excluding the impact from the higher severance and other restructuring costs, our free cash flow expectations for 2026 remain unchanged. We ended the second quarter with $4.2 billion in cash, cash equivalents and available-for-sale securities. Remaining performance obligations, or RPO, came in at [ $2.732 billion, ] representing an increase of 7% sequentially and 38% year-over-year. Current RPO or 64% of total RPO and grew 35% year-over-year.

Moving to guidance for the third quarter and full year 2026. For the third quarter, we expect revenue in the range of $736 million to $737 million, representing an increase of 31% year-over-year. We expect operating income in the range of $129 million to $130 million. We expect an effective tax rate of 20%. We expect diluted net income per share of $0.34, assuming approximately 374 million shares outstanding. For the full year 2026, we expect revenue in the range of $2.864 billion to $2.870 billion, representing an increase of 32% year-over-year. We expect operating income for the full year in the range of $443 million to $445 million. We expect an effective tax rate of 20%. We expect diluted net income per share over that period to be $1.25 to $1.26 and we expect approximately 374 million shares outstanding.

In closing, the second quarter reflected the strength of our underlying business. Our strategic position leading the paradigm shift of the agentic Internet has never been stronger and the opportunity ahead of us is larger and more defined than at any point in our history. We remain committed to capturing it with disciplined execution, durable growth and long-term focus. And with that, operator, please poll for questions.

[Operator Instructions] Your first question comes from the line of Saket Kalia with Barclays.

Great to see the acceleration in the business. Matthew, maybe for you. I'd love to dig into the changing profile of traffic and how that's monetized. You've talked about how the majority of traffic now is nonhuman, and we all see the inflection in traffic. Maybe the question is, do you see that driving more new logos or are existing customers buying more? And then from a different related lens, is it driving more business with large customers or smaller ones?

Yes, Saket. I think it's a pretty wild time. I was asked in the end of 2025 -- in November 2025 when I thought that nonhuman traffic would pass human traffic. And I -- we pulled all the data, we ran all the numbers, and we were pretty confident that it was going to be the second half of 2027. I was asked the same question again in March of 2026. And we did the same exercise. And we're surprised to see that it had moved up to be -- that would cross in the first half of 2027. So I was quite surprised when in May, of this year, our team came to me and said, you won't believe it, but nonhuman traffic is now passed human traffic online.

And to give you a sense of how this trend is playing out, and with the big caveat that I have called it wrong at every point along the way, the -- if the current trends continue, we think, in 5 years, nonhuman traffic will be as much as 1,000 times as much as human traffic. In other words, humans will be a rounding error on the Internet, not because human traffic goes down, but that's just how fast we're seeing nonhuman traffic grow.

And so that's resulting in a number of things. The first is that for some of that nonhuman traffic, it's malicious. And that could be malicious like it's hackers or bad guys. It could also be -- it's malicious from the perspective of a particular customer's business model where it's traffic that is maybe an AI company trying to take the content from a media company that relies on advertising. And in those cases, we block that traffic, and we don't charge the customers anything more for blocking up that traffic because we think that, that's the right thing for us to be doing and delivering, and that's part of being a security company.

At the same time, though, there are some people who want that traffic. And so we're doing everything we can not only to serve that, but to make it as efficient as possible to serve it because if we're going to have 1,000x as much traffic online, we've got to get a lot more efficient and companies like Cloudflare are critical to be able to support that for customers, whether they're large and small.

The thing that I think is most interesting, though, is that increasingly, as we talked to the AI companies and over 80% of the major AI companies are Cloudflare customers and rely on us. And we have a great relationship with them. And as we talk to others, it's clear that as agents are accessing all of these sites and the volume that they're accessing them on to sort of give to get that you have with human traffic is different. And so if you look at some of the things that we've announced this week, during agents week, things like cloudflare.pay, that's us setting the foundation to be able to say how do we charge agents some -- again, what will be a very, very small fee fractions of a penny for every request that goes through but for the request that pass through that traffic because somebody has to pay for the bandwidth, somebody has to pay for the server. Somebody has to pay for the people doing the work to create the content.

And I think that the business model of the Internet for the last 27 years has been largely defined by advertising and really defined by Google. I think the business model of the next 27 years of the Internet is going to be very different, and there's no company in a better position to define what it looks like than Cloudflare.

Totally agree. Thomas, maybe for my follow-up for you. it was great to see the stabilization in gross margins this quarter. And we talked a bunch about that at Analyst Day. But maybe you could just talk to us a little bit about some of the puts and takes there. And particularly, how you see gross margin sort of ebbing and flowing through the second half?

Well, as we said at Investor Day, the important metric for us to focus on is total unit economics because the behavior, of course, margin across the various products is just very different. You can see with all the metrics that we published today that we are tracking ahead of everything we track. So we're quite confident that the total unit economics for the -- across all products is going to increase over the course of this year.

Gross margin, I think, will stabilize around the level we have. But as I said, more importantly, it is to focus on the unit economics. And here, we will continue to see expansion over the course of the second half of the year.

Your next question comes from the line of Matt Hedberg with RBC Capital Markets.

Congrats on the quarter, just stellar results here. I wanted to start -- you guys had -- I think, Matt, you said 2 million developers out of this quarter. That was obviously impressive. I guess, with so many options for developers and questions about open weight and open source models these days. What are the most important elements about workers that gives developers the confidence the platform on it, really, as you know, the Internet is fundamentally changing to support pushing to machine traffic?

And then maybe kind of a follow-up to that. There obviously been a lot of buzz about [indiscernible], and we saw the open source this week, that's great. When we think beyond the developer, talk about how [indiscernible] opens up the opportunity to really every employee in an organization?

Yes. I was -- I made the team triple check the developer numbers because we added more developers in a quarter than we did in all of last year. And we thought last year was pretty good. I think it's being driven by a handful of different things. The first is that Cloudflare Workers is turning out to just be the perfect platform for building agents and agentic workloads. It's extremely lightweight, it's -- you only get charged for when it's actually doing work. You can spin things up and spin them down very, very quickly. And so it has become the go-to place for sophisticated developers to be able to launch code.

And we're pretty conservative how we can -- there are big companies that have 1 developer account, even though there's lots of people that are working behind it. But across the board, and you can see this in part by the paying customer count. There's just lots of small developers that are driving more and more use of the platform. And again, I think that's great because those small developers, some of them will turn into nothing, but some of them will turn into major things as well. And that's really always been the key story of Cloudflare.

I think the other thing is it kind of dovetails into your second question, which is we're seeing that the number of developers generally is going up massively, where people who never thought of themselves as being able to write code or create things are doing exactly that. My [indiscernible] is using some of the vibe coding platforms in order to build tools in order to run things around her house and she's writing code and deploying it to Cloudflare. And that's really remarkable.

And so if you look at companies like Lovable and [indiscernible] and BACE 44 at Wix and others that a lot of times that where those -- that code is actually getting deployed and where the preferred target of that code is going is actually to Cloudflare. And again, I think that's driving more and more developers.

Cloudflare is -- we talked last quarter, and I think there was some head scratching in some corners about -- we said that we were just seeing so much more efficiency across our team, not just our developers but our finance team, our legal team, our procurement across management and how are we getting that? And the answer in large part was we've really built a set of tools with Cloudflare that allowed everybody on our team to be able to take advantage of some of the tools that I think in most organizations are somewhat limited to the developer platform. And we got enough questions about that, and we were proud enough about the work that we -- just the other day, open sourced that project. And already the reaction has been amazing to see how many large companies are saying, wow, this is exactly what we need.

I think the thing that's -- there's a lot of companies that are doing this, including a lot of start-ups that have really promising potential. But I think the thing that's unique about us is that because we started out as a security company, and I think describing Cloudflare just a security company today is obviously misunderstanding us. But we still have that DNA at our core. It allowed us to hook into things like our systems of record in a way that was very secure, auditable that we could actually sample again, so we put controls around that gave us the ability to allow people who are across the organization to have the access of these very powerful tools, but do it in a way where we felt it was still safe and responsible.

I think that's where most organizations are hitting kind of road blocks internally in terms of turning more AI tools loose across the organization. And I think one of the reasons that we've been able to move faster than others and one of the reasons that we're so excited about open sourcing Cloudflare OS, is because of the fact that at our foundation, we're a security company, we understand how to do these things securely, and that's allowing us to move fast without breaking things.

Your next question comes from the line of Sanjit Singh with Morgan Stanley.

I want to pick up on Matt's question. I think probably one of the other reasons why you're seeing such impressive developer traction that you guys have, one of the most thorough sort of age and stacks across infrastructure, run time, model serving, the data layer of durable objects, pretty impressive in the market. At what point that do you think that there's going to be like a security opportunity around not just running and orchestrating these agents, but securing those agents? And what is the security portfolio for agent orchestration agent themselves look like?

Yes, Sanjit. First of all, I mean, I think kudos to all of the team that's been working on the developer platform, especially the agents part. I mean, they've just done an absolute animal mode, really seeing more and more features and really building, first and foremost, the tools that we need ourselves, but the tools that we need ourselves is turning out the rest of the world needs as well.

I think the opportunity is already there. I mean, the #1 thing that's causing our phone to ring from big companies is them saying, listen, we know we have to do AI, but we need to do it more securely. And so it's been especially interesting in sort of the SaaS Zero Trust space. I think that if you listen to a lot of the Zero Trust vendors that are out there talking about human seats and things like that. I think we've sort of shortcutted a lot of those conversations by saying, like, do whatever you want with the humans in your org, but you're going to have even more agents and you've got to have a proper security model for those agents.

So I was just in London a few weeks ago meeting with a large government agency there that was well down the track with one of the sort of first-generation Zero Trust companies to implement that across a big chunk of the U.K. government. And we sort of started talking about agents and what their plan was and how they were thinking about it. And very quickly, it became clear that the vendor that they were considering really hadn't thought about this whereas it's been core because of the fact we have the developer platform to how we did things, and they literally canceled the RFP and are now reevaluating this with the sort of agents first approach.

I think you're going to see that more and more across organizations. And because of the fact that we come at this from a developer platform place, I think that's what's really allowed our SASE and Zero Trust platform to actually gain share pretty significantly in the last 6 months.

I'll be clear to see how that plays out. As a follow-up, Tom, on the Investor Day, you had some like great slide going through the different revenue models. And particularly with respect to the pool of funds, I think you had a chart there about some of the larger customers really early renewing on a pull of funds basis and some of the revenue timing and revenue headwinds associated with that. As we got into Q2 and as we look into the back half, do you think -- are we at a point where those customers have sort of leaned into usage and were sort of passed those initial revenue headwinds [indiscernible]?

Yes. The trajectory of the business is clearly up and to the right for both revenue as well as operating income. But as we said at Investor Day, their business model is also going to evolve. It's moving away from a purely ratable SaaS model towards a much more diversified mix of pool of funds, consumption-based structures and also what we call [indiscernible]. As the business accelerates, that also leads to more and more customers burning through their [ T-shirt sizes ] faster, pool of funds are getting consumed faster and getting renewed. So we are really stacking these things on top of each other. And that leads to you would call it a higher standard deviation across the individual customers.

So the momentum over a multiple quarter is clearly pointing up into the right. You not only see this in revenue. You also see this in RPO -- but moving -- how we move from quarter-to-quarter becomes a little bit more difficult to forecast. That's why we continue to be prudent in terms of how we set the guidance. But the trajectory is clear is clearly up into the right.

Your next question comes from the line of Gabriela Borges with Goldman Sachs.

Thomas, your comment now on the momentum in the business are very clear, and certainly, we can see it in the acceleration over the last several quarters. My question for your higher level is [indiscernible] you'll have been in this very consistent growth rate for the past 3 years now in the high 20s, low 30s. And I think about the trends that are happening in your business, is there a scenario where the structural growth rate of Cloudflare actually ends up being higher over the next 3 years? How is all of these charts come together?

I'll take a step on it and then Matthew can jump in. A lot of factors, I think, play to that. The one for sure is the innovation flywheel and that we address and disrupt a TAM that becomes bigger and bigger from $30 billion at our IPO to north of $300 million at this point in time. I think the flywheels that come with X3 and now especially with X4 are accelerating every time we move to a new act, the maximum deal size goes up by factors, maximum $1 million contracts [indiscernible] double-digit million dollars [indiscernible] triple digits in the X3. So it's a combination of a lot of factors.

And then we clearly said this today, the larger the customer cohort, the faster the growth rate. So I think -- there are a lot of independent growth vectors coming -- pointing all in the right direction and then generate the momentum that you see in the numbers.

Gabriela, the only thing I think I would add is that there's probably some downside to having a CEO who is the founder of the company, co-founded the company. But one of the -- maybe downsize or upside as you look at it is that I remember when our growth rates were a lot higher and I missed those days. And so we'll be upset if we get back to even higher growth rates. .

Yes. My follow-up is on the triple play around monetization gateway wallets an ID that you announced intra quota. Talk a little bit about why this is hard to do and how you think about the moat here? Who do you think the competition will be? And I remember with that for your comments role is look, we need to get to critical mass. We need enough of the industry to accept this as a status. How do you think about the adoption curve for the monetization gateway while it's IP triple play?

Sure. I love the question because it's not -- like we don't -- at Cloudflare think about how we create moats. We think about our mission as being how do we help build a better Internet. And that has turned out to be a way of creating a really durable business. But in this case, I think -- I like the other part of your question more, which is why is this hard. I think when we started to think about this, and we said the business model of the Internet is going to change dramatically. What does it change, too? So something like micro payments was going to be a part of it. And so then the question is, how do you pull that off?

And again, to give rough numbers, and these are not audited, but they're directionally kind of accurate. But we handle, say, about 0.5 billion requests per second through Cloudflare's network. We roughly estimate that somewhere between 1% and 10% of those, you could monetize through some sort of a micro transaction. Again you need to be a tiny fraction of pennies. But -- but that means that you day 1 on launching something like this, you need to be able to support, call it, $10 million transaction -- financial transactions per second, and be able to scale up to, call it, $100 million financial transactions presenting.

To give you some sense, Visa -- and again, these are from memory, but Visa, which is the largest payment network in the world at peak during the holidays, handles about 20,000 transactions per second. So you have to build something that's 3 orders of magnitude bigger than Visa in order to pull this off. That is -- we didn't actually set out to build that ourselves. We thought maybe we could partner with someone and maybe someone else could do it. But unfortunately, a lot of the world is building kind of these next-generation payments networks things that they're competing with Visa as opposed to things that they're trying to figure out how you help build a better Internet and how you ensure that there's a healthy and successful business model for the Internet going forward.

And so like as we do this, if we're able to do it, I think that the scale that we're able to operate at and how we do that efficiently is incredibly important. Not to mention the fact that today, north of 20% of the Internet already sits behind us. And if we can make it 1 click simple for them to turn this on. And 1 way to think of this is a lot of that is our free customers. What have we made it less than free? What is being part of Cloudflare, we actually sent you money for being part of us because we were generating that through a series of micro transactions largely serving agents. I think that, that then just continues to accelerate the flywheel across all of our business. And as I said, I have never been more excited for the future of Cloudflare. I've never been more excited for the future of the Internet. And the Internet business model is going to change, and there is no company better positioned to lead that change than we are.

Your next question comes from the line of Fatima Boolani with Citi.

Matthew or Thomas, I was hoping you could put a little bit of a quantitative framing on how big the book of business so to speak has become for you on workers and workers AI combined just so we can have a sense and feel for the magnitude of impact this is having on your top line in terms of proportion. And to the extent there are any particular capabilities or SKUs inside these portfolios that are driving the outsized monetization momentum for you.

And then on a related matter, there has been a multiyear focus on beefing up the enterprise go-to-market motion and having those seats at the table on a more consistent and more strategic basis. But I'm wondering with the center of gravity sort of shifting back into the hands of developers and everything that you're doing from an innovation standpoint to cater, to developers. Do we see a little bit of a reversion back to sort of more PLG, I'd love to get some of your perspectives on that.

Sure. So I will start out not fully answering your question and then Thomas can continue to not fully answer your question. I think we haven't broken out by product area, what our revenue or revenue growth rate is. But I can say that the Workers' platform continues to be something that really drives an enormous amount of both new customer adoption and real revenue growth.

I think for a while, we were saying that we were playing for adoption, we weren't playing for revenue. I think that we have -- that is tipped and we are now very much -- again, not trying to maximize every dollar by any means. But that Worker's has been kind of a meaningful contributor to revenue. I think it shows up in the anecdotes that I gave about customer wins. More and more of them are signing pool of funds deals because of the fact that Worker's is part of it and that their developer teams are saying, "Hey, yes, fine with Cloudflare, use their Zero Trust, use their reverse proxy services, but also make sure that we've got something available where we can use their Workers' platform. And so that tends to drive that and drive a lot of consumption through that as well.

I think -- so for sure -- and again, I think you see this in the paying customer count that the PLG kind of product like growth is definitely going great. But you're also seeing, if you just look at like -- you can do this across earnings calls, you add up we give basically the same number of customer win examples every time. It just added up the dollar amount of those contracts and figure out what the average is. Those have been ticking up and up and up. And I think that, that is still very much a function of a great enterprise sales team where people want -- if you're going to sign $10 million plus contracts, [indiscernible] putting on a credit card. Those are deals that humans negotiate with each other, and there needs to be a level of trust and relationship and acknowledge that if something goes wrong, someone is there to fix it. And that's what great enterprise sales leaders do.

What I think our enterprise sales team is doing is getting more technical. And again, even those people who were not -- who were great salespeople, but they weren't particularly technical, I think we're giving them the tools to become much more technical. And I think over time, I mean, what I've challenged our team and the sales team, and I think that I will -- challenge being met fairly quickly is who's going to be the first sales person who is in a meeting and while sitting in that meeting, delivers a feature -- before they even leave the meeting, it gets shipped to production. We're building the system to be able to do that. And I think our sales team has leaned in to say, yes, it's really important to still go build relationships, but we need to be technical. We need to be able to understand and talk to the technical leaders that are on the team. And at the end of the day, the best product is still what's going to win.

Your next question comes from the line of Jonathan Ho with William Blair.

Congratulations on the really strong results. At Investor Day, you set a goal of GAAP profitability by 2028 at the latest. Just based on the guidance that you gave for the second half, how are we tracking against that goal? And what are some of the dials that you have in terms of relative or incremental opportunities?

Yes. I think the keyword was latest by end of 2028. We are pacing ahead of that very clearly without the impact of the restructuring charges, the GAAP net loss would have been around $18 million. So that is almost striking distance. I think what is the levers we have and that we are pulling are, of course, the acceleration of top line, building in operating leverage on the automation, we talked about Cloudflare is helping right performance and discount analysis. So we're putting a lot of levers, but we are clearly ahead of our targets at this point. .

Perfect. And then is there a way that you could maybe rank order for us the relative strength in pipeline growth that you're seeing just based on the act or product families? Any color would be appreciated.

Well, I think what the visibility we gave at Investor Day continues to hold true. The worker products are still the fastest-growing act we have and across the complete product stack in features stack of 3 and followed by the SASE products [indiscernible]. Nothing has changed from Investor Day to this earnings call regarding the dynamics behind the business and how the pipeline builds.

Your next question comes from the line of Adam Borg with Stifel.

Maybe for Matthew, just on the channel opportunity. So partners reached 31% of revenue this quarter continues to track up to the right. And as I think about kind of the growing mix of Act 2, which should drive more channel usage or necessity relative to Act 3 to Act 4, which may not require as many partner touch point, how do we think about the partner involvement going forward? Have we kind of tapped out of this mix? Or is there still room for the partner contribution to go higher in coming quarters?

Yes, Adam, I think that there's definitely room for the partner mix to continue to go higher. I think that, that has been a real improvement across our business is how we've interacted with partners, how we've made sure that they can be successful, how we train them, how we've given them very clear rate cards, which was something that we were traditionally very bad at. And that, I think, we're seeing and when I talk to partners, they really appreciate how we've been a good partner to them and we've had them succeed alongside us.

I think the partner ecosystem is certainly changing. I think that the people who are winners yesterday may not be the winner tomorrow because AI and other tooling is going to really change how people deliver services. We have focused the partners very much on sort of our Act 2 products to SASE and Zero Trust and seen great success there, but then willing to attach other products to that. But [indiscernible] place where we definitely see partners winning.

We're also seeing some partners who are taking the developer platform and running with it. And so for [indiscernible], which we talked about before, we've had a lot of interest from large systems integrators and saying that they want to -- for people who don't want to deal with the open source version of this, they want to take that and run with it and implement it. And that's again one of those great partner opportunities because it's not just how do you get software installed, but then how do you get all of the various systems of records connected. How do you make sure that kind of sources of truth get incorporated. How do you make sure that all of that is put in place. And that's an incredible opportunity for partners. And we think that, that could drive very significant business because that, in turn, we expect in lots of cases, will drive more of our developer platform adoption.

And so I think I would guess that our partner -- our share of revenues through partners will continue to tick up over time. I don't think it'll ever get to sort of 90% plus of some of the first-generation Zero Trust companies or hardware companies have old. But I think it will continue to tick up over time. And I wouldn't be surprised if you get north of 50%, even maybe over 60%.

Your next question comes from the line of Ryan MacWilliams with William Blair.

Formerly with William Blair many years ago. But -- so maybe still early here, but as agents become more complex and require more calls to LLM over the course of what a agent is doing, does that make edge inferencing more important for AI aided use cases? And are you seeing more companies build with high AI performance, low latency requirements in mind?

Yes, Ryan, I think that it's -- the inference is one piece of a more complicated puzzle. And so what I think that we're seeing is what agents need is the ability to have very femoral low-cost places to create code, do inference, access the network, coalesce information, store some things and pack all of these things together. It's not simply the inference that matters. The real key is how do you orchestrate all of those pieces together. .

And so -- and in the ideal case, your agent doesn't run just in one place. It runs in many places. It might -- it might be operating in literally different places around the world where I have to access information, you have to get different things. So you want effectively for the entire network to -- as the old [indiscernible] become the computer, the network is the computer, [indiscernible] when Oracle bought Sun, they failed to register the trademark. They reregistered the trademark of the network as a computer. And so now Cloudflare runs that trademark, which I think is appropriate. We've never really used it for much, but it's a fun sort of thing.

I think that's exactly what agents see -- they need the network to be the computer. And so yes, inference needs to exist at -- and again, I've always shied away from the term edge because I kind of think it's meaningless. [indiscernible] edge? Or are we the core? Or are we everything? I think what we are is we are wherever you need storage, wherever you need compute, whether you need inference, whether you need network, we're there and are able to orchestrate that across the global network that we have.

And so that's what we're seeing agents need. And I think that in addition to the network, the fact that we've got these primitives where if back when I was in college, if you wanted to stand up a new service, you have to buy a physically new server then VMware came along and was able to take 1 server and make it to many. That was still pretty heavy weight. And so you had folks like Docker that came along that made containers. That really drove the first generation of the cloud. The agents need a new generation of the cloud. And that new generation of cloud, even things like containers are too heavy. As Rita on our team walked through at Investor Day, if every knowledge worker on earth ran an agent that was running in a container, we don't have enough CPU to actually power that. We need to increase the amount of CPU that's accessed by many orders -- many, many times.

And so what we've done, which is a new version, which is a much lighter weight in terms of our sandboxing technology, which we called isolates. That is -- that gives you the ability and the scale to actually deploy and run code in a way that is going to keep up with the demands that agents have. And so I think, yes, it's the network. Yes, it's the fact that, that network is everywhere and whether you want to call out the edge or whatever, it is what it is. But it's also that we've got the underlying platform that allows us to scale for what's coming, which is going to just be an incredible increase in the use of the Internet generally.

Yes, I thought your agent development life cycle release this week was pretty cool, like giving agents everything to build other agents [indiscernible] more finance question. Along looks like the big deals in the quarter and some new workers' wins, along with some ramping pool-funds deals. Just any help that we can get on modeling net retention or thoughts there for the rest of this year as implied by the guide?

What I would say is that the performance was really driven across the board. There's not really one thing to point out. This was an impressive quarter in many ways. But what stood out most for me was that the strength came across all products, across all customer cohorts, across all types of contract interactions we have with our customers. So there's not really one thing to point out. It was strength across the full spectrum and plus of what we are doing.

We have time for one more question, and that question comes from the line of Patrick Colville with Scotiabank.

I guess, look, I got the last question, I might ask a 2-parter, please. Matthew, can I ask you, I mean it's clear that the Cloudflare engine is humming with multiple cylinders firing. One of the decisions that you've made is not to participate in the AI CapEx kind of arms race. We're seeing hyperscalers and Neo Cloud sign these kind of massive GPU AI infrastructure contracts, including tonight when your competitors disclosing an aggregate nearly $3 billion AI infrastructure contract value. I guess, Matthew, what was your thinking kind of not joining this AI infrastructure arms race?

And then Thomas, if I may, I mean, very clear from Matthew's comments throughout this call that the cyber risk is rising. Can I ask you, Thomas. I mean is that rising cyber risk hitting the Cloudflare financial model in 2Q? Or is it kind of still conversations in 2Q, and this is more of a kind of back half 2026 thing?

Sure, Patrick. So I think not all revenue is created equal. And if you're selling what is just commodity compute, if you're basically letting an AI company user balance sheet, and your credit rating in order to buy servers that are the same as everybody else's servers, then like that's just not an attractive business for us. And so I think I think we're very disciplined. [indiscernible] we don't get asked. But I don't think that's particularly interesting for us. I will say that we have a lot of interaction with the largest AI providers that are out there. And they're -- again, they're all customers of us.

And so I think that it's -- that you could imagine ways where -- that we would find something that where we could actually be a bigger and bigger player in providing the services to power more and more of the serious AI players that are out there in more meaningful and deep ways. But ultimately, like we're not going to do that just selling commodity hardware because that, again, is bad business for us.

I think the other thing that's different about us is that we've just always focused on how we can squeeze as much out of every CapEx dollar as possible. And so how is it that we can -- not just throw money at a problem, but actually say, how do we get more utilization out of that. And so our team is constantly trying to figure out how can you get more inference, how can you get more CPU, how can you get more out of the memory or storage that we have in every box that we have. And I think that, that's just a very different attitude than you see elsewhere.

And so again, we know what the typical kind of inference loads are at the hyperscalers and the GPU utilization, and it's super low. And that's not the hyperscalers fall. It's that what they're selling is just a box and it's up to the customers in order to actually maximize the utilization and customers often don't have the diversity of traffic or the ability to really schedule things in order to get the highest possible utilization. And so if we can, and in some cases, get 10x as much utilization out of every CapEx dollar, then we can play these games and continue to grow revenue, but it's really high-quality revenue and it's revenue that doesn't require us to just throw money at it.

Again, I think it's to worth repeating things [indiscernible] before, which is we're just in a very different business than the hyperscalers. The hyperscalers, the traditional first-generation cloud are in the business of buying a server and then trying to sell it back, lease it back and get 5 turns of revenue off of it. We're in a very different business, where we're selling actually work getting done. And then instead of it being up to the customer to get as much as a server as possible, we need to do the work to get as much out of the underlying equipment as possible and build the scheduling and the efficiency and everything else that's out there.

And I think that, that is fundamentally the difference in our business and why -- like I don't think we want to get into the business of renting servers because it's -- over time, it's a commodity business, it's not very attractive.

Thank you. Matthew Prince, I would like to turn the conference back over to you.

I appreciate everyone at Cloudflare, just delivering what has been an exceptional quarter. We're hoping to find the future of agents. We're helping to find the future of the Internet. We're helping make sure that the Internet has a successful and sustainable business model. Some of the product work that I've done in my career, and I'm really proud of everyone and everything that they've done in order to make that happen. Thank you so much for being part of the earnings call today, and we'll see you back here again next quarter. .

Ladies and gentlemen, this does conclude today's conference call. Thank you for your participation, and you may now disconnect.

honored guests, please welcome your host, Phil Winslow.

All right thank you for coming this year. I know it's a busy time of the year, and it's not just because of the Knicks are in the finals. But obviously, there are a lot of things going on in the investment world this week, let alone just the volatility that's in the market. So we're going to make best use or for time with a very packed agenda here.

We're going to first start with a fireside chat with our CEO and Co-Founder, Matthew Prince, you'll then hear from Rita Kozlov and [ Sam Ray ] on our AI strategies, both for our customers, but also for Cloudflare. [ Stephanie Cowen ] is then going to take the stage to help everyone in this room better understand how Cloudflare is the only cloud built for the Agentic Internet. And then Mark Anderson will update everyone on our go-to-market strategies. And after that, Thomas Seifert, our CFO, will delve in our financials. And then as usual, we'll end the day with open Q&A with Matthew Michelle Zatlyn, Thomas and Mark.

And so with that said, I would like to invite our Co-Founder and CEO, Matthew Prince of the stage for yet another addition of our fireside chat in last year.

All right. So we're back this year. There was...

I think a lot more people showed up because the Knicks lost last night. Otherwise there will be a lot less tired eyes in the audience.

Exactly [indiscernible], but I appreciate you all came with that though. And so this year, as far as last year, there was a good fill in [ Evelfield ]. This year, I have a prop, so I brought receipts. This is actually the founders letter from 2019. Wow. And so that was the -- it was a very first founder letter that you and Michelle wrote. And I thought it was actually good to actually go back to that and start at a sort of a high level here. And I'm going to quote something here...

I tell Phil not to tell me what the questions are in advance, and so this is look I'm surprised on my face.

It is real. So we're all going to experience...

You did tell me there was going to be a prop, but he didn't tell me what.

I would probably be, so I gave a little hint that there was some prop, but I thought this line was particularly interesting. It's still very relevant today.

Many great startups pivot over time. We have not. We had a plan and been very purposeful in executing against it since our very earliest days. we are helping to build a better Internet. That was an audacious goal, and it changed both our business model and our technical architecture in ways that we believe differentiate us and provide us with significant competitive advantages. I think that still holds today.

And so my question is, let's start at a 30,000-foot view, and we're going to zoom in to questions after that. My question to you is thinking back to that what you wrote that in 2019, what makes Cloudflare different? And how has that changed? What makes us unique? How do the little pieces fit together? The small decisions that were made along the way, interconnect to make the whole greater than some of the parts.

So I think there are some parts of that, that I think are still very true to us. I'll close with saying, though, there's a little bit of that, that I'm not sure totally holds true today. And so let me talk about what is true and then what maybe we're actually starting to think about a little bit. So the first is I think from the beginning, Michelle and I were both students of [ Clay Christensen ], and we really did believe that whoever had the lowest cost to serve customers would win over time. And so we focused from the beginning on how we could just drive the cost of delivering our service down as much as possible. How could we buy equipment for as inexpensively as possible? How could we put it on the network as cheaply as possible? How could we actually deliver bandwidth and actually drive the cost bandwidth to as close to as possible? And get those advantages of scale.

And I think that so ingrained in the ethos of Cloudflare. And it's been super important for us over time. But what's really interesting about it today is when humans make decisions on what service they're going to buy or build, a lot of times, it's, oh, it's what I'm used to, or oh, it's who has the best advertising or what brand makes the most sense. What we're seeing is that as agents make those decisions actually make those decisions on much more sort of foundational rational facts. Just because you have fancy airport advertising doesn't necessarily influence the agent to choose. They choose based on actually what's the cheapest, what's the fastest, what's proven to be the best.

And I think that that's been really interesting. We're seeing OpenAI built their latest new platform, which they launched last week. On top of us. We partner with Anthropic on this, we're seeing more and more of the sort of Agentic web choosing us because of the fact that we've actually built these fundamental cost advantages where we can serve request just so much less -- so much more efficiently than anyone else.

The one thing I will say that was interesting, we had a Board meeting about a month ago. And [ John Grandin ], who used to be our CTO and now is on our Board, was sitting there. And we were presenting some of the stuff that actually [ Sam ] is going to show you later today. And he sort of looked -- and he was clearly -- he was thinking about something. And he said, "I wonder if our mission is too small". And it was the first time I was like, oh, it's really interesting, where it always felt audacious to say we're helping build a better Internet, but it's increasingly more than that. And that's actually something that I think we're internally trying to figure out that that's been so core to who we are, but is it actually something that's even bigger going forward.

Great. Well, let's continue from here because what you just described is sort of the beginning, obviously, that starts, we call it, connecting and protecting origins, the properties. But then it became the developer platform where we building these agents, building the applications and then sort of they call it the bot side of an internet interaction, then obviously connected with the most interconnected network in the world. So let's drill into those three, but let's work backwards.

Starting first with act 3, and so I'm going to rewind back in 2017 because you put out a blog post about this new product called Workers. And I was super excited about that. I was like, this is the future. This is the archetype of the next thing. I remember I had a call with you, [ John Gram Cummings, Dan Kinneg and Thomas ], because I was so excited about it. I remember at the end, you said, but Phil, Remember, a developer platform take 10 years. And as usual, you're right because we're now 9 years into end of this and when you said something seems to be changing.

I mean you just -- you can look at -- again, we shared some stats publicly, but there's also things that you can see from like [ NPM ] downloads and all things, we're just hit this hockey stick inflection where, again, I think because -- and I wish I could say that we were so prescient to see what the future was going to look like. We were always just building what made sense for us. And that's always been the key to what Cloudflare does is we were like. If we could start with the blank slate of paper, how would developers want to build. How would we -- our own team want to build these things. And I think that it turns out when we build things for ourselves, and then we show them to customers that's really powerful.

So for example, Sam's going to show you what we've done with Cloudflare OS later today. We didn't intend to be a product. but we keep showing it to people like what are you guys doing around AI and we show it to customers and they're like, that's exactly what I want. And so again, I think it's over and over again that we're building what are kind of scratching the itches that we have internally, and those turn out to be because we sort of think about these things from first principles, those tend not to be the really powerful tools that everybody else is going to want for that.

And again, there's this flywheel, which is happening now where as you make a platform which is cheaper, was faster than anyone else is. You have a lower cost of serving everyone that's causing more and more people to adopt it, which is getting more examples online more things are training against, and that's just accelerating over time. And again, I wish I could say that it was because we were so smart I think it's actually a we have a team that's always saying, if we could do it the right way from the beginning, what would we do? And now more and more people are saying, that's an opportunity where we can build on top of it.

Now let's also talk about security because obviously, if you think about how we began, business plan was called Project Web Wall. I said one of your friends said, actually, I think what you're trying to build is firewall in the cloud, you should call it Cloudflare. And so let's talk about security because that's obviously been a huge topic this year with Mythos [ PD 55 ]. Just almost equally as dominant is talking about AI and the agents in the news on. People are acting how do we protect when attackers have -- or be able to access these latest models. So my question is, how do you think AI impact in the security industry, both near term and long term but also what's cloud players role in this new security paradigm?

So I think for the next 2 years, every security company is going to be wildly busy. We're going to have a sort of [ log 4J ] like vulnerability, which was probably one of the worst kind of vulnerabilities that was out there. And for the last 10 years, we're going to have about one of those at least a month, maybe one of those a week. And that's going to keep the [ Crowd Strikes, the Sentinel Ones, the Cloudflare, the Zscaler ] incredibly busy over that period of time. But I think that what's going to change is that going forward, software is going to get a lot more secure. I think there's going to be a flip probably about 2 years from now, where all of a sudden, it's sort of like, huh, it's gotten much harder to find where those bugs are.

And I think there's still going to be an incredible role for sort of network security and how you're going to sort of partition data, like there's always going to be a purpose for that. But I think that security as a stand-alone category, if you're just a security company, will be really exciting for 2 years and then kind of maybe fades away after that.

And so I think that, that's -- again, I think we're going to be incredibly busy around that time, but it's going to -- that's going to just become table stakes as part of the platform over time. What's been really rewarding has been working with the major model vendors to be able to share they see. And we get early feeds on, hey, we found this new vulnerability.

And what's powerful and the reason that they partner with us in a very special way is because when -- because so much of the Internet sits behind us, very quietly effectively patch the vulnerabilities that they find and keep the Internet safer. And as you see these model companies who, I think, genuinely do worry Anthropic, we've had access [ Smiths ] from the beginning and Anthropic, just today kind of released a version of it, which has a lot more guardrails than the version that was sort of in limited release. But they really do want to make sure they're not making the world worse. And we're an incredible partner to them the fact that we can say, if you do find something, we can have a very effective way of protecting as much of the Internet as possible before everyone can go out and pass the software.

And again, I think that, that's one of the reasons why our customers trust us to stay in front of these threats and the folks that are finding these new vulnerabilities, including these new model companies are such a great partner set.

That's great. Now let's go back to the founder's letter here. One of the things we wrote is we never wanted to anything to discourage any potential customer from routing any amount of traffic, large or small to our network. So there is a question I get a lot, so I'm going to ask you about this too. Why is the free tier valuable to Cloudflare? And has this answer changed over the years? Because with the rise of AI and the opportunity to help build the new business model of the Internet.

You can actually -- so when Michelle and I started Cloudflare, it started as a school project at HBS. And the business plan is on file at the library at Harvard. So you can go out and actually read it and verify whether this is the case. And what we thought from the beginning was that some part of the business was going to be something that looked ad related, where we were going to do something where we would present ads to either on error pages or if you had a 404, a page not found page, we would sort of give you a ways of hints of getting around or we do other things like that. I never quite worked because we never could figure out how we could do something that was significantly better than what Google was offering for it.

And so that was always a piece of it, but the theory of it was always if you got enough of the Internet, passing through, there must be some way to actually generate some value to the people who are using that and be actually less than free, actually return capital back to -- return money of revenue back to them in order to encourage that.

What's interesting is it's starting to become true in a bunch of different ways. So some of the things that we're playing with are like no matter what the business model of the Internet is going to change over the course of the next 5 years. Like bots don't click on ads. So like ads are going to change.

So imagine if you had the ability to say, "I know a bot is coming to a page, and we're not going to show traditional human ads but we'll take those spaces off the page and will let people experiment with actually putting context on the page." We know what the bot is looking for. So we'll say, "Hey, by the way, I know that you're looking shopping for this", maybe there are some places out there. So imagine sort of a new version of like a Google AdSense but targeting bots.

Like again, I don't know if that's going to work. But we're in an incredibly unique position to be able to experiment that and play with it.

I mean the other thing which I think some version of this has to happen in the future, which is like as these agents are exploding in volume and to give you some sense, I thought that -- I mean, from everything we saw, we thought that the end of 2027 bot traffic, agents and everything else would exceed human traffic. We updated that about 3 months ago, bringing it in saying in the first half of 2027, it was going to exceed it. It's been exploding at such a rate just a few months ago, we actually pulled it hit again and it turns out bots have already exceeded human traffic online and it's growing exponentially.

Someone has to pay for that infrastructure. The system that's paid for in the past has been advertising based largely, that's going to change. It's going to go away. It's going to become something different. And so someone has to pay for that. And I think that what we're thinking about is we see about 0.5 billion transactions through our network on any given month. We think that somewhere between 1% and 10% of those, you can have some sort of micro payment to actually transfer something to that. And so we're trying to figure out how can we build the rails to support that.

And what's interesting is all the AI companies are like, yes, that makes sense. Like that's something that seems actually pretty fair that when we access these pages, we'll pay a fraction of a penny in order to get that content, and that's a way that there's still a way to encourage contents out there.

Again, I think we're in a very unique position to be able to deliver that. And if that's the case, if that's going to be the future, of course, we want every bit of content behind us. Of course, we want to support that as much as possible. And the reason we can do that cost effectively is because from the beginning, we've always focused on how do we drive bandwidth cost to 0, how do we make it as efficient as possible to serve this? How do we build a routing engine where always for a free customer, have excess capacity somewhere in the network in order to deliver that. And all of that has allowed us to get into the position to be able to take advantage of whatever the next business model the internet is going to be.

All right. Let's talk about how we're positioning ourselves in the next thing. And going back to the founder's letter, you said we will continue to invest in R&D so long as it demonstrates a significant return. Our investment philosophy is around making small and inexpensive bets, quickly term rating those that don't work, increasing investment in the ones that do. So let's fast forward to the day now with AI. How has AI changed your view of the potential for the pace of innovation at Cloudflare?

I mean it's just been amazing to see how much it's increased the pace of innovation. Our best engineers have gotten massively more productive. We're hiring as many people to build products as many of the developers as we can because we see that their pace of innovation has just gotten so fast.

What's powerful though is it's not only gotten faster, but it's gotten significantly better. So like we measure -- again, every once a while, we have a big public incident. You all read about that. We have lots of little incidents all the time. We have 10 years of data on these. We trained a model on all of that information on all of those incidents and we put in place to check. Not only code release, but every time a configuration change goes out. So you go to our dashboard as a customer, you push a button, there's actually an agent that checks that it gets pushed out to the rest of the system. And it was dramatic.

[ Jeremy ], who runs our production engineering team. It was running along, and you can see here's the rate of incidents. It was all like this, and then it just falls off a cliff. And it's down like this. And what that's doing is it's making us more comfortable to actually move faster. And again, not break things as we do it, which is really -- which is this incredibly powerful thing.

So I think this won't just be Cloudflare. I think at Cloudflare, we're learning from -- we're always one of the fastest shipping companies that are out there. And I think now we see like what Anthropic is doing, what OpenAI is doing, we're like, "Oh, like we can be even better" and that's encouraging us. But at the same time, I think we and across the entire industry, and again, somewhat similar to what we talked about security, you're going to see just a massive increase in the quality of software over the next period of time, and that's going to just turn into better product and really, again, a flywheel, which will continue to maintain our advantage over time.

So speaking of products, you said earlier that some of our best products. If you think about our developer platforms. Actually started with us building solutions for ourselves, the problems that we face. So if you think about the pain point that we're solving right now internally, what do you think that next thing is that we're solving internally that customers face as well?

Yes. So I think that the -- that you can't overstate how much that's been important to us. Like we started with the idea of how do you put a firewall in the cloud, and we had a problem that we knew that only to make that a successful business, we needed to sell to like big banks and governments like that. But they would never trust us unless we had really just robust data, but we couldn't get data unless we had customers. So is this chicken-egg problem. And so being plucky businesses, we're like, "Oh, let's create a free version of the service, we'll get a bunch of data and we'll use that".

The problem is that all of a sudden, the people who signed up for the free version of service were really a bunch of hacker kids who are all attacking each other like crazy. And then a bunch of like civil society and humanitarian organizations around the world who are all getting attacked by nation states all the time. So all of a sudden, we started and we were like, we, we never thought we had to be like a DDoS mitigation service, but I guess we have to go build that because like [ Arbor ] and the folks at [ Prolexic ] at the time, wouldn't like take us on as a customer.

And then like we're hacker kids like almost stole our domains because the registry that we were using wasn't successful. It was security. So then we had to build our own registrar, like we started having distributed teams and we looked at the [ Zscaler ] and others of the world where just like we don't this performance isn't up to what our standards were so we built Zero Trust, the developer platform was all because we built it ourselves.

And so core to what we are for better and worse, and there's downside to this, too, is we're very kind of inward looking at problems, solving them the way makes sense. We're not driven by what [ Gartner ] tells us to build. We're really driven by how do we build what the right solution is. And that, again, I think has very well for us over time, but it accelerates even more in a world where you have these sort of emotion-less agents that are making decisions on what's actually the right answer because that's what we've built. Over and over again. And that's why we're seeing more wins in Zero Trust. That's why we're seeing more wins in developers why we're seeing more wins in even our earliest products.

I think what -- to your question though, customers are really trying to figure out, have to implement AI but I have two massive problems. One is this is terrifying from a security perspective. How do I do this in a secure and safe way. And then the second is I did it. I put this budget aside for it, and all of a sudden, I've blown through my entire budget overnight.

And so the two questions that we get from every CTO, every CIO is how are you doing this securely and how are you doing it cost effectively. And we have those exact same problems. The amazing thing though was we had all the tools to solve them. So because we had Gateway, we could actually put in place very strict rules on what data individual users and their agents had access to and all of a sudden -- and again, you'll see this later, all of a sudden, you were able to give these tools access to all of these different sort of systems of record, but do it in a way that you could contain and audit what was going on with them.

On the other side, again, because we're partnered with all of these different model providers, we run things ourselves, we partner in other cases. We actually build very sophisticated routing systems that said on this particular user's use case, this particular query, that's actually not that valuable steer it to one of our internal low-cost models that we can provide very inexpensively. On the other hand, if there's something that's really high stakes, maybe we'll add it to some things that's more expensive. And so as a result, we've been able to very much control what our AI costs are. And as you can see, that's the #1 thing that customers are asking for, and that's actually driving a huge amount of the adoption, especially in our Zero Trust portfolio, where it's turning out that AI is kind of the killer use case to drive Cloudflare's version of Zero Trust.

And when I listen to some of the other security companies that are out there, I kind of scratch my head saying like, why do they sound like they're kind of talking yesterday's game and not tomorrow's. And the team is like because they don't have a developer platform because they haven't actually experienced this pain the way that customers experience it. And I think that's giving us this just really unique perspective and advantage, which is going to drive a lot of business going forward.

That's great. All right. Last question here. In Cloudflare, we have tendency to say we live in future.

Just sometimes we're better.

Exactly. And so -- but we just celebrated obviously, our 15th birthday. Last September in Workers, as we said, turns 10 next year. When you think about Cloudflare 20th birthday when, let's say, when Workers turned 20 like what excites you most about our potential as a company when you stare into the future?

I can't emphasize enough how insane the -- what's about to happen is across the Internet. There's this giant system that's the Internet. We're going to see traffic across it grow. I mean in 5 years, it will probably be 10 to 100x the traffic volumes that we see today, just because that's the amount of additional load that these systems are putting on it. You're going to actually see, I think, a lot more application a lot more content creation.

For a really long time, the Internet kind of plateaued. And then in the last 18 months, it started growing sort of exponential rates again, it's growing at the same rate that it was growing kind of in the mid-2000s, which is extraordinary to see again. And it catching on.

And so I think as that happens, no matter what, the business model of the Internet is going to change. And so what I think is so exciting is that we've got a front row seat to figuring out what that new business model is. And I think that as we do, like I hope that we can do it in a way that catalyzes not just the people who are creating content and applications, all of those things to be more successful, but actually solve some of the things that we learned over the last 20 years of the Internet, weren't all that healthy. So I think the most interesting question of the next 5 years is going to be what's the future of the business model the Internet going to be.

And I think we're the company that everyone is looking to figure out what that is. And so that's -- I think that's a ton of responsibility. But it's something that our team takes incredibly seriously. I don't think that -- we've always said that our mission is to help better Internet. We don't think we can do it alone. So we're partnering with amazing companies, both the model manufacturers, a lot of the people in the financial services, people who are supporting small businesses in order to really think about what that healthy future looks like.

And again, I can't imagine any more interesting question than what is the future business model of the Internet going to be because it's not going to be advertising. It's not going to be subscriptions.

That's great. Well, that's the end of our fireside chat. That 25 minutes went super fast. But Matthew, we'll be back up here for open Q&A at the end. So thank you, Matthew.

All right. I have the honor of entering our next speaker, her third year being here, I was talking about AI. And so to the stage, I would like to welcome Rita Kozlov, Vice President of Product for Cloudflare's Developer Platform.

Hello, everyone. My name is Rita Kozlov, I lead product for Cloudflare's developer platform. I'll share a little fun fact with you. Just a couple of days ago, I got to celebrate my tenth anniversary at Cloudflare. And I'm being very sincere when I say that even though it's been an incredible 10 years, there hasn't been a time that's been more exciting to be at Cloudflare than right now.

Now I'd like to start by talking about Cloudflare's mission for developers because that's the reason that we're doing all of this. That's the reason that I get up out of bed in the morning. And it's to provide developers with a modern cloud platform that accelerates their velocity and enables them to build better end user experiences.

In other words, we want to help developers bring their ideas to life. And we want to make it as easy as possible from the first line of code that they write to the first million of users that they get and the millions to come after that.

The other thing that I like to start with is the big, bold nonconsensus bet that we made 9 years ago when we started on this journey of building a developer platform. The reason this was such a big bet is because every single cloud that had been built before us has been built on the foundation of virtual machines or containers. But we decided to take a different approach with the technology called isolates. And this came with a big trade-off. We knew that it meant that it would be harder to lift and shift applications onto Cloudflare.

But it also meant that if you were building something new, if you're building a greenfield application, we could actually make it really easy for developers and remove the burden that we saw come with the clouds where they still had to worry about infrastructure and regions and scaling and performance.

And when you're making a big decision like this, you always want to bet on the future, not on the past. And it's a good thing we did because the greenfield opportunity today is so much bigger than it's ever been. AI is making writing code incredibly cheap. And that means that there's a whole explosion happening of applications that are being built every single day. And the whole calculus around lifting and shifting applications is changing too because organizations are no longer stuck with these decisions that they made 10, 20, 30 years ago, they can actually choose to rebuild something and not have it be this ambitious project that never completes. These are real conversations that we're having with our customers right now.

Lastly, developers are feeling really inspired because there is this whole new use case for them to build around agents.

Now this is my third time up here, and I feel like this is your opportunity to hold me accountable to the things that I promised the predictions that I made a year ago. So let's see how some of them have played out.

The first thing that we talked about was how AI workloads were going to shift from training and inference to end-to-end automation or agents. And guess what? That's exactly what we're seeing unfolding right now with tools like [ Open Clock ], which is an open source consumer agents, growing so quickly in popularity that overnight, it surpassed [ React ], which has been the biggest front-end framework for the past 10 years. That's how much demand there is for agents.

The next thing that we talked about was how in the next 5 years, more code was going to be written in the entire history of software that came before it. And again, we're well on track to surpass that goal. Every day this year, there are more websites, more new iOS apps being built than the year before. There is so much code being pushed up to GitHub every day, that GitHub is barely able to keep up. And we're seeing the exact same thing from our users. With [ Wrangler ] downloads being up nearly 1,000% year-over-year.

Last but not least, and perhaps most importantly, we talked about how Cloudflare was going to become the go-to platform for developers to build and deploy these AI applications. Because Cloudflare could uniquely offer developers the best of cost and scalability, performance and developer experience. And again, you don't have to take my word for it. You can take the word of 5.5 million developers that are building on Cloudflare today.

And that's not all. Just a week ago, we announced the acquisition of a company called [ Void Zero ], the company behind [ BEAT ], which has been one of the fastest-growing developer tools that we've seen over the past couple of years. The reason it's so popular is because it powers the technology that underlies almost every framework out there with the exception of [ Next JS ]. And what's been incredible to see is that bet has been growing in popularity, Well, so has Cloudflare's plug-in that's been growing with it just recently representing as much as 10% of total [ BEAT ] downloads.

But okay. This is all a year ago. So now that hopefully, I've earned a bit of your trust, let's talk about what's going to happen next. And I'm going to start with making a small adjustment to something that I said a year ago. Well, yes, automation and agents are going to be the next phase of AI. What we're actually seeing happen is that this is occurring in two distinct phases.

First, the models have to get smart enough to be able to actually operate these agents and to end. And what's going to happen next is all of these agents are going to need an execution environment to operate in. So let's dive into that.

I'd like to start with a recap of how Agentic systems even work. Agentic systems are made up of three main components: the AI piece, this is the model or the brain of the operation that's going to come up with a plan, constantly reevaluate it to make sure that it's continuing to deliver on a goal. There's the workflow aspect of it that's going to make sure that the next steps are actually being taken every single time. And there are the APIs. These are the tools that the agents have at their disposal so that they're not just making plans, they can actually [ action ] on them.

Let's start with the first component, AI, which is where we've seen so much improvement happen over the past year and really even over the past 6 months alone, since November, December, we've seen a massive step function improvement in the productivity that models have been able to give to agents. With models like [ OPIS 4.6, GPT 5.5 ], representing a massive unlock in agent productivity. And what's been remarkable is that open source models have actually been very quick to follow here with models like [ Kimi 2.6 and GLM 5.1 ] being just behind them.

The good news here for Cloudflare customers is that they can always get access to the latest models. And this is especially important when new models are coming out so quickly, that you don't want to be locked into a single vendor.

The Cloudflare AI Gateway, developers are able to access hundreds of models in a single place. And if you're choosing to run one of the open source models like [ Kimi ], you can do so on Workers AI in a truly serverless way where you don't have to pay for pre-provisioned resources that might go and use you can actually only pay for the inference that you end up using.

But now that that's behind us, let's talk about the more interesting piece, which is the workflows and APIs. And this is where you've used an agent recently, and you found that while, yes, it is really incredible. You can't quite leave it alone entirely unattended. You can't fully delegate a task to it and trust that it will fully get done. Why is that? I wager that the main reason behind it is that the agents absolutely must have the ability to write code. Now that's a bold statement. So let me explain why.

This all starts with something called the context window. The context we know you can think of it as the memory of the model. It's everything that the model is able to hold in its head throughout the transaction of an agent. If you hear [ Model Labs ] announcing new models often announcement start with this. We've extended the context window.

Now here's the thing about the context window. The context window of the largest models today is about 1 million tokens. But into it, you need to be able to fit in all the tool definitions that tell the model what tools it has access to, the conversation, the system prompt, the tasks.

But comparison, Cloudflare's API surface alone to tell model everything that Cloudflare can do, you need 2.5 million tokens, okay, more than double what's today's biggest context windows are.

And this is where you start to see where things happen, like the lack of accuracy that you see when agents are operating. For example, just this morning, I asked my agent to create an event for later tonight for dinner at 6 p.m. at La [indiscernible], great restaurant, by the way, if you haven't been. And well, upon first glance, it seemed like it successfully created the event. When I look closer, I realized that it actually created it for January 1, 2024, and which is not today's date. Why did I do that?

Models have no conception of what today's date is. The way that they obtain that information is through doing tool calling, but we just talked about how you have to be very conservative with what tools you include in your agent because of how quickly it blows the context window. So agents start to come up with all these inefficiencies and inaccuracies.

What this really comes down to is that agents are not good at tool calling. LLMs are bad at tool calling because there's no such thing as tool calling in nature. Tool calling is actually an entirely artificial process. Tool calling gets injected into the model after the model has already been trained. It's not actually code which is where models start to get easily confused. If you have two tools that have similar names, they can conflate them. very inefficient using these tools which using a lot of tokens, which adds up to a lot of cost for our customers. And this is not just me saying this about the model. The model labs know this about tool calling. This is coming directly from Anthropic.

Asking LLMs to do tool calling is a bit like asking Shakespeare to take a month-long course in Mandarin and then asking him to write a play in it. It's Shakespeare, so it will be good, but it's not going to be his best work.

Now what LLMs are really, really good at, what they're exceptional at and have been trained on lots and lots and lots of is code. LLMs can write code extremely efficiently and quickly. And LLM can write a few lines of code to figure out today's date in a matter of milliseconds.

But LLMs are good at writing code, that means that they now need a place to be able to execute and scale all of that code. Well, if they need a place to scale and run all that code securely, why not just do that the same way that we've scaled and run applications all of these years, which is why not just run them on hyperscalers?

Well, the reason for that actually sits behind the way that we scale applications today. So let's talk about that for a moment.

Every application generally begins as a monolith, as a developer, you write that single application. And actually, after you write that code, it scales pretty well to the first 100, maybe even for 1,000 users of them coming in and running through that code that you wrote. But eventually, you get that 1,001 user and that monolith is no longer scales for that. So what do you do? You create more copies, more instances of that exact app.

For the past 20 years when we've been talking about scaling applications, and we've been talking about micro services, this is exactly what we're talking about, running the same code over and over again. And when you need to make an adjustment you write that code, you thoroughly review it and you release it to your users.

You just don't work like this. It just don't need more copies of the same code. Agents need a distinct new piece of code written not just for every user, but for every single task and sometimes even sub task. What we've been doing simply doesn't work here.

To use an analogy, the way that we've been scaling applications is actually very similar to the way that fast food restaurants work. You have a chef, they come up with a menu, maybe they update it every couple of months. And you have this kitchen that's been designed to very efficiently execute on that exact same menu producing hamburgers, fries and milk shakes.

Now when you have a lot of customers, how do you scale as a fast food business? Well, you create more franchises, right? More copies of that kitchen designed to produce even more hamburgers, fries and milk shakes.

Now the really incredible thing about the year 2026 and the time that we live in is that we accomplished what we previously thought was impossible. Now every person can have a personal chef in the form of an LLM that comes up with a menu fully tailored to that user and what they want on that given day. But we're still stuck with that exact same kitchen that's only optimized for executing on hamburgers, fries and milk shakes.

What you really want in this situation is something more akin to a pop-up kitchen where you can quickly spin it up, have just the tools, just the ingredients that you need and be able to produce that meal on demand quickly spin down that kitchen and execute on the next one.

Now it's not just conceptually that this doesn't scale. The math shows it too. Let's run through what it would take to power agents for the entire United States alone. There are about 100 million knowledge workers in the U.S. day. Now being conservative, let's imagine that we run one agent per person. And you can fit about 10 agents per CPU if you're running containers. You need about 10 million CPUs to be able to power all of that.

For context, current global CPU server production is about 35 million to 45 million a year. So to power the U.S. alone, that's about 20% to 30% of global CPU production. But we're not alone in this world, right? So now let's run the math globally.

In the world, there are 1 billion knowledge workers. And realistically, we're not running one agent per person. They can run very efficiently in parallel. We could run two of them. They can run on were a sleep, 1 were at dinner. Now if you do the math, you need 1 billion server CPUs. That's 20x current global production.

If you think about the phases of AI adoption, each of them has been defined by a different bottleneck. The training era needed consistent GPU resources for training for inference, you needed flexible GPU resources to be able to adjust to unpredictable traffic. The first phase of agents, the bottleneck was models getting smart enough.

And the next bottleneck is going to be on access to the CPU resources. And the cloud that got us where we are today, it simply doesn't get us to where agents need to go.

But you know what does. Remember that big bold bet that we made 9 years ago on isolates. Well, this is exactly what isolates were built for. Isolates are perfect for this moment. Because unlike [ VMs ], isolates don't need to bring an entire operating system. Every time they need to run a line of code. Or unlike containers, they don't need to spin off the entire language run time every time they need to run a line of code. They can actually just bring in the application code or the code that's been written by an agent and execute it on demand.

This means that we can be 100x more efficient than any type of compute out there, being able to import the code that the agent just generated and move on to the next one.

And it's not just about our efficiency. This results in meaningful cost savings for our customers, too.

I'll start with a counter example actually, where if you have a regular application that maybe only grows 10% year-over-year, and you've really invested in optimizing your one server running on Cloudflare and spinning up a new worker renting it every time that you need to run something might actually end up being about 15% more expensive.

But when it comes to running agent-generated applications, Well, if you need to run about 10,000 of them, that's going to be about 63% more effective on Cloudflare. If you need to run 1 million agents that's going to be nearly 75% more cost effective on Cloudflare. And at the scale that our customers run at, this adds up to really significant cost savings for them.

Now it's not that the cloud got everything wrong. In fact, [ Andy Jassy ] was absolutely right about one thing. And it's that it is all about primitives. It's just that the primitives that got us here are not the primitives that we need for the Agentic era. If the last arrow was defined by perimeters like [ S3, EC2, SQS ], the next generation of applications is going to need primitives that look a lot more like durable objects, artifacts, workers and workflows. And this is not just me saying this, our customers are starting to think about this already, too.

Just recently, I was talking to Mark Smith, who is the Head of Infrastructure at Discord, one of our customers. And he was telling me that he said, I know what hyperscalers are going to look like in 10 years. They're going to look exactly the same as they do now. I'm looking to Cloudflare to define what the next-generation cloud is going to look like because like us, our customers want to bet on the future, not on the past.

And we're about to live through a major generational shift on the web. The last one of these that we saw was mobile. And while the cloud proceeded mobile, it was actually mobile that made the cloud take off, right?

Before mobile, we could only access the web when we were tethered to our computers. But smartphones made it possible for us to access the web anywhere we want. And these new devices were so much cheaper that they became accessible to a larger portion of the population, making Internet traffic explode. We're about to see the exact same thing with agents. We're already on track to surpass human traffic with bot traffic, as Matthew was just saying before, earlier than we even expected.

Stephanie is going to talk about how unlike humans, they just never tire. We get bored after clicking on 5 links. Agents can look at thousands of links before they reach a conclusion. Traffic on the Internet is about to take off, and it's going to need a different type of cloud to be able to power it.

And if the iconic companies that were born out of the mobile native era, or Airbnb and Uber and Netflix and they were all built on clouds like [ ABS ]. While the next generation of iconic applications, these agent native businesses are going to be built on Cloudflare.

And this is not a prediction. This is happening right now. Look at the past months of announcements alone, [ Figma, Lovable, Anthropic, OpenAI ] the world's fastest-growing agent native companies are already betting on Cloudflare.

We're building the fastest, most secure, most cost-effective place to build, deploy and scale agents and the code that they write.

And I'll leave you with this one final thought. There are 8 billion people on this planet. And we're starting to rely on agents not just for our jobs but for our day-to-day lives. We're going to need multiple agents running per person, and these agents are going to generate infinite lines of code, and we can't just build more servers to scale our way out of this. We need a cloud with a fundamentally different approach. And Cloudflare is the only cloud that's going to be prepared for this next agentic era. Thank you.

All right. With that, I'm excited to invite Sam Ray to talk about how we do AI class layer.

Thank you, Rita. Good afternoon, everyone. My name is [ Sam Ray ]. I'm Cloudflare's CIO. And today, I get to share with you how we equip and enable everyone at Cloudflare to use AI. Now Rita just told you about how their team has been creating the world's best platform to build and deploy AI agents. And my team is lucky my team, we have the privilege of getting to use that platform first, becoming the first users of these technologies.

We are our own most demanding customer, and we're also a very active customer. Because recently, we needed a way to give our entire team a turnkey solution for managing agents, for automating work for getting their jobs done in the AI era. So we started building one. We started using all of those components that you heard about to create something that can really supercharge our workforce.

And the result has become something that we call Cloudflare OS. It's this intuitive platform that gives every single member of the Cloudflare team the ability to automate work, to build and manage agents, to create apps and documents all running on the infrastructure that you just heard about.

Now Cloudflare OS starts with a cloud-based agent running in our infrastructure that users get to with a single click, no more forcing your entire team to fumble around with commandline tools. In fact, no requirement for local tools at all, which means your agents and your workflows keep running when you close your laptop. And it's all happening inside of an environment where we have total control and total visibility.

And team members, they're not constrained by the number of hours in the day that they're on a keyboard. They can build and schedule agents that get jobs done autonomously. It comes batteries included with our organization's brain. We've invested the time to make sure it has the right context about how we work, about what our priorities are and safe and secure access to our company's data.

All while making sure, like you heard Matthew mentioned earlier that everyone on the team can use the right model for the right job by running all AI inference through our AI Gateway.

So here's the architecture at a high level. it starts up there at the top. An employee starts their day by opening their laptop and within seconds, inside of their browser, they have a full comprehensive agent workspace. All without needing to take time configuring local software. For IT, that means we get everyone productive immediately. And for our security team, that means we move away from what I think is really an anti pattern, this idea of putting all the data back on the device. Instead, it's all taking place in an environment where we have total control and total visibility.

Now that employee, they start going them out their day by accessing our context library. This is a collection of workflows, runbooks, material, but give them the ability to solve problems and accomplish jobs with a single click inside of the system.

And then on the bottom right here, like I mentioned earlier, they also have the safe and secure access to our systems of record, all of them because for a lot of teams, giving everyone access to all of your systems of record is both a security challenge and it's a context window challenge because like you heard Rita mentioned earlier, when you access these kinds of systems, it can explode your context window.

It makes the AI agent or they add workspace overwhelmed, but not with Cloudflare OS because we get to use something that we call code mode, where our agent workspace doesn't interact directly with the systems of record, interacts with an isolate, running in our network, that executes code to then go interact with those systems of records that we can make sure every employee has access to the hundreds of various tools inside of these systems without overwhelming the context windows, again, for our security team, all happening in a way where we have total visibility into every tool call.

And then when those workflows, those agents, those user sessions when they need to use AI, we send everything through our AI gateway, no exceptions because that gives us the ability to dynamically route entrance to the right model for the job, and so for the IT team, this is fantastic because not everyone needs to spend $20 with the world's most powerful model just to summarize an e-mail. And for our security team, they love it because they have total visibility into all AI usage at Cloudflare.

And finally, our team members can build applications inside of Cloudflare OS that also run as isolates, and they can schedule automated agents and workflows to get work done when they are not at their laptop.

Now Cloudflare OS, it's our open-ended internal platform, but it's not the only way that we deploy AI Cloudflare. It's part of a larger funnel. Sometimes users, they build great applications in Cloudflare OS. And that gives them an incredible opportunity to then just immediately share those with their teammates, all safe and secure behind our Zero Trust layer.

But sometimes those applications are really great. We see experts across the organization in various functions, building the applications, they always wanted the tools they always dreamed of to get their jobs done. And in those cases, we're happy to partner a little bit more closely with them so they can kind of run these through an AI car wash and they become more formally supported tools in our organization. It's really incredible to see the creativity that gets unlocked with everyone becomes a builder on top of Cloudflare's compute and storage platform.

Now thanks to our AI Gateway we can treat these models more like a choice, like a commodity picking between the right model for the job while controlling token usage. With enough external models to power workflows, but we can also steer jobs to the models that are running on our own GPUs, where we run inference with an incredible efficiency.

And this also means that the latest models are available to every team member without any additional configuration. We're not handing out tokens or sign-in keys. They opened this workspace. They have access to the latest greatest models and they have access to models that map to the work they need to do.

Now we have all these incredible technology primitives that power Cloudflare OS. But like everybody else, you still get faced with this blinking cursor problem. You have people in AI workspace and you say, go and they say, what do I do here? And that's why we've really spent a lot of time investing in a central context layer. Our organization's brain because we have thousands of employees in various functions, and we want to take those manual jobs to be done and map them to skill files to workflows to instructions that team members can schedule or on-demand invoke with a single click, all from a visible, transparent place inside of the ecosystem.

All right. Time for the fun part. Let's walk through how team members use Cloudflare OS starting with that context player. So inside of Cloudflare OS, there is the centrally managed skill file collection. These are all jobs to be done in our organization. And these map to various roles, to various functions, and each can be invoked with a single click or scheduled with a trigger. It can take action while people aren't at their laptop. And people can also come bring their own context, their own skill files, so they can make this environment tailored to the way that they do work.

And so within those people who are doing that work, let's zoom in on two personas, builders and sellers, how do we equip and enable them with Cloudflare OS to do incredible work.

First, when developers use Cloudflare OS to write code, how do we make sure it meets our quality bar. AI allows us to write more code faster than ever before, but that's only useful if it's not introducing more problems. So you heard Matthew mention this earlier, but every submission to the code bases at Cloudflare runs against an AI review, that's not just a generic AI review. It is powered by this context layer. We debate and argue over because we want to make sure that everything that gets introduced into every code base at Cloudflare maintains or improves our quality bar.

And when it doesn't, it gives the developer the instructions that they need to go make the fixes and they can hand those over to an AI agent to go solve that problem for them because Cloudflare OS lets our builders focus on higher order challenges because they can just give Cloudflare OS a ticket or a problem. And it will go about fixing it. It will provision a sandbox in our environment to test it. It will go create monitors to make sure that the builds are passing that the test suites are passing. And if they aren't, it will come back and fix it for them as well. And that means our developers can focus on higher order challenges or, quite frankly, they can go get a coffee because, again, this doesn't rely on them walking around the office with their laptop cracked open, it keeps running in our infrastructure so that they can focus on other things.

But what about everyone outside of R&D? How do we make it so that everybody at Cloudflare can own more of their workflow. They don't have to wait on five different specialized inputs to every stage in their workflow, how do we equip them we built Cloudflare OS to solve that as well. So here's an example of somebody doing research ahead of a conversation with a prospect customer. Before Cloudflare OS, they would have to go manually conduct this research, the hours, finding this material, referencing our blog and other material that we have internally, we want them to be able to spend that time with more customers. So with Cloudflare OS, with a single click they can get a tailored research report for one or dozens or hundreds of customers ahead of the conversations they go have with them.

And this -- we don't need containers to go do this. What you're seeing on the screen is all running with our agents SDK orchestrating this so that it scales to meet our entire go-to-market function and more. And so we went Cloudflare OS -- we went sellers hit the ground running with Cloudflare OS so that they can, for example, start their morning without needing to go into a dozen different vertical applications to figure out what they're supposed to be doing that day. They can schedule workflows to run before their day starts. We'll go look at our sales force, our JIRA, their calendar, their e-mail and say, "This is where you ought to focus today" all before they sit down. And this works for various functions inside the organization.

Actually, right about now, I have an agent that's reviewing some updates to our MCP servers using my own personal context file, so it can provide feedback while I'm here with you.

And as that workday gets started for the rest of our team, for example, our sellers who used to need to request things from competitive intelligence or our brand marketing team to put together a deck before they go meet with that customer from whom they've done some research, they can create in a matter of minutes, a tailored prescriptive deck using that organization's brain to think about how we talk about our products, how we position ourselves, what are the case studies we want to reference? Gone are the days our single corporate deck, our sellers can own more of this workflow.

And like I mentioned earlier, some colors projects, they really deserve to be packaged up as formerly supported tools in our organization. And that's increasingly from people outside of the R&D team, which is really exciting. This is an example from members of our advanced marketing team, who they plan hundreds of events like this one and small events all around the world using dozens of different spreadsheets and various tools, which inflicts an extremely high coordination cost. That's really painful.

And in fact, one of them sitting in our Lisbon office got so sick of this that she said, I'm going to build something better and use Cloudflare OS to prototype and idea that removes all of that coordination cost, building the tool that they always wanted. And we've now partnered with them to make it something that's more formally supported in the organization. And I think you're going to see this happen in every part of our team.

And in addition to reducing the coordination cost of our organization, we also want to shorten the time spent waiting in various parts of our team.

So imagine that seller, they did that research for that customer. They had this prescriptive deck created for that customer. They met with them. They felt confident about the meeting because Cloudflare OS helped them out, and the conversation went really well. Customer said, "Great, I want to get started".

So in the old days, the seller might think, "Oh, gosh, I need to get scheduled time with a solution architect or a bunch of other people get some expert opinions, come back, put a document together, get it reviewed", not with Cloudflare OS, because it can pull from our organizations bring that investment we've made in that context player, they can come in and just tell Cloudflare OS "hey, conversation went well today". The customer really wants to talk about our developer platform and our Zero Trust products and it will create this prescriptive solution architecture for them.

And of course, our teams are also global. I'm visiting her from our Lisbon, Portugal office, and we have teams around the world meeting customers in every industry in every geography. And previously, that slowed us down. Because we would need to wait on teams to do specialized case study, go find different advocacy stories, go find different localization tools, make sure that the materials that we've been talking about today. We're accustomed to a given region, given industry, given segment. We don't need to do that anymore. We can make everyone all around the world of Cloudflare more productive by giving them the tools to adapt everything that we generate to the customer, the region, the industry that they're meeting with all in a matter of clicks.

Now I feel like I'm the luckiest CIO in the world because I get to build on top of everything that Rita and team are creating. But being the lucky CIO in the world, that's only partially true. The reality is every one, every CIO, every team in the world can use these same components to equip and enable their organizations to use AI.

I'm so excited for us to keep building and more importantly, sharing and learning from our customers because at this exact moment, every organization on the planet has this exact same problem. So I appreciate your time today. And with that, I'm going to hand it over to [ Stephanie Cohen ], our Chief Strategy Officer, who's going to talk about the future of the Agentic Internet. Thank you.

Thanks, Sam. It is so great to be here. My goal for today is to build and we heard from Matthew, from Rita and Sam and demonstrate to you why Cloudflare was built for this moment.

The Internet itself is changing. We all see it in our daily lives. Cloudflare is the only cloud designed for this new Agentic Internet. And not because we decided it was the cool thing to do, it's because we repeatedly made the right architectural choices over the last 15 years.

The new Agentic Internet, it's turbocharging our existing businesses, and it's opening up new opportunities because on the Agentic Internet, every single interaction is a chance for commerce.

So yes, I don't think it's an exaggeration that orange is a new cloud. And that's coming from someone who lived right here in New York City for a really long time, and I prefer to wear black.

While we, along with everyone else eagerly await for what this new Agentic Internet holds, we are already seeing the benefits in our financial results. In the sectors that are most affected by the Agentic Internet, we're experiencing revenue growth north 37%. And that's driven by new customers like Reddit and People Inc., and existing customers such as LinkedIn.

Our retention and our expansion are higher because many of our products are now strategic imperatives. Those bots on your site, they're no longer just a nuisance or a security threat. They can either be a chance for new monetization or an existential threat, depending on the decisions you make and the tools that you choose.

When we go to meetings on application performance and security, they're now with the CEO, in addition to the CTO and CSO. Cloudflare has never been more important to the C-suite.

So before we go deeper into these results, let's start with how we got here, and you heard this from Matthew. A new Internet really is emerging.

Okay. So you're all here, here in the context of your job. But I wanted you to take a minute and think about how your own daily interaction with the Internet is changing. We're going through a platform shift at the fastest rate we have ever seen. The numbers on this page are staggering. It took 14 years for the Internet to reach 1 billion users. Social media was almost 9 years, smartphone 6, generative AI in less than 3 years. And generative AI just keeps growing. It's now at more than 2.5 billion active users. That's over 30% of humanity in just 3.5 years.

And the curve isn't just getting steeper, it's going vertical. Never before have we seen such a dramatic change in how we interact with information and how we perform our jobs. And this platform shift is not just faster. It's actually more extreme.

In the AI era, the way all of us use the Internet and actually spend our time is changing. The other platform shifts mainly fed the Internet, but this one is taking our attention away from it. The previous platform shifts taught us that if you just put your content out there and you allow bots to scrape it, you're going to get rewarded with traffic, and you can monetize that traffic with ads, subscriptions and affiliate revenue.

But traffic from humans on the Internet, it's disappearing. For every hour that we all spend today, searching online for information, only 15 minutes of that is actually spent on the Open Web.

When was the last time you actually clicked on a link before you just read the answer and why? Because for most things, you just want the answer, you don't want a treasure map a [ BlueLinx ]. These numbers don't lie. Traffic from humans for many news and content sites is quickly heading to 0. The [ faustian ] bargain between search and content is no longer just a bad deal. It doesn't even exist.

Many of our customers are preparing for what they call Google Zero, a moment where none of their traffic comes from search. The answer is not to go back to the way the world was before, but we do need a new ecosystem because high-quality original content is valuable, it's valuable to all of us, and it's also necessary for AI to flourish.

And while media is the canary and coal mine, they're not the only ones. Retail, software, IT and finance are all seeing meaningful declines in human traffic on their sites to the tune of almost 40% in less than 1 year. And this matters. It matters not just to those sectors and companies, but it matters to all of us. The Internet is a critical part of the economy, and it's also a really important conduit of information.

And it's why we at Cloudflare come to work every day. It's the North Star for how and why we deliver value to our employees, our customers and our shareholders. For that reason, we are uniquely suited to help build and power the business model for this agentic Internet.

Providing the control plane for the agenetic internet requires a unique combination. You need supply and you need demand, but you also need technical capabilities that support both sides and they bring them together so that everyone can benefit.

Let's start with the supply side. I think you may have heard the internet runs on Cloudflare. Because of that network that Matthew talked about earlier, more than 20% of the internet runs on us. But we also have depth. 36% of the most visited sites sit behind us. And more than 40% of the Fortune 500.

We also have a large amount of hard-to-find unique user-generated information that comes from communities that really matter, both to humans, but also to AI. But we said this before, this is not just about news and content. Every site is being impacted by the change in how the Internet functions.

Every site needs to understand what it means to be agent enabled. You agree to talk about all those agents. Those agents are going to all of our customer sites. And they all depend on Cloudflare to help them, whether it's media or software or retail or finance or health care.

The largest sectors in the economy are all experiencing historic changes and traffic to their sites. And they depend on us. They depend on us to help them understand what the world is going on to give them insights, but also give them the tools to adapt.

But what about the demand side? That part of the Agentic ecosystem that want access to the information, they all depend on cloud fair, too.

From OpenAI to Anthropic, almost 80% of leading AI companies use Cloudflare. But it's not just the big companies. We also support the long tail. And the long tail really matters here if we want a sustainable ecosystem to emerge. We can't just have a small handful of large LLMs deciding what information matters to the world.

Because of Cloudflare, this long tail is actually able to exist, and they can compete with those really, really large companies, and they do that by leveraging our network, our scale, and innovative products that you've heard a lot about today.

We are uniquely really easy to use. We're efficient. We are global out of the box. And as you heard from Rita, our TCO for agent-generated applications and agents is 60% to 70% less expensive. I think we need to pause on the 60% to 70% less expensive because that's how we get an ecosystem to emerge.

It is our unique innovative capabilities that bring everything together. It makes Cloudflare end of one when it comes to being a platform for this new agentic internet.

Our position of sitting uniquely at the intersection of the internet and AI, that's been built over the last 15 years. It's not an accident. The world chooses cloud player because they are at their best when they build on us.

And while there's, of course, dependent on products, it starts with a set of beliefs that are anchored in our mission to help build a better internet. That helps us keep making the right architectural choices over and over again.

You heard from Rita on the isolates. This is something that's embedded into who Cloudflare is. We're simply the best partner for every player that is building for the agentic Internet now and in the future.

We sit in between the supply and demand and have world-class primitives that work together seamlessly and will allow this nascent agentic Internet to flourish.

Well-functioning markets like the one who's building we're sitting in today, they don't just appear. They have to be built on the right platform with the right capabilities. And for us, these capabilities include Act 1 products like Bot Management and the full suite of developer products from Act 3.

Let's start with Bot Management. My guess is I don't have to tell this crowd, but there is no sustainable market without reliable scarcity. No one pays for things that they can get for free. And of course, that's true for AI companies.

The world's largest companies with the most sensitive data, including the AI companies themselves depend on Cloudflare because no other bot management provider has the breadth and depth of data that we have.

We see every IP address on average 7 times per day and sit 50 milliseconds from 95% of the world's population. Our win rate show that we're just better and faster every time. But for most of our customers blocking, that's just the starting point. What they really want is they want to optimize and monetize their content, and information for AI.

But in order to do that, they need analytics, the dashboard that we originally built for the security teams is now for the C-suite, business and marketing teams. This idea not only expands our customer base, but it actually makes us stickier. Once you help someone drive revenue rather than just reducing cost or risk, a product that could have become a commodity is now price inelastic.

And what about those AI companies who are spending hundreds of billions of dollars crawling and Internet that was meant for humans, not machines. They need our help, too. Those same signals that make our management products best-in-class, they help AI companies decide what content is scrape and how often.

Consider for a moment how easy it is for us to tell if a site behind us has been updated or where the domain behind it is trusted. AI companies value that information on top of the value for the content. Our agents SDK makes it easy to build an agent that obeys site owners request. And therefore, has better and also faster access to what it means on the Internet.

And we've said this before, the vast majority of AI companies and agents want to do the right thing. Okay. So I want to pause here because there's a lot of people who believe that AI and agents will just prefer to ignore site owners preferences. The data on our network tells us that's just not true. The reality is it's just that not easy to traverse the internet. Just like we want to extract away infrastructure for builders. We want to do the same for helping build agents. We uniquely can help builders to build while helping site owners maintain control.

But it's not just about the tools, it's about the ecosystem itself. We, as a society, I think we can all agree, we need a level playing field. This chart on the right, this is just not good for the world. It shows that one large LLM has access to almost 2x more information than the next LLM. And look at how quickly those bars get smaller, how can new model companies ever compete. We want to create an ecosystem where small model providers have this opportunity to actually compete with the world's largest companies by being able to access the same information.

Right now, large existing companies are leveraging their advantages in one market to help them win in this new market. Cloudflare uniquely knows where the asymmetries exist. And can help AI and agents access this harder-to-reach information and price what it's worth.

With the finance crowd, every functioning market, what do we need? We need a payment system to support it. Today, there is no payment system that can match the throughput of the internet and the Cloudflare network. We process more than 500 million transactions per second. The largest payment company they can process less than 100,000 transactions per second. I can assume we can all do the math, it takes us less than 2 seconds to process the number of transactions that they process in a day. So it's not surprising that these large payment networks have decided that Cloudflare is a really important partner to them. All of them have embedded the ability for us to cryptographically verify bots at scale into their payment protocols.

Okay. So that's how it works today, but there's a much larger opportunity. We spoke earlier about how the Internet is changing and Rita alluded a bit to this, but we didn't cover what that means for traffic. When a human goes to research a bot, they may go to 5 to 20 sites, your agents going to thousands. And while traffic has always been a positive, growth in traffic has always been a positive for Cloudflare, growing agentic traffic is even more interesting.

Today, we monetize request by providing security and developer products. Tomorrow, we can monetize requests by simply being in the flow, the way a payment company does.

To start, with [ Coinbase and Stripe ], we created the X402 protocol so that machines can pay other machines at scale. Already today on the Cloudflare network, there are more than 2 billion 402 responses a day. A 402 response means payment required. So it's not just a block, it's say, please pay me.

Today, these -- the vast majority of these 402s, they don't actually result in a transaction. But the scaffolding is there. It's built into the Internet. With the capacity for a higher number of machine-to-machine transactions than the largest payment networks in the world.

X402 is payment rail agnostic, but it is perfectly designed for stable coins. For the dream of agentic commerce to be a reality, a lot of things need to work together. No other company has all the pieces to enable this agentic Internet, building the agent verify the agent and allowing for the payment.

But it's not just about agentic commerce. It's about changing the surface area of the economic opportunity. On the human Internet, I think we can all agree, payments can be a bit of a nuisance. But on the agentic internet, they are the oil that allows new ecosystems to emerge micro payments for access to content and systems of record, those just weren't viable before.

And Cloudflare sits in between all of it. We go from protecting websites and applications to monetizing trusted, automated demand.

Yes, the agentic internet is nascent, but we already have proof that it's working. The largest media companies in the world are moving a Cloudflare to ensure that they're agent enabled, People Inc, [ Condas ], Financial Times, just to name a few. And these companies are using Cloudflare to negotiate large deals with the LLMs.

The best example is People Inc. In July of last year, they implemented bot management. And by September, they had seen a significant shift in their negotiating leverage. What they do, they created scarcity and they used that dashboard that we showed you earlier. And yes, the splashy deals between the big publishers and the LLM, those make the headlines, but it's not just about them.

Clay, a start-up. That provides AI-powered sales and marketing tools to automate lead generation, prospect research and data enrichment, they're using Cloudflare for signals, so they can quickly and efficiently access updated information on the Internet.

Before they started using Cloudflare, they were constantly scraping the same sites over and over again. That creates unnecessary load for them and for the sites. It's very expensive. But now they're using our network and our storage capabilities. So they -- we can just serve them the updated information. This capability is where we expect to start with most AI companies. we make it faster and cheaper for them to obtain access to information that they already can get to.

But the next step is us getting them information from hard to reach places, remember that chart, the difference between the LLM that had most information and the one that had the least, this is the ecosystem we see emerging.

To that end, we have large AI companies working with us to get access to this niche, hard-to-reach content. Only Cloudflare would know that a top LLM was unsuccessfully attempting to crawl a specific health care site, almost 800 million times in a 30-day period. We can see where the demand is and price it not just once but programmatically.

The promise of the agentic internet, it's already here. It's helping us to win new customers. grow existing customers and increase stickiness and decrease price sensitivity. We are executing content fields and building stable coin payments into our products. We are building an ecosystem where both sides can benefit.

Act 4, it's not separate from the rest of Cloudflare. It's only possible because of the foundations we built from Act 1 and Act 3. We're the only platform that can build secure and enable this future. There is no more consistent with our mission and nothing that will deliver more value over the next few years.

And with that, I'm going to turn it over to Mark. He is going to talk to you about how -- he's helping and all of us are helping our customers adopt our platform at scale.

All right. Hey, there folks. Excited to be here with you today. I'm looking forward to sharing what we're seeing and hearing from customers around the world and how I think about the go-to-market team evolving during these really extraordinary times.

Throughout my career in tech, I've tried to spend more than half of my time communicating directly with customers. Cloudflare, given the sizable transformation I've helped orchestrate I think the -- it's the best way for me to learn what customers know about us and why we should matter to them. Also learn how well equipped my team is to articulate the compelling value that we can deliver.

So let me start about what I'm hearing today. So I've been in this role now for a little over 30 months. When I first joined Cloudflare, a large portion of my time was spent introducing the company to enterprise and government customers.

Conversations kind of started with the basics. Who's Cloudflare? What do you guys actually do and should mean my big bank care? At the same time, I was sharing a consistent message that AI is going to change everything. Organizations are going to need to modernize their infrastructure, their security architectures and their application environments to be able to take advantage of AI.

I challenge customers to think differently about the future. Who are you going to pick for your infrastructure partners for the next 5 years? Guess what? It's not going to be a lot of the infrastructure partners that you've had for the last 5 years.

So fast forward to today, the contrast it's remarkable. I spend far less time explaining who we are. Instead, customers are coming to us because the challenges that they're facing increasingly align with what Cloudflare was built to solve.

This market is moving towards modern cloud-native globally distributed architectures, and Cloudflare is already there. Now that shift is showing up in our brand awareness, our enterprise engagement and ultimately, in our business mix. And the conversations haven't just become easier, they become far more strategic.

I've lived through some pretty big platform shifts as an operating executive over the last 25 years. But since the end of 2022, an OpenAI announced ChatGPT, the pace of change has just been astonishing. In these last 2 years, I kept hearing from customers, chat bots, copilots, they're interesting novel, not really moving the needle with regards to delivering real enterprise value.

But since the beginning of this year 2026, well, agentic AI is all anyone wants to talk about. In every single industry everywhere I travel in the world. AI and agents are changing everything. And earlier this year, the market began to understand just how quickly this transformation is occurring.

Companies like Anthropic have demonstrated the enormous value that AI can create inside large enterprise. But what's even more important is what AI enables. And they're beginning to think -- sorry, they're beginning to rethink how work gets done, how applications are built and how employees operate and ultimately, how businesses create value. I think we're entering a world where software doesn't simply support work. Software increasingly performs work. And that requires organizations to rethink the technology architectures, application infrastructure stacks that currently support their business.

And one of the most significant shifts I've seen since the beginning of the year has been the speed at which AI adoption has evolved. If you go back roughly 9 months, most conversations centered on enablement. Organizations wanted to experiment with AI, but they were equally focused on governance and controls and guardrails. The question was, how do we safely allow employees and developers to start using these technologies.

Again, fast-forward to today, the conversation has changed dramatically. The urgency level has gone from 0 to 100. Organizations are no longer asking whether AI will impact their business, they're asking how quickly they can operationalize it.

I'm seeing customers move from pilots and experimentation into production every single day. And increasingly, those deployments involve AI agents interacting with applications, infrastructure, data and their business processes.

Now looking ahead, we believe that autonomous agents will become embedded across nearly every enterprise workflow. And that is where the opportunity for Cloudflare becomes increasingly compelling.

I think AI is changing how work is done. And that trend only accelerated with the rise of agents. Just in the last year, agentic traffic on our network, as you saw, grew 1,700%. When you move more than 20% of the world's web every single day, the insights that only Cloudflare can glean from those traffic patterns is a super power for all of us here at Cloudflare.

Also, in the past month alone, we've commercialized major relationships with AI native companies like Anthropic, OpenAI and Lovable that understand that our workers' developer platform is the very best place to build and scale AI agents and the code that they write.

And this is just the beginning of the beginning of the agentic cage. And it's not just changing how business is operating. Of course, it's changing how attackers operate. And what you're seeing here is the dramatic compression and time between vulnerability disclosure and active exploitation.

Historically, organizations might have had months or even years to be able to identify, prioritize and remediate those vulnerabilities. Today, that window is shrinking rapidly.

Today, it's seconds. And I think AI is going to continue to accelerate that trend. What I'm hearing today and seeing from CSOs is fear. They've realized that this year, legacy security approaches are just not effective. Organizations need architectures that can detect, adapt and respond in real time. And that's exactly why we're seeing so much urgency around security modernization.

Now as we talk with customers around the world, we're seeing two powerful forces emerge simultaneously. On one side, every organization is exploring how AI agents can improve productivity, automate workflows and transform operations. As discussed, those same advances in AI are changing the threat landscape.

Frontier AI models are making attackers more capable, more automated and much more sophisticated. Individually, each trend is significant. Together, they're creating urgency inside virtually every enterprise, and I've never seen an environment like this before.

Now customers are being asked to move faster, while simultaneously reducing risk. They're modernizing infrastructure, the rethinking security architectures, and they're looking for platforms that simplify complexity rather than add to it.

I believe Cloudflare sits at the intersection of those needs. And that's why we see the demand going through the roof like they are.

What's particularly interesting to me is that all of this is happening against the backdrop of relatively flat budgets, IT spending is growing modestly. Headcount growth, it's kind of slowed significantly in IT and Infosec and technology leaders aren't being handed dramatically bigger budgets. They're being asked to accomplish substantially more with roughly the same resources, which means every single decision matters.

And despite those constraints, AI spending is accelerating. [ Gartner's ] research shows that nearly 90% of technology leaders expect to increase their AI investments this year.

Security remains a major investment area. Application modernization remains a major investment area and cloud platforms remain a major investment area.

Now what's important is that these aren't separate initiatives. They're increasingly connected. AI adoption requires modern infrastructure, modern infrastructure requires security, modern security increasingly requires a platform-based approach.

I'll hear different perspectives from executives every single day. For the CSO and governance teams, the primary concern is managing this new wave of risks. Including employees using unsanctioned shadow AI tools. For the CIO or infrastructure team, it's about unlocking productivity. You want to empower your teams with the best AI tools, but you need to do this in a way that's both safe and agile.

And for the CTO or technology teams, the goal is to innovate quickly empowering your developers to build the next generation of AI architectures with security embedded from the get-go, not sold separately, not as an afterthought. No matter their role all IT leaders are almost exclusively focused on how to best accelerate AI adoption to drive their businesses forward.

Ultimately, it's about speed and security, no trade-offs. Customers don't need entirely new architectures, infrastructure stacks to be able to adopt AI. They need infrastructure that was already designed for this reality, infrastructure that can support users and agents, infrastructure that can connect, protect and help build an infrastructure that can adapt as technology evolves and boy, is ever evolving.

That's where Cloudflare is differentiated. Our network is the earth. If you step back and look at everything we've discussed so far, a pattern emerges. Organizations are moving faster than ever, they need to embrace AI. They need to defend against a rapidly evolving threat landscape, and they need to do all of that without dramatically increasing cost or complexity. That's exactly the type of challenge that Cloud there was built to solve.

I had a unique vantage point before joining the company as an operator because I spent nearly 5 years on our Board of Directors. What attracted me back then was this architecture, one global network, one platform, security, networking and developer services working together.

Today, it feels increasingly obvious to me the market is moving towards the architecture that Cloudflare has been building for years. And we're seeing these demand signals translate into larger opportunities deeper platform adoption and accelerating enterprise engagement.

Now when you take a look at the market, most vendors participate in only one part of this AI journey some focus on security, some on infrastructure, some on development. Customers increasingly need all three, and they need them to work seamlessly together. And I think that we're unique because we help customers connect, protect and build these AI applications on a single platform on a single cloud-native modern network.

That breadth becomes increasingly valuable as organizations move from experimentation into production. Customers don't want more complexity. They want fewer, larger platform partners that can do more for them. And the build portion of this story is becoming increasingly important. Enterprises aren't just consuming AI anymore. They're building AI-powered applications. They're deploying agents, orchestrating workflows and creating entirely new customer experiences. You heard what we're doing with [ SAM ]. Everyone is going to be doing this. They need infrastructure that helps developers move quickly without sacrificing security governance or performance. That's where our developer platform becomes a critical part of the story.

The same global network that secures applications can now accelerate applications. The same platform that protects AI workloads can help customers build AI workloads, the same architecture that simplifies operations also improves performance and resilience, programmability, automation now that is a very powerful combination.

So far, I've talked about what's happening in the market. The obvious next question is, are we prepared to capture this opportunity?

Because opportunity alone doesn't create outcomes. Execution does. Over the last 2 years, we've been transforming our go-to-market organization to position Cloudflare for the next phase of growth. Everything we've done falls into six areas. Recruit and develop stage appropriate talent, stage appropriate leadership, improve segmentation, drive customer adoption, selling the platform, leveraging value-added partners and increasing sales productivity quarter in, quarter out. None of these initiatives were designed to operate independently. Each element reinforces the others. Together, they form the foundation of a world-class go-to-market organization capable of meeting the moment that we're in today and especially the moment that we'll enjoy in the coming months and years.

The first step was bringing in stage appropriate leadership. We brought in experienced operators who scaled businesses at every stage of growth. Just as importantly, we strengthened frontline management throughout the field, pretty sizable change in team dynamics that you see here because great sales organizations are built manager by manager, team by team, that investment created the foundation for everything else that you'll see in the following slides.

And we've refined our sales strategy to be much more targeted and more structured, enabling an improved go-to-market approach to best address the unique needs of our diverse customer set. We have clearly defined market segments ranging from digital native, start-ups to large enterprise to public sector, and this allows us to apply specific sales models with the right resources to solve the specific challenges for each opportunity.

We have had and continue to have incredible success with [ SMB ] and mid-market companies from our freemium origins. We're stacking larger and larger customers on top of that. And in this graph, you're going to see the inflection in the slope of our $1 million-plus customers from the second half of 2024 with the momentum building through 2025 and through Q1 of this year.

And I believe we are still in the early innings with regards to large enterprise adoption. Although we count 42% of the Fortune 500 as customers, I'm not going to be satisfied until they're all Cloudflare customers.

Now our innovation engine has always been the very heart of who we are here at Cloudflare. Our incredible R&D teams continue to build more and more amazing new products and features for our customers.

As an organization, we've become far better at listening to these large customers and building products and features that they need for their AI revolution. To me, this is a massive competitive differentiator for our field team.

Our platform, as you've heard, built from the ground up with a full understanding that the entire infrastructure stack is going to have to shift to the multi-tenant cloud compute, storage, security and networking, cloud native, massive scale, single-user interface. We didn't glue together a bunch of acquisitions, a true platform where every service runs on every server in every rack in every data center all around the world.

And our field team understands that customer value this kind of solution. Simply put, platforms win. And that's where we continue to lean in.

Another important shift has been our partner-first approach. The largest opportunities increasingly involve complex transformations. Customers want strategic guidance. They want implementation support. And they want long-term operational expertise that partners bring. And they have to move fast.

Partners help us extend our approach. They give us feet on the street to start telling the Cloudflare story and they help us accelerate adoption. Drive consumption. As a result, partner contribution has steadily grown as a percentage of our business. Partners around the world are creating demand, fulfilling that demand and providing value-added services that amplify what we do every day for our customers.

We view this as an indicator of both scale and maturity in our go-to-market motion. So welcome back to my business-to-business go-to-market university. From the last 2 years, you'll remember my equation for momentum. It's the product of mass times velocity. And in sales, momentum in [ ACV ] is the product of sales capacity times productivity.

Ultimately, all of these investments need to show up in productivity, and that's exactly what we're seeing. Sales productivity has continued to improve and recently reached new heights. That's important because productivity is one of the clearest indicators to me that a go-to-market system is working.

Now I consistently believe that when you have a differentiated suite of products at a time when customers are just realizing that they need this differentiation. And you can continue to build and ramp high-performing teams to articulate a compelling value proposition, you can roll up market share in a very meaningful way.

My team certainly thrive with these dynamics at F5 and especially at Palo Alto Networks over the last couple of decades. But for us at Cloudflare, the opportunity is much larger.

Given our massive total addressable market, my plan is to continue to invest in sales capacity, and I expect that sales productivity will continue to grow for a very long time.

As a reminder, when I first joined in 2024, any of the CIO, CTO, CSOs had little or no contact with Cloudflare. Most didn't know who we are and why we should matter. Again, we brought in stage appropriate talent to build these relationships and earn the trust with the largest companies and governments around the world. Our momentum to bring the right AEs and Sales pod members to customer teams globally is going full steam ahead.

So where do we go from here? I believe that AI will transform not only what we sell, but how we operate. The next phase of our journey is applying AI to our very own go-to-market organization, not as a science experiment or side project but as a practical way to increase capacity, improve customer experiences and help our teams operate at greater scale.

We spent the last several years helping customers modernize their infrastructure. Now we're applying those same principles to ourselves. We intend to be one of the best examples of how AI can transform a modern enterprise because the future of go-to-market is in humans or AI it's humans and AI working together to create better outcomes for our customers and for our partners.

One principal guides our approach. As Sam mentioned, we're our own most demanding AI customer before we ask customers to trust these technologies, we want to use them ourselves. We want to understand how they create value. We want to understand where they fall short. And we want to understand how humans and AI can work together most effectively. That experience helps us become a better operator and a far better customer to our customer base -- sorry, a far better partner to our customer base.

And today, sales professionals, they probably spend less than half of their time actually selling. AI gives us an opportunity to change this big time. Across the customer life cycle, we're seeing repetitive activities that can be automated, accelerated or augmented. The goal isn't about replacing people. The goal is about allowing our teams to spend more time with customers and less time on [ administrivia ].

The next phase, our AI journey isn't just about how we sell. It's about how we serve customers throughout their life cycle with Cloudflare. We're applying automation and AI across onboarding, adoption and renewal.

For example, during onboarding, we're helping customers reach value faster through automated planning, deployment preparation and follow-up.

During adoption, we're using usage signals, business reviews and customer engagement data to identify growth opportunities and proactively address risk.

And as customers approach renewal, we're automating insight generation, forecasting and expansion planning. This will allow our teams to spend more time delivering value and less time managing process. AI is fundamentally improving the customer experience while also increasing the scale and efficiency of our customer-facing teams.

So folks, the opportunity in front of us the Cloudflare has never been larger. The market is moving towards the architecture that we've been building for years. Our platform continues to expand and grow. Our enterprise momentum continues to strengthen. Our go-to-market organization is operating at a higher level than ever before.

And we're just beginning to unlock what AI can do for our customers as well as for ourselves. I am excited about the road ahead. Now some of that excitement comes from the confidence that I have in the team that I've assembled and the executives that you see here in the room, I think we have an incredible opportunity ahead of us.

But folks, I've been at this for a long time. And it's with mixed emotions that I'll share with you at the end of this year, I'm going to retire. Psyched about it, my family is psyched about it. I'll be stepping off of our Board back in 2024, I had this incredible conviction that we could conduct that we could complement our product-led growth motion with the right people, strategy and infrastructure to dominate, where half of our total addressable market exists, our enterprise and large government. That's half the spending on what Cloudflare does.

And I'm really proud of the progress that we've made and I'll continue to cheerlead from the sidelines in 2027 and beyond. We're launching a public search right away. And I know we'll find the very best person to take the baton and guide this go-to-market team. and their efforts into the future.

In the meantime, I'll be squarely focused on our FY '26 plan and wrapping up the most successful fiscal year in Cloudflare's history. I want to thank you for your time, there are some friendly faces here in the room that I'm going to miss. But I really appreciate the support you've provided over the years. And with that, I'm going to pass it off to our amazing CFO, Thomas Seifert.

It is quite a while until the end of the year, but this is the Investor Day, and it falls on wherever it falls. So I want to acknowledge that it has been a true pleasure working alongside you Mark for the last 7 years.

You and I have been both long enough in this business that you got to know how important it is that our two roles have a really trusted relationship, and I truly want to thank you for that.

So I know it's a long time until the end of the year, but in front of our investors and shareholders, I want to acknowledge that. Thank you.

So I'm Thomas, the CFO of Cloudflare. I was standing up here yesterday. And one of you guys last year reminded me Thomas, keep in mind, we count CFO smiles, that is a really important KPI for us. And I said, finally, I get my moment where a CFO can smile.

I should have saved this comment for this year. This year is a way more appropriate place for smiles. And especially now after Mark [indiscernible] acknowledged that I work at a wonderful, magical company. So here we go. So off we go.

This is a slide you have seen two times already today. And I think it's one of the reasons, as Matthew said, this change that are behind the significant increase in traffic on our network is probably shaping up to be the biggest tailwind for Cloudflare moving forward.

So how does this impact our financials? When does it show up in revenue? How does it show up in revenue is what I would try to take you through over the course of the next 20 minutes. And in order to get started off correctly, we -- I take you back a moment and try to familiarize you again with how revenue and Cloudflare works.

So today, as you can see, we have four primary models of how we generate revenue. That's easy ones. The flat rate, this is where we came from, where the world was easy or new IPOed and almost 100% of our revenue was subscription-based. There is a larger share of pure usage-based revenue today, still under 10%, where customers pay us in the rear. The complexity comes from the two other components of how we interact with our customers, cat models and pool of funds. And you'll find out there's a lot to do with T-shirt sizes and Nike swooshes. And why is that?

So the majority of our contracts are based today on what we call a cap model. And requests in these contracts are the most significant measure revenue. So what does this mean? Think of a T-shirt. You buy a customer buys a T-shirt, commits to a ceiling of utilization of a product, a [ WAF ], for example, type by a medium-sized T-shirt 10 million requests per month. And it gives us predictable and smooth revenue as long as the customer is using under this umbrella of what he has committed to.

But what happens if a customer outgrows that medium-sized T-shirt and has to go out and buy a larger one, let's say, small from medium to large to $20 million requests per month.

The traffic is increasing more than the customer has planned, he needs to buy up. But however, remember, the customer is not going to buy what he consumes in the moment. He's trying to estimate what this future consumption is going to be. And this transition is letting our revenue grow, but it will lag to the inflection point of usage and makes a little bit of sympathy from all of you. Our job of guiding and forecasting a little bit more tricky.

I brought you a good example. It's a very large leading technology company in retail space, coding. So the company had already to increase their caps twice in the past year. The customer is working where agents are. It's highly likely that they will be up soon for another bigger T-shirt size. And keep in mind the lag to revenue will come with this model because the usage is lagging behind the revenue recognition. That is all still manageable. It gets a little bit more complicated when we get to pool of funds.

Pool of funds have become a meaningful instrument for us and extremely beneficial to the platform adoption to get to larger customers have frictionless expansion. In the fourth quarter of last year, 20% of our new ACV was coming in pool of funds. And pool of funds have a very significant impact in how they shape our financial model.

You'll see here two cohorts of pool of funds for 2004 and 2005. And you see what would have happened if we had recognized this revenue in a more ratable model the smooth line from the cap model disappears. It becomes much more a ramped model. And you can see that revenue lags compared to a ratable revenue recognition in the early parts of the contract before it catches up and exceeds it.

That delta becomes bigger, the bigger the pool of funds contracts are. And you can see that in 2005, we were even more successful signing larger deals, so the gap becomes more pronounced.

What happens now when the customer is really burning through that commitment really fast. This is an example of one of our largest customers in the AI space, increasing the pool of funds commitment already twice. And in the fourth quarter, there was a significant step-up in commitment, north of 215% of ACV commitment. The customer, of course, is negotiating better rates.

So in the first and second quarter, and you get a little glimpse into the second -- in the current quarter, it takes a while until revenue catches up and overtakes what revenue run rate with that specific customer was before. But you can also already see what momentum is building up behind the dip in the first and second quarter.

You can also imagine what happens to us when we stack a couple of these extensions in the quarter on top of each other, the headwinds are a little bit more pronounced, make our job a little bit more difficult. You also have to keep in mind that becomes very important when we talk about gross margin later. While the revenue dips, consumption or usage on the network is not going down.

So not only is a headwind from a revenue recognition perspective until the run rate catches up, there's also a slight headwind to gross margin for the transition.

So with this out of the way, moving forward, I want to focus on frame the discussion around the two primary drivers of value creation for us, growth and profitability. Starting with growth. I want to walk you really fast through the six or seven most important growth drivers. The [ closet ] business model always excelled itself, because we have these multiple independent growth vectors that have been carrying us forward.

So how do we think about the components at this point in time. First of all, there's a Today, we have about 75 revenue-generating products on the network, including Act 4, a massive opportunity. I remember when I joined, I think we were at [ 5 ]. Our total addressable market is forecasted to be about $238 billion for this year. That is more than 7x up from 8 years ago. It's almost $40 million, up just from the numbers we looked at last year just by adding developers this year for the first time and the $100 million up for '28 and a for all the good that Stephanie was talking about is not yet included in this number.

A massive opportunity moving forward. And we have been excelling and turning and disrupting this TAM significantly, pipeline income. You heard that from for Mark is really, really good. We have increased our opportunity in the market, exiting the first quarter of '26. Our carryforward pipeline is already up 40% year-over-year.

And as Mark mentioned, we have made significant progress in our sales organization over the last 3 years.

Productivity. You saw that from Mark is reaching all-time high, and we are not finished yet. This is where all the good stuff that Sam showed you if you look carefully at the skill files that went down their menus, they were -- a lot of them were go-to-market related. So we think we can significantly optimize our cost to book revenue moving forward, and you'll see this on a later slide.

So productivity is reaching new all-time highs. And this gives us the confidence really to continue to invest in quota carrying capacity, not only has it been increasing throughout '25 in the first and second quarter for this year. We forecast based on the confidence we have in pipeline creation and the momentum behind the business, the sales capacity is going to continue to increase for the remainder of this year.

You have also seen with Mark that the upward market on the enterprise side has been really successful. We are now -- we have 42% of Fortune 500 companies behind our network. And the fastest-growing cohorts are our biggest customers. And I think this is probably the reflection of the combination with products enterprise-ready and the account structures that Mark and team has put in place are the right structures.

We are seeing success in driving platform adoption. One of the reasons why you see [ DNR ] coming back up again is platform adoption. So the multitudes of products, the cohorts of products that customers have now and more products for customers, the largest cohort.

And at the same time, across the 75 products we have and sell, there's hardly any revenue concentration. So it's a widely diversified revenue footprint across the product portfolio that we have.

So finally, there is accelerating traction in the key product areas, including developers and as in 2025, ARR for Cloudflare 1. Our [ SC ] product suite is up 43%, and developer grew 137% and last year -- year-over-year in ARR. That's an important number. It shows the acceleration. It is also an important number to keep in mind when I try to give you context around what gross margin does and how gross margin will develop moving forward. So a significant acceleration in the key product areas.

Well, growth was easy. How do we think about profitability moving forward? As you heard us now say many times, we are focused on maximizing long-term shareholder value by investing in growth, while we remained committed to the underlying economics of our business.

Exiting 2025, we increased our unit economic margin to about 44% despite the declines in gross margin. So we'll come to that in a minute. We have also started to realize that the business has become complex. The dynamics that drive the business are very different. So we have to be way more granular in how we manage the business, how we allocate resources across the various products. And in order to give you some transparency how we think about this. I want to drill down more than a line item in terms of what the impact is on our P&L.

Let's start with gross margin. So there have been two primary headwinds in -- starting in the first quarter of '24. The first one is and we've talked about it, is the free-to-paid mix transition. So free traffic moves to paid and over the last 2 years, it has accelerated. But this is simply an allocation topic. Cost move from sales and marketing to cost of goods sold. And this is a wash on the P&L, but it's a headwind to gross margin. That is what you see in the blue colors. I think moving forward, we've bottomed out it's flattening out. We have also made some adjustment also a useful life of equipment. That is also flattening out.

The other mix shift is specifically coming from product mix. There's some little impact in there, like how contracts work, how the caps transition, how the pool of funds transition. But it's also a recognition that today, the developer product has slightly -- has a lower gross margin than the rest of the portfolio. And we'll talk about this in more detail what that means, whether that is important, where we make it up because how can it be the gross margin came down and unit economic margins still went up.

So while the gross margin might continue to hover at this level for a while. We expect the unit economic margin across the four acts to increase and to increase significantly. So let's talk about the economics for each one of the four acts and understand how it impacts our KPIs, what differs between the acts and how we manage the business taking into account these products, these acts are now very different in how they influence our revenue mix and different line items on the P&L.

So you might recognize this graph. So all the credit [ grills ] to our esteemed Chairman of the Audit Committee, [ Mark Hawkins ]. But -- he's inspired by him. But the unit economics define literally the growth and margin frontier. So the lower the cost to book are, the higher the line is on the Y axis, and the lower the cost to serve and the lower the attrition is the further to the right -- the line will cut the x axis and where the line crosses the x axis is exactly where the economic margin sits today at 44%. So now let's talk about the four acts and how they differ.

Let's start with act 1. So you're not surprised. You'd expect that. Act 1 is our original and largest business. So it most closely resembles our overall unit economics. It's the foundation for all of the acts that follow. As you heard from Rita, from Matthew and from Stephanie. It has a cost to book our average cost to serve our average. And because we have a higher mix of small- and medium-sized businesses, the attrition rate is slightly higher. So this frames pretty much where most of the business from a volume perspective is.

How does act 2 compare to that? Very different. You heard this before, to significantly benefit from the infrastructure we have put in place with act 1. It's like taking the infrastructure for free. We compared it to planes that are flying to the eyeballs empty, and we fill them up with data for -- on their way back for literally 0 to very low marginal cost.

So very low cost to serve, very low attrition, very sticky products, significantly higher cost to book. Longer sales cycles. This is where Mark's graph on increased partner share comes into play. We have to share revenue and margin with customers significantly higher cost to book.

Now act 3 as discussed earlier, our developer products currently carry a lower gross margin. That's it. Lower gross margin doesn't necessarily mean that the unit economics of the product are significantly lower. The combining impact of getting a platform business to scale is getting us to superior unit economic margins. And there are good examples for it. You all know them, right? If you look at what -- how the cloud is -- how the business develop, how AWS developed, how Microsoft Azure developed.

So platform -- and the reason for that is the platform businesses have this unique capacity that they can structurally shift their front here. And we've seen this now already happening in individual products in our developer platform portfolio.

Rita showed you good examples for it. The lower cost to run agents on our infrastructure, a result of the progress we have been making, taking costs down for us. So the curve is shifting. Scale is shifting this. One of the biggest levers is still in front of us. We are from retail that unpredictable CPU workloads for agents are the next bottleneck for AI deployment.

Well, if there's one thing that we are really good at, better than anybody else than it's making the most of our CPU capacity. So that's a significant lever that we are going to pull. So the curve is shifting up into the right. And at scale, we think that act 3 could surpass act 1 and act 2 from a unique economic perspective.

That is a good basis to start. And finally, there comes act 4. And you keep in mind what -- how Stephanie defined it, is that act 4 for us is a combination of act 1 plus act 3. So of all the acts we have of all the products we have, it will have, by far, the lowest cost to serve. It will have, by far, the lowest cost to book because the traffic is already on the network, and it will have significantly lower attrition costs. So this unit economic curve for act 4 could literally shift us off the chart from a performance and from a potential perspective.

So let's put all of this together and talk about what it means for our financial targets moving forward. So just for housekeeping, we are clearly on pace to surpass the $5 billion annualized revenue target before 2028. With the expanding TAM, you saw the momentum you saw on traffic and pipeline. This simply will be the next revenue milestone.

And as we execute on this massive opportunity of us. We have been talking about what guided us through all the years since the IPO. The North Star for us was what we call the rule of 40 is a very serious KPI for us since fourth quarter of 2020 on a rolling fourth quarter -- on a rolling average forecast, we have achieved 40 or above for 22 consecutive quarters. That's quite an achievement. So this is a very serious KPI for us.

How the year is shaping up, the momentum we are seeing from a pipeline, from a traffic, from a customer engagement perspective, allows us to see enough momentum that we exit the year, have clear line of sight to a rule of 50. And I think this is why we think it's time to raise the bar for us, for '27, we will move our North Star up to a rule of 50.

More importantly, we have already enough signals in the data we see that it will not take us another 9 years for that rule of north to shift. We think we will exit with enough visibility to push the rule beyond a rule of plus 50% exiting next year. And that is good news, but it takes into account everything that you heard today. So a very important milestone that we have reached today.

So we cannot talk about the Rule of 50. The impact and all the good news on the revenue side and keep our long-term operating model the same. So we are also now at a point where we're adjusting our long-term operating model. The first step we take is make the range of our gross margin wider, not to signal that we will fluctuate between 70 and 77 all the time. It is more a reflection of the journey we will go through over the next quarters recovering from where we are improving from where we are and getting back to where we came from, primarily by a lot of good engineering work that is happening, but also by a shift in revenue mix.

So with that said, where do OpEx fall in display. Sales and marketing will come down a bit compared to where we were before. This is a reflection of two things: it's a reflection of an increase in revenue share in act 3 and act 4 products that have significantly lower cost to book. But it's also the results of what Sam and Mark has been talking about.

In the past, we have been building go-to-market organizations where we scale the support of quota-carrying employees buy ratios. And these roles, these ratios are literally contracting and falling away. So we think we're moving forward already with the work that has started, we can significantly reduce the support infrastructure across support roles for us will not let those dollars fall to the bottom line. We will use it deploy it for more quota carrying capacity, which automatically means for every dollar that we continue to invest in sales and marketing compared to where we were before, we will see higher sales capacity, higher quota carrying capacity.

R&D will pretty much stay in the zones where we are today. However, we have some of the best engineers in the world. Now we give them the tools to be super productive. These are just three recent examples how supercharged the engineering organization has become in terms of deploying new core, rebuilding new products. So we think even if we keep the run rate at where it is today, the productivity we get out of our best engineers will jump and the flywheel that is driving the cloud innovation wheel will just continue to accelerate.

Where we will see the most significant changes from a productivity perspective is in the G&A line. So here, literally AI is transforming the G&A workflows that we have been operating with. This is where currently Sam's team is most busiest across the organization. I see this in my own finance organization now every day, how roles are collapsing, how functions are moving from manual and monolithic ERPs to agent-driven execution, whether that is account receivables, revenue recognition, even IR, how we trough our cues and case is impacted. So we think the frontier of where G&A costs and benchmarks are going to be moving forward will significantly fall and we will be leading that charge.

That means our operating margin targets will go up from plus 20% to plus 30%. And with that, our free cash flow margins. will significantly increase moving forward to about 30% to 35%. So quite a dramatic impact, I would say.

And last but not least, having all this in sight. Some of you are looking for that. They have been looking for it myself. We will achieve GAAP profitability latest by '28.

So in summary, I've been with the company for a while. It's the most exciting time to be with Cloudflare to have a front seat to what is happening in the world. the momentum and the opportunity in front of us is enormous. I'm proud to work with this team to make it all happen. And with that, we move to our Q&A. Thank you.

All right. Well, good afternoon, everyone. We have some time for open Q&A now. So -- and please wait for the mic before you do ask the question for the webcast, please.

Hey, good afternoon, thank you for all of the new details throughout the presentation. I would actually love to talk to you all about some of the infrastructure you've built internally with Cloudflare OS. And I would love to hear what are some of the technical challenges that you have to solve to get there what you think some of your peers are doing? Are there alternative approaches? And really, the question is, do you think that's something you can actually commercialize and monetize over the long term?

I'll start. Sam is here, who can -- and he should corner him to ask more specific questions. I think what I hear talking to customers is that everyone is facing the same set of challenges. And everyone is trying to figure out how to build it. What I think was unique to us is that we had the primitives to be able to go very, very quickly and build out what has turned out to be really a transformational tool.

We were using AI on our developer teams. And I think a lot of organizations are doing that. But expanding it to use it across the rest of the organization has been something that Cloudflare OS has really enabled that I don't see our peers actually building. And so whether that's Fortune 500 companies, whether that's new startups, every time we show this to people, they're like, that's what I want. I need that. Which again has us questioning we didn't think we were going to turn this into a product. But again, over and over and over again, that's been sort of the story of Cloudflare.

So any of your organizations are interested in something like this, we'd love to show you how it works. And right now, we're kind of just letting people play with it and use it, and it's been really powerful.

I think the really amazing core thing that enabled us more than anything else was actually -- and the sort of secret hidden benefit where I think everybody struggles is the fact that we started as a security company.

That allowed us to have hooks that were auditable and controllable into a bunch of different systems. And when I talk to peers, where they all -- the second question in every presentation is what's the security model of this? And because we had that infrastructure built in from the beginning because that was sort of core to the -- every primitive that we built, I think that's allowed us to develop the tooling a lot faster than anyone else.

I would say the biggest part of that momentum, and Sam can follow up with you afterwards is how do you get enough momentum on the skill files. How do you get the organization to document the jobs that need to get done? Because this is where the effectiveness of the outlook and the multiplication starts. Sam had this great idea with an e-mail alias where people -- employees literally could say, this is what I want. This is the skill I want to build. This is the skills I want to have literally in the beginning, we had humans building skill size just to get momentum going.

But we lied to them. We told them that a system behind the scene. People would ask more honest questions if they thought they were talking to a machine, not at least seeing their colleagues time, but really it was a team that was answering that, recording everything.

And then we started with what we call build sessions at every site where what Sam demonstrated literally got put on every laptop of every employee, and we help them think about how that tool can help them become more productive, how they schedule does, how to build their own agents, what are the most burning workflows that need to get automated.

I'll just add that when Matthew mentioned Cloudflare OS on the earnings call, we had a number of inbounds from customers and some of the trusted relationships we have with long-term customers. I've stopped taking count of how many of them want us to do this for them.

Thanks so much for hosting, Saket Kalia at Barclays. I'd love to dig in to act 4 for a little bit. There was a chart, I forgot which presentation that showed some websites where traffic is down precipitously. One of them was PC magazine. And remember, we're not too far ago kind of thinking, "Oh, which laptop am I going to buy" and going to PC Magazine looking at that when was the last time we ever did that, right? We just looked for the answer. We just kind of get it like that. Right, we just kind of all want the answer immediately.

But I'm curious, maybe, Matthew, Mark, anyone on the panel that wants to chime in, how are your conversations with those content creators going now? Because it feels like such a pivotal moment. And when does that start to unlock the value that you could provide with act 4. It feels like that should inflect at some point, when does that happen?

Yes. So first of all, the charts show that what human traffic is down. You look at the overall traffic to that -- those sites, it's up massively, but that additional traffic is agentic traffic. And so I think the key question is to figure out what's the fair value exchange for agents and the people who are feeding information into those agents. And that's what we're figuring out.

We're seeing early success where once they create, as Stephanie talked about scarcity in the market, you can't get to my content unless you make a deal with me. Instantly, you're seeing the largest publishers that are out there striking much bigger deals. And they're all then talking to each other, which means that they're all then -- it's become this virtuous cycle where they say, wow, we've been able to strike much better deals with the AI companies because we're using Cloudflare and they're doing that.

I think the next stage though, is how do we make that more automated and systematic across the infrastructure. And so I think over time, there's going to be a lot of different ways of doing that.

Yes, there are going to be content deals. There are also going to be things that just so we say, every time you access 1 of my pages, you have to pay me a fraction of a penny. And at first, that's going to start out as a relatively small amount. But if you just see the growth in agentic traffic, and you forecast that out, that turns into significant models. What I think is going to change is what content is the most valuable.

So if you look at the media industry historically for the last 20-plus years, the way that you built a media company was you told basically the same story that everyone else told, but you defined a very specific audience for it. I remember sitting with an executive of a kind of fast-moving media brand that you've all heard of. And this person was bragging about how they created articles as inexpensively as possible and then wrote the most incendiary headlines to kind of piss off their audience as possible so they would click on them, and that was the way that they drove that traffic. I think that business model for media is in real trouble. I don't think that's going to exist.

On the other hand, like my wife and I own the local newspaper in our hometown. I think we'll make more money off AI licensing deals this year than we do off online display advertising. Why? Because we own the newspaper in Park city, Utah. If you guys are going on vacation in Park city, Utah, you want to know what's the best new restaurant, what's the best hotel to stay out, what concerts are going on. We're the exclusive information for that.

And so interestingly, I think what's going to change is that the media companies of old may get very much disrupted. They may be something different. Some of the media that has disappeared over the last 20-plus years, local media, very specific things, unique content actually might be much more valuable in what the information age of the future is. And exactly -- I think there's going to be a lot of change. But I think what I'm optimistic about is that if we figure this out, traffic has always been a terrible proxy for value. For the first time, we can actually reward content creators for creating content that has actual real value that accelerates human knowledge.

Fatima Boolani from Citi. I appreciate you hosting this. I wanted to ask you about Cloud managed agents. I think all of us have lived through seven lifetimes of announcements from some of the printer labs. And what was striking to me is almost all of them seem like and probably are your customers.

So Matthew, this is probably best suited for you. And then please feel free to chime in because there's a commercial implication of my question as well. Cloud managed agents versus what your purpose building at the edge for agents to be persistent, agents to be ephemeral. Where is kind of the line in the sand as it relates to cooperation, coopetition and competition. And ultimately, where will Cloudflare shine? Is it going to be the orchestration brain? And how do you get paid for that when you talk to customers who are thinking about building autonomous workflows to run their businesses from here on out?

Yes. I mean I think that we are incredibly closely partnered with all the major frontier AI labs. And so if you look at the latest open AI models, they've all been trained on what we call our code mode systems. We work very closely with Anthropic in terms of thinking about their future of managed agents looks like.

And what we believe is that if you are the place that has the highest performance the lowest cost to serve that you're going to be the place where that code naturally runs.

And so very much what you're seeing right now with like Open AI's recent announcement in terms of where their code is shipping. Cloudflare is the default where it goes to. I think that's going to become more and more the case across these different things.

And again, the AI companies are really good at actually building AI systems. But we're really good at then running the output of the code from those systems extremely effectively.

And what I think people don't understand and Rita's talk is actually worth going back and rewatching is it's not just code like the way you think of developers writing code. If you ask an AI agent to plan your next vacation, it needs to write code.

If you ask it to pick what restaurant you should go to or who you should invite to our next birthday party, the way that AI works best is if it's actually generating code. And so every single request that in the future is going to have to spin up some code in order to do its work and the most effective way to do that is with isolates technology, and we are the only ones that have it.

Adam Borg with Stifel. Thanks so much for the time. I appreciate the discussion today. So Matthew, I thought your comment with your fireside chat would Phil earlier was interesting around how AI is really driving demand for your Zero Trust platform. So maybe a 2-part question.

First for Mark, maybe give us a state of the union of the SASE Cloudflare 1 in the field. Obviously, the hour growth you talked about, Thomas was really impressive.

And then two, maybe for Matt you talk a little bit more about how AI is drilling drives demand and how that should play out in coming quarters and years?

Yes. I mean from a security standpoint, I think we made the point earlier that as you start to build and proliferate these agents, you need a security framework that's very differentiated from legacy point product technologies. And so I think the opportunity for us to have a conversation that traverses agents into security, into global network performance is super easy for us. It's a lot harder for other people.

So we've seen a lot more mindfulness of the rating acceleration towards Zero Trust model. It's funny. We're going to have our Customer Advisory Council here tomorrow. I started, I think the week we had on 2.5 years ago in New York, and we surveyed customers at the end of every CAC and say, what's your big strategic opportunity this year for your company and 75% of them said Zero Trust.

1 year later, same survey 65% of the people said Zero Trust. Like Zero Trust is a journey. It's not a point in time in which you buy some stuff. It's a journey to aspire towards. And I think AI and agentic deployments really accelerate that?

Yes. I think that -- I mean, we're really honest with ourselves. When we were selling Zero Trust to people who are looking to buy Zero Trust. The challenge was like we were late to the game. Like there are a whole bunch of people that had kind of dominated the category. And so we were good if we could get in the conversation, but we weren't always in the conversation.

And so I think what's changed is really two things. The first is the conversation started in a different place. The conversation today starts with you're trying to build an AI system to do effectively what Cloudflare does and whether you're adopting that or you're adopting your own version of it. Now you have to have a security plan that actually makes that happen. Oh, it just so happens that's what our Zero Trust platform does. And it's built very much for that.

And so instead of the conversation starting with, let's go sell you Zero Trust and you're starting with the CSO Instead, you're starting with the CIO, the CTO, or even the CEO, who's saying, I need to be able to do this, but I need to be able to do it in a secure way. And we've got the complete package to do that. That's how I think the Fortune 500 are coming to us and where we're winning those deals much more. And we're any more of the conversations because they don't start to 0 trust conversations. They actually started AI conversations where Zero Trust is kind of the killer app to make it and enable it to work.

I think the other piece is we're really the only zero-trust platform that has a self-service option. And so when things like open claw came out and everyone said, "Oh my gosh, how powerful it is if we give it access to everything. And then everyone said, "Oh my gosh, how scared it is have we given access to everything everyone then quickly search for what's a solution, and they figured out that we were this really powerful and compelling solution.

And so for a long time, the average consumer didn't care about Zero Trust things like Open Claw have made them care about it. And as a result, you're getting much more adoption and then those people are bringing themselves to work. That's a very natural kind of cloud for like sales pattern that we had not usually been able to tap into, but now we're more able to tap into it today.

Ryan MacWilliams, Wells Fargo. 10 to 100x Internet traffic over the next 5 years with AI agents. It seems more and more realistic each day. Is there anything you would need to change or reorient or invest in as the Internet will move towards that level of activity in the future?

Yes, and by the way, if you challenge me to take an over/under bet on that, I'd probably pick the over. Like it's growing so fast, like we usually feel like we're pretty good at predicting the future. This is growing much faster than we ever anticipated. So obviously, it's going to require investment into what is there. And then the question is how do we pay for the investment. And that's very much what Stephanie was talking about with to figure out as that traffic happens, who's paying for the infrastructure in order to support it? And the answer can't be -- it all goes back to the content creators, who all of a sudden aren't generating any revenue anymore.

So again, I think as this natural cycle where everyone is like, we have to create some new version of this. And then I talk to the leaders at all of these different AI labs, they're like, listen, we're happy to pay as long as you treat us fairly as long as everybody else pays as well. And I think that we're getting to a point where we're going to be able to do that, and you're even seeing folks like Google who are very much moving away from a traditional search model in that process, they're much more saying, okay, there's got to be a different way of having a value exchange with the people that are creating the content. But I think, again, this -- we might be on the cusp of a gold native content creation. Because if we figure out this way to actually reward real knowledge creation as opposed to just what generates traffic, I think that's actually a really optimistic view of what the future can look like.

Matt over here from RBC. Thanks for doing this. And by the way, Mark, great run. It's really been turbocharged...

Thomas, a lot of good information there. We'll give one to you. I thought the way that you described the different acts from a cost to book, cost to service was great. When we think about that gross margin range that you presented, I think it's helpful context that you sort of provided, but I wanted to double click on that. So in other words, we should expect maybe a little bit of pressure on that in sort of in the near term and then an upward trajectory towards that? Or how should we progress in that range?

Today, I would describe it best as a journey where we are, and we will improve from where we are. Will it be a straight line, and we might have a dip again? Could happen, but it would come with the benefit, right? A dip in gross margin would mean that something else accelerate significantly, either growth rates or operating margin because a significant step-up -- for us to dip even more, it would need to be a significant amount of additional revenue on top of the acceleration of the growth line we already have projected, right?

So that additional revenue with maybe pressure on the gross margin would come with other benefits. Our path to on the rule of whatever, north of 50% would accelerate. Our operating margin would increase as we have already seen. Apart from that, I think it's a journey that takes us from where we are, to back to where we come from. And we are public I can say that you will see already a stabilization in the second quarter results.

And usually, the fear with gross margin pressure is that you're seeing more competition. We are not seeing more competition. In fact, we are seeing a place where we are oftentimes selected as a category of one and that's the case where the gross margin pressure is really coming from is the fact that we've just seen so much growth in our developer platform. And again, as you see, that's actually a really healthy thing even though it is at a lower gross margin.

I think one of my favorite parts of the presentation that was Rita's presentation around we're not going to be able to CPU and VMware way out of this. When do you think that reality sort of hits the enterprise? And maybe it already is with developer growth being 137%. But in terms of breaking that inertia, breaking that modality that those people have when it comes to building net new applications, when do you think we hit that breaking point?

I think the -- so I think that we're seeing that with companies, we're seeing that realization with companies that are really adopting AI. So the big hyperscale or the big AI labs, the reason that they're adopting our platform, the reason they're using the things that they're doing is because they see that it doesn't work to do it the old way. I think you're still going to see. A lot of the reasons why AI adoption at large companies is still more of an experiment that a really embedded system is because they don't see how to do it. The security model seems too hard. It seems too costly based on the token spend. It seems too costly based on all these VMs that you have to stand up.

And you hear even digital native companies like Uber coming out and saying, "Oh, I don't see what the return is here." There's a better way to do it. We can show them the better way to do it. That's part of the reason why like we may just give Cloudflare OS away for free, just to show people, here's a way you can do what you want in a cost-effective way, where we've been able to significantly lower the cost of using AI, while enabling more of our team to do it. So I think what's really going to drive it is when you see more and more of those companies say, "Okay, I really have to get on board? And is that going to happen in the next quarter, in the next year, in the next 5 years?

I think faster than that because you're going to see that adoption expand. But I think that's going to be the real thing where people say, I can't do it the old way. I have to shift to a new model, and we're the natural place to provide that new model.

Well, we're out of time, everyone. We actually have to get down to the closing bell. Yes. Maybe you want to take one more?

I said the market.

I said, we'll, wait. I was like -- but we have 15 of our longest tenured Cloudflare that are going to be joining us. Many of them who are actually in this room to ring the closing bell and we also have a bunch of members from Cloudflare New York office joining us as well. So thank you for your time today. I certainly learned a lot. I hope you did too, and we'll be around afterwards to chat. You know where to find me. So thank you, everyone, today up here for joining us.

My name is Jay, and I will be your conference operator today. At this time, I would like to welcome everyone to the Cloudflare First Quarter 2026 Earnings Call. [Operator Instructions] I would now like to turn the conference over to Phil Winslow. You may begin.

Thank you for joining us today to discuss Cloudflare's financial results for the first quarter of 2026. With me on the call, we have Matthew Prince, Co-Founder and CEO; Michelle Zatlyn, Co-Founder and President; and Thomas Piper, CFO. By now, everyone should have access to our earnings announcement. This announcement as well as our supplemental financial information may be found on our Investor Relations website. You may be making forward-looking statements during today's discussion, including, but not limited to, our customers, vendors and partners, operations and future financial performance, our anticipated product launches and the timing and marketential of those products.

Our anticipated future financial and operating performance, and our expectations regarding future macroeconomic conditions. These statements and other comments are not guarantees of future performance and are subject to risks and uncertainty, much of which is beyond our control. Our actual results may differ significantly from those projected or suggested in any of our forward-looking statements.

These forward-looking statements apply as of today, and you should not rely on them as representing our views in the future. We take no obligation to update these statements after this call. For a more complete discussion of the risks and uncertainties that could impact our future operating results and financial condition, please see our filings with the SEC as well as in today's earnings press release. Unless otherwise noted, all financial numbers we talk about today other than revenue will be on an adjusted non-GAAP basis.

You may find a reconciliation of GAAP to non-GAAP financial measures that are included in our earnings release on our Investor Relations website. For historical periods, a GAAP to non-GAAP reconciliation can be found in the supplemental financial information referenced a few moments ago. We would also like to inform you that we'll be hosting our Annual Investor Day on Tuesday, June 9.

Now I'd like to turn the call over to Matthew.

Thank you, Phil. We had a very strong start to 2026. We achieved revenue of $639.8 million, up 34% year-over-year. We now have 4,416 customers paying us more than $100,000 per year, a 25% increase year-over-year. Revenue contribution from these large customers grew 38% year-over-year, contributing to 72% of revenue during the quarter, up from 69% in the first quarter last year.

Our dollar-based net retention was 118%, down 2% quarter-over-quarter and up 7% year-over-year. Our gross profit margin was 72.8%, and we delivered an operating profit of $73.1 million, representing an operating margin of 11.4%, and we generated strong free cash flow of $84.1 million during the quarter, again exceeding expectations. The strong momentum we've seen in our business continued to build through the first quarter.

Some highlights. Sales productivity increased year-over-year for the ninth consecutive quarter. Growth in hiring sales force capacity also accelerated in the first quarter, increasing at the fastest pace since 2023. Deals over $1 million were up 73% year-over-year, the fastest growth rate since 2024. We added a record number of our largest customers in the quarter, those spending more than $5 million with us annually.

In fact, we added as many $5 million-plus customers in Q1 as we did in all of last year. Bookings from new customers increased at the highest rate since 2023. New pipeline generation grew sequentially at the fastest pace in 5 years, and we exceed our planned target by more than any other first quarter since 2021. Our quarterly gross retention reached its highest level in 4 years, reinforcing that customers understand cloud lowers a must-have rather than a nice to have.

We are a significant beneficiary of many of the most powerful trends across the economy. To give you some sense, we had 1 million new developers in just the last quarter. Our products were made for this moment, and we are helping our customers build the future on our platform. That's a good segue to talk about some of our customer wins in the quarter.

A leading technology platform, expanded their relationship with Cloudflare, signing a 2-year $10 million pool of funds contract with initial use cases for application services at our workers' developer platform. With our full portfolio now unlocked under a single rate card, we won workloads from both a hyperscaler as well as point solution competitors. Looking ahead, we are also in discussions on AI paid for Crawl to control and help monetize AI bot traffic. .

A rapidly growing technology company in APAC expanded their relationship with Cloudflare signing a 2-year $8.7 million contract for application services and our workers developer platform, driven by the boom in AI-powered by coating, this company has seen explosive growth, and Cloudflare has become core to their infrastructure, intelligently routing billions of daily requests across the globe. This customer chose Cloudflare over a competitive bid from a hyperscaler due to the strength of our unified platform and our seamless low-latency security.

A Fortune 100 technology company expanded their relationship with Cloudflare signing a 2-year $8 million contract for our privacy proxy solution, the fifth privacy engagement with this customer, solidifying Cloudflare as their go-to privacy partner. They approached us with an urgent need to handle massive scale with precise geolocation accuracy for user-initiated agentic traffic. We delivered a fully operational solution within 1 week, demonstrating the speed, trust and engineering depth that continues to set us apart.

A leading insurance company in EMEA expanded their relationship with Cloudflare, signing a 5-year $5.1 million contract for application services and our full SASE portfolio. Driven by years of acquisitions, this customer's IT environment and bloated to over 600 vendors with some employees literally juggling up to 4 laptops to access essential applications. By standardizing on Cloudflare, they displace 6 legacy vendors at signing with 10 more displacements already underway, targeting over $1.3 million in annual savings.

Their CTO put it simply. He wanted a high-performance "Formula 1" level architecture with Cloudflare as the engine. The Fortune 500 aerospace and defense company expanded their relationship with Cloudflare signing a 3-year $5 million contract for our Zero Trust products, including browser isolation, access and gateway. After a major security breach force this customer to move from on-premise hardware to the cloud, they discovered that their first-generation Zero Trust vendors browser isolation solution could not meet critical government compliance requirements.

Putting $5.5 billion in government revenue at risk. Cloudflare delivered a fully compliant solution in a matter of weeks where the incumbent could not. A leading AI company expanded their relationship with Cloudflare, signing year 1 million contract for application services. As 1 of the most visible targets for cyber attacks globally. This customer needed a security layer to protect their massive infrastructure build-out, despite a strong build over by mentality, they chose Cloudflare trust a battle-tested network that has proven its resilience against the largest attacks.

This is a customer that moves fast and pushes boundaries and they're already testing our AI gateway for their AI workloads. Another leading AI company expanded their relationship with Cloudflare signing a 10-month $2 million contract for Argo Smart Routing coming just 1 quarter after signing a workers' developer platform deal. This customer wants to be the fastest and most reliable AI provider in the market, and Cloudflare is delivering.

After deploying Argo, they immediately reduced their average global latency by 30%. And -- in the AI space, that kind of speed is a real advantage that our hyperscaler competitors simply can't match. In nearly every customer conversation, it's clear. the emergence of generative and agentic AI is not just redefining the economics of the Internet and software companies, they are redefining the business models of all companies fundamentally reshaping how organizations are structured, operate and create value.

At Cloudflare, we don't just build and sell AI tools and platforms. We are our own most demanding customer. AI and agents are no longer pilot projects at Cloudflare. They are now core parts of our workforce. It's been an interesting journey. We've been selling pick and shovels in the AI gold rush for the last 4 years, but we ourselves were cautious users wanting to ensure there was real ROI before making significant investment.

We afforded a lot of the performance of AI some companies engaged in Internally, the tipping point was last November. At that point across our teams, we began to see massive productivity gains. -- team members that were 2, 10, even 100x more productive than they had been before. It was like going from a manual to an electric screw driver. Cloudflare usage of AI has increased by more than 600% in the last 3 months alone.

For team members in R&D, 97% use AI coding tools powered by the same workers' developer platform we ship to our customers and 100% of their contributions to our production code bases are now reviewed by autonomous AI agents. I think across the industry, you're about to see a massive uptick in reliability every code or configuration range can now have a tireless and uncorrelated set of eyes trading on every incident from the last 10 years, checking to avoid problems.

At the same time, the impact on developer velocity is clear. We've never seen a quarter-to-quarter increase in new co-generated bug squashed and technical backlog burn down like we did last quarter. It's been wild. Beyond the product and engineering, employees cross-cloudflare from HR to marketing run thousands of AI each day to get their work time.

Those agentic workflows rely on dozens of MCP servers to reach data and systems of record and use hundreds of centrally managed skill files as well as many more that have been created and shared within individual teams. The hardness that we've built, which we call Cloudflare OS allows teams across the company to quickly get up and running. We've asked our team to think what the fundamental job to be done is and then reimagine how we can make the work to achieve it more efficient, reliable and joyous. .

At Cloudflare, the way work is done has fundamentally changed. That means being intentional in how we architect our company for the agentic AI era in order to supercharge the value we deliver to our customers and honor our mission to help build a better Internet for everyone, everywhere. As a result, we announced significant actions this afternoon to further accelerate our evolution to an genic AI-first operating model. That unfortunately means saying goodbye to teammates who have contributed to building Cloudflare where we are today, resulting in a reduction of the size of our team by more than 1,100 people.

This decision is not a reflection of the individual work or talent of those leaving us. They were critical in getting us to where we are today. Instead, we are reimagining every internal process from engineering to finance to sales to run on an agentic AI backbone on our workers' platform. This isn't a cost-cutting exercise or an assessment of the individual's performance. It's about defining how a world-class, high-growth company operates and creates value in the agentic era.

Deciding to part ways with teammates is the hardest part of this decision and is the responsibility of the entire senior leadership team at Cloudflare takes personally. We believe that acting with empathy isn't about avoiding hard decisions, but rather about how you treat people when those decisions are made. If we are asking our team to be world class, we have a reciprocal obligation to be world-class in how we treat them.

By taking decisive action now, we provide immediate clarity to those departing and protect the stability of the team that remains. We are also pairing the directness of these measures with severance packages that lead the industry because we want to ensure that those who have invested their time and talents and Cloudflare's mission are taken care of as we move into the next phase.

It's the right thing to do. It's the honest thing to do, and it reflects the values of the company we are continuing to build. On a personal note, this has been a hard day. A number of friends will no longer be colleagues, but I'm confident they will land at other great places and bring with them a set of skills they learn, building Cloudflare to where we are today.

The group leaving us will build many future great companies, and I'm confident that our reshaped organization will be even more nimble and innovative as we continue to build the future, not an easy day but the right decision. .

With that, I'll turn it over to Thomas to walk through the numbers. Thomas, take it away. .

Thank you, Matthew, and thank you to everyone for joining us. Before I begin my customary remarks and our results for the first quarter, I would like to provide additional details on the actions we announced this afternoon to accelerate Cloudflare evolution to an agentic AI-first operating model. .

Cloudflare's history proves our business model innovation is as important as our technical innovation. These 2 forces don't sit side-by-side a Cloudflare rather, they compound on each other in ways that provide us with meaningful competitive advantages and create significant value for both our customers and Cloudflare. AI is driving a fundamental replatforming of the Internet as well as a paradigm shift in how software is created and consumed.

And is shaping up to be the biggest tailwind for both our network and our workers developed a platform that we've ever seen in Cloudflare's history. From this position of strength, we are again applying the same winning formula of compounding technology innovation with business model innovation by fully embracing an agentic AI-first organizational structure and operating model, as through Cloudflare revenue scales, our efficiency and productivity will scale even faster.

Unfortunately, this decision means studying ways with colleagues who have helped build a strong foundation Cloudflare stands on today. resulting in a reduction of the size of our team by approximately 20%. These reductions are across all functions and geographies and reflect how broadly AI is accelerating our operational velocity.

Importantly, however, we continue to expect growth in the net capacity of our quota carrying sales force to accelerate in 2026 with today's actions compounding productivity to fuel our growth. These actions will result in severance and other restructuring charges of $140 million to $150 million for full year 2026.

Approximately $40 million of which is noncash. With the majority concentrated in the second quarter. Our expectations for free cash flow for 2026, however, remain unchanged with approximately 25% to 30% of full year cash generation in the second and third quarters. By decoupling our ability to scale from the traditional dependency of the past, Cloudflare will be structurally faster, more innovative, more productive and more efficient.

Now turning to our results. The first quarter was a strong start to 2026. With momentum building across multiple areas of our business. We continue to see rapid growth from AI and the agentic workloads across our network strength in our largest customer cohorts continuing returns from our go-to-market transformation and rebid adoption of our workers developer platform.

Total revenue for the first quarter increased 34% year-over-year to $639.8 million. From a geographic perspective, the U.S. represented 49% of revenue and increased 34% year-over-year. EMEA represented 28% of revenue and increased 31% year-over-year. APAC represented 15% of revenue and increased 34% year-over-year.

Turning to our customer metrics. We ended the quarter with roughly 4,400 large customers, representing an increase of 25% year-over-year. Revenue contribution from our largest customers was 72% of revenue during the quarter, up from 69% in the first quarter last year.

We again saw significant strength in our largest customer cohorts, including those that spend over $5 million with Cloudflare annually, which grew 50% year-over-year and added a record number of additions both quarter-over-quarter and year-over-year.

Our dollar-based net retention was 118% during the first quarter, down 2% sequentially and up 7% year-over-year. As we've noted previously, there can be some variability in this metric quarter-to-quarter, with growth this quarter driven by a meaningful acceleration in business from new customers which grew at the highest rate since 2023.

Moving to gross margin. First quarter gross margin was 72.8%, representing a decrease of 210 basis points sequentially and a decrease of 130 basis points year-over-year. Paid versus free traffic on our network continued to grow both year-over-year and quarter-to-quarter, again, driving additional allocation of network costs from sales and marketing into cost of revenue.

Our workers developer platform products, which currently carry a lower gross margin than our corporate average, delivered another quarter of significant growth. In fact, developers on our platform increased to more than 5.5 million at the end of the first quarter, an increase of 1 million developers in a single quarter as compared to an increase of $1.5 million in all of 2025.

While our developer products are not yet as optimized on gross margin, they also have a lower cost to book. And we will continue to focus on driving further efficiency improvements as our developer product scales. While gross margin may continue to trend down in the midterm from these dynamics, the scalability and efficiency of our network remains intact, and we expect our unit economic margin will continue to increase.

Network CapEx represented 9% of revenue in the first quarter. As a reminder, there can be some variability in this metric quarter-to-quarter, and we expect network CapEx to be 14% to 15% of revenue for full year 2020. Turning to operating expenses. First quarter operating expenses as a percentage of revenue decreased 3% year-over-year to 62%. Our total headcount ended the quarter at approximately 5,500.

The majority of new hires during the first quarter were in sales, with a particular focus on continuing to add quota-carrying account executives. Sales and marketing expenses were $227.5 million for the quarter. Sales and marketing as a percentage of revenue decreased to 36% from 38% in the same quarter last year. Research and development expenses were $101.5 million in the quarter. R&D as a percentage of revenue remained consistent at 16% compared to the same quarter last year.

General and administrative expenses were $63.6 million for the quarter. G&A as a percentage of revenue decreased to 10% from 11% in the same quarter last year. Operating income was $73.1 million, an increase of 31% year-over-year compared to $56 million in the same period last year. First quarter operating margin was 11.4%, a decrease of 30 basis points year-over-year.

Turning to net income in the balance sheet. Our net income in the quarter was $94 million, or diluted net income per share of $0.25. Free cash flow was $84.1 million in the quarter or 13% of revenue compared to $52.9 million or 11% of revenue in the same period last year. We ended the first quarter with $4.2 billion in cash, cash equivalents and available for sale securities.

Remaining performance obligations, or RPO, came in at $2.543 billion, representing an increase of 2% sequentially and 36% year-over-year. Current RPO was 64% of total RPO includes 34% year-over-year.

Moving to guidance for the second quarter and full year 2026. For the second quarter, we expect revenue in the range of $664 million to $665 million, representing an increase of 30% year-over-year. We expect operating income in the range of $90 million to $91 million. We expect an effective tax rate of 21.5%. We expect diluted net income per share of $0.27 assuming approximately 377 million shares outstanding.

For the full year 2026, we expect revenue in the range of $2.805 billion to [ $2.817 ] billion, representing an increase of 30% year-over-year at the midpoint. We expect operating income for the full year in the range of $418 million to $421 million. We expect an effective tax rate of 20.5% and -- we expect diluted net income per share over the period to be in the range of $1.19 to $1.20.

We expect approximately 375 million shares outstanding. In closing, the first quarter has set a strong tone for the year. Our strategic position heading into this paradigm shift of the agentic Internet has never been stronger, and the opportunity ahead of us is larger and more defined than at any point in our history.

We remain committed to capturing it with disciplined execution, durable growth and long-term focus. Before opening the floor for questions, I want to again acknowledge our colleagues who will be departing Cloudflare as we move into our next chapter. They will always be part of the Cloudflare story, and we are sincerely grateful for their service to our customers and their commitment to our mission.

With that, operator, please poll for questions. .

[Operator Instructions] Your first question comes from the line of Matthew Hedberg of RBC Capital Markets.

Great. Matthew, first, on your strong Q1 results, I find it interesting that some of your Act 1 competitors don't seem to be benefiting from monetizing agentic traffic the same way you are. I guess First of all, why are you seeing some strong tailwinds there?

And then as a follow-up regarding the announced restructuring, really in light of the strong results, it seems to really be coming from a position of strength I guess the question is why now? How is it going to make cost for stronger? And Thomas, have you embedded any conserves in the guide for this action?

Yes, Matt. Maybe I'll start with the second because again, I think it's I think sort of the honorable thing to do. This wasn't an easy decision but it's the right decision. We've just seen that there are roles at Cloudflare that just aren't the roles that we need for the future. I think just because you're fit doesn't mean you can't get fitter.

And so I think that what we've seen over -- especially of course in the last 6 months, is just the productivity that we've been able to gain from the people who are directly talking to customers and people who are directly creating code that has been just such an incredible efficiency gain and a lot of a lot of the support people that have provided support behind them, that's -- those roles aren't going to be the roles that drive companies going forward.

I think we've always looked a little bit in the future. And I think that you're going to see companies around every industry starting to realize the gains that they can get from these tools. And in the process, I think that that's going to change companies pretty dramatically. We are, I think, early beneficiaries of that. I think that we will continue to -- we believe in people. I think that we will continue to hire people and we'll continue to invest in them because the people that are embracing these tools are just so much more productive than we've ever seen before.

And I would guess that in 2027, while more employees than we did at point -- but the roles are changing dramatically, and you've got to do something dramatic in order to make that shift, and that's why this was the right time. In other words, we're the fittest we've ever been but we're going to get even fitter to win the next chapter.

To your first question about traffic, I think the key here is something that we've always understood, but I'm not sure the market has understood as well, which is not all traffic is created equal. A lot of the times, if you were sort of a traditional CDN customer company. What you're trying to chase was things that drove lots of bandwidth, so video streaming of live events and things anything like that. And again, that was an okay business. I think you had a lot of pressure. It was largely a commodity, but that was kind of the game of being a CDN.

We never saw ourselves that way. We thought what we wanted to get in front of was the most essential traffic that's out there, not bandwidth and video streaming, but APIs and applications and seeing that. And I think that in this new world of agentic commerce agentic transactions, -- our approach is showing kind of the wisdom and durability of that.

So today, literally, we're seeing hundreds of billions of agentic requests per month and that number is growing exponentially. They are interacting with us, and we are setting with the rails and the guardrails are for that and that's driving our Act 1 business. And then on the other side, in our Workers platform, we have built a platform that allows you to build agents that are just significantly more efficient than anyone had before.

And so across all of the parts of our business because even in the Zero Trust and SASE space, it turns out that having more fine grain control about data is exactly what you need if you have of these somewhat new agents running around doing things, you want to make sure that they only have access to the things they should. It's -- I wish I could say that we saw all this years ago and built Cloudflare for it. But I think that the reality is that we happen to have built exactly the right set of tools for this moment. And I think that's what's separating us from some of the people that we sometimes get compared to.

Yes, let me take your second part regarding guidance and how this action is reflected. First of all, I would say we've been rather thoughtful. And you heard in my prepared remarks that while this action pretty much affects all teams at Cloudflare. The only exception really is our AE and quota-carrying capacity that is sitting in front of the customers. We hardly touched that -- and I think we've been careful trying to reflect whatever residual risk remains in the guidance that we provided. -- for the remainder of the year. So though I think we -- as usual, we try to be thoughtful and prudent in how we think about what is in front of us.

Your next question comes from the line of Adam Borg of Stifel.

Awesome. Maybe, Matthew, for you, 1 of the things that we keep on hearing about is just how AI costs internally are really expensive, especially around R&D coding agents internally. And so -- how do you think about balancing R&D adjusted coating, adoption with the cost? What AI efficiencies are you looking to see across the organization? And of course, how much of this risk is to offset some of what I have mentioned.

Yes. So first of all, we have seen as used has gone up 600% in the last quarter. We have seen costs go up. And I don't think it's got nearly as much as some others. And that's driven by a number of things. I think the least important is in terms that most of the big AI labs are Cloudflare customers. And so we have very good relationships with them. We work with them.

We have ways of making sure that we can get the best pricing in the best models from them. I think more importantly, though, is a lot of times, we're able to run those models instead of on their infrastructure on our own infrastructure. And so -- we have a fleet of GPUs, and we have all of the tools with cloud floor workers and workers AI to be able to build and use those tools themselves. And so most of the use of various AI coding tools, isn't even leaving our network.

It's running on our infrastructure because we're very good at routing to wherever there is capacity we're able to get a lot out of that. And so I think that's 1 of the reasons why we see a significantly higher utilization across our GPU resources than some of -- than any of the hyperscalers and then any even of the AI labs are able to drive.

And then when we built what we call Cloudflare OS, we've paired that with our AI Gateway product. An AI Gateway product allows you to route different requests based on what's the right model for the right task. And so -- that means that if we have a task which we can evaluate as being relatively simple, then we can wrap that to a model that might be running on our own infrastructure and be able to be delivered at essentially no marginal cost to us.

Whereas if we have to think it is more important, we might send that off to 1 of the frontier models. -- and pay more for that. I think that, that model -- again, I feel like we're living a little bit in the future. I think that's what you're going to see a number of companies, Danas we demo Cloudflare OS to other companies what we're seeing overran by CIO after CIO after CIO is, well, we want that, too.

And so we already have kind of the strip down version of that with Cloudflare as AI gateway. But I think you might see us increasingly take some of the tools that we've built internally and actually make those available to other companies. And that's actually very normal for cluster. Almost every product we've had that's been successful, start is something that we needed ourselves and then it turns out that what we built for ourselves is valuable to others as well.

That's really helpful. And maybe as my follow-up, Fortinet last night talked a lot about this opportunity that they're seeing in soft and SASE -- and when I think about cloud player and just your global network and reach and data residence requirements globally, how do you think about the data localization opportunity, sovereign opportunity, not just in AI, but across the ...

Yes. I mean I think that we're uniquely positioned for a world where they're increasing regulatory or even just practical requirements around keeping data in particular jurisdictions. We were present in more than 120 countries worldwide, more than 350 cities. And we're designing all of Cloudflare systems in order to really think of data as being able to just stay wherever your permissions are.

So if you're a customer that really cares about I need to make sure that all my data stays in Germany because I'm a German, I don't know, a health care provider, then we can actually set that up now where your data can actually stay resident in Berlin, Frankfurt, Munich, the various data starts that we have inside of Germany. And we can do that with a level of granularity that no hyperscaler can match.

And so I think that we've -- that will increasingly become an advantage of Cloudflare's network because of what we're doing. And then if you layer on top of that, the Zero Trust and SASE tooling, that's going to allow us to make sure that agents can only get access to the things that they have permission to get access to. And I think that that's actually going to be a bigger tailwind to that space. And again, we won't be the only winners there. But I think across the board, you're going to see a lot of these principles that sort of the most forward-leaning companies had adopted previously.

That those will become much more standard with what people are doing with things like open cloud. I mean it's amazing how many -- how the fact that we have a self-service version of our Zero Trust products as people are experimenting with things like Open Cloud, they need to have that find grain data control, and we're basically the only game in town to deliver it.

Your next question comes from the line of Saket Kalia of Barclays.

And nice start to the year. Thomas, maybe for you. I know the growth in different acts of course, has different impacts on gross margin. And of course, we spoke about sort of the proactive optimization here in the business as well. As we get through those changes here, I think you said by the end of Q3, how do you think about sort of the net impact -- how should we sort of think about gross margins as those other acts continue to grow?

Yes. It's a topic we've been talking about for a while that the margin structure is different across margin structure is different across the various acts with the developer products being the gross margin weakest. But we also said that despite the fact all products are pretty much equal when we look beyond gross margin and look at performance from a unit economic value. And I think this will be a transition that we will get you ready for on Investor Day that our operating margin becomes a better measure for competitiveness of products than gross margin. Just keep in mind that the biggest move in gross margin this -- the last quarter was free traffic moving to pay traffic and cost moving into cost of revenue.

While that is decreasing the gross margin, it's literally a wash from an overall P&L perspective. And I think you'll see more movements like that. It will be up to us to give you the right insight. But what is clear to us across all products with the opportunity in front of us that the unique economic margin and value is going to increase over time. we -- with the guidance that is in place already today, we are getting north of 46% from a rule of 40 perspective, -- and we think we have visibility to reaching north of 50% next year. So that all shows you how much potential is, and we just need to give better insight in how the various parts are coming together and provide force in that direction.

Yes. The 1 thing I would add to what -- sorry, I was just going to say I want to add to what Tom said. -- is just to put a finer point on what I think is the most important there, which is for -- since our founding, Cloudflare always had a free version of our service. And that provided a lot of different benefits. One of the largest ones that gave us data to provide security models and do everything we did. And we haven't actually worked not hard to convert free customers into paying customers.

What -- and so the traffic that was associated with this frees customers went into -- as a marketing cost. What's been fascinating though is that, that giant pool of free customers turned out, a lot of them are developers. And so as we've now built incredibly compelling developer tools, most of our developer tools aren't free. They require at least some payment. And so you're seeing a lot of that free traffic turning into paid traffic.

And so it actually might be that the cost of customer acquisition that we have for those those really high-growth products like our developer platform are really fueled by what we've built over the end of the beginning of Cloudflare and some of those Act 1 products. And so well, that shows up somewhat finally in gross margin, -- it actually is a sign that more and more people are adopting the paid products, including the developer platform as, which is the part of our business, which I think I am certainly among the most excited about.

It's a great reminder and certainly a geography shift to your point, Matthew, maybe for you for my follow-up. I'd love to hit on Act 4 a little bit and Cloudflare's ability to sort of manage that relationship between AI tools, scraping type of tools and content owners. Maybe the question is, what are some of the milestones that you need to see in order to see that business inflect. Is it big lighthouse accounts? Do you need industry consortiums? It's just an uncharted territory. I'm curious how you think about it.

Yes. I think the way that we think about Act 4 is that the business model of the Internet, which has historically been advertising and subscriptions is about to change dramatically over the next 5 years. And exactly what it changes to I think it's still an open question. And I think it might not be 1 thing. I think it might be several things. Because of how much of the Internet sits behind Cloudflare we have a seat at the table of defining that. And so I think there are a number of different things that I'm watching. So 1 of them is -- I think that some part of this is going to be some kind of micro transactions for any request that agents are making.

It might be fractions of fractions of the pennies. But if you think about the -- I don't know, that $500 billion robust pass-through Cloudflare in any given second that -- some percentage of those -- we think that there's going to be some ability to have some micro payment that is made for that because something has to the infrastructure. And if you look at the growth in agenetic traffic. If you look at the growth in sort of nonhuman traffic on the Internet, somewhere in 2027, we think it's going to surpass in traffic, and it's not going to slow down.

And so we've got to figure out something else to build it. So I think 1 of the milestones we'll look for is how do we figure out what that sort of micro payment infrastructure looks like. And the challenge is like nobody can handle the volumes right now. And so we're looking around to partner with people, we're looking around for everything. But right now, the sort of transaction volumes that are -- that people are excited about, like 1 million transactions per second.

We need something that's significantly larger than that. So that's -- I think that's 1 thing. I think on the other side, first of all, it's really important to be clear that the answer isn't that everyone wants to be blocked from AI or everyone wants them to pay. For example, Cloudflare has a whole bunch of developer documents. We want those developer documents to be in every single LLM that's out there. So we make it as easy as possible for LLMs to crawl our developer documents. On the other hand, if you're an ad-supported business, then your being crawled is actually a threat. So I think we're trying to provide tools on both sides of that.

The side that you focus on the folks that want to block it, the ad-supported folks that are out there. And I would say that the first milestone we've seen is that we went from being relatively low in terms of our penetration in the media space. to today dominating that space. And so I think that's the first sign. And what I hear from media company execs is they are signing better deals with AI companies because we've given them the tools be able to control who has their content.

That's the early Lighthouse signal. I think the question is, how can we then take that down to the long tail. So sure, the [indiscernible] of the world can negotiate their own deals. How do we make that available for everyone that's on the Internet. And I think that's going to be really some lighthouse sort of deals with some of the foundational model companies. And I don't know exactly when that will come. But I do -- I will say that when we listed what our top 6 priorities, were for 2026. One of the 6 was making sure that we can make real progress and see the first revenue that we can then pass back to that long tail of the Internet in order to help make sure that we continue to create a healthy ecosystem for content creators. And I'm pretty confident we'll make that goal.

Your next question comes from the line of James Fish of Piper Sandler.

Thomas, you just mentioned Rule of 50 there potentially -- but given what you're seeing with the demand behind inferencing, I guess what's the team's willingness to go after more of this opportunity and really drive more megawatts behind the network to host more of the inferencing use case like what you're seeing out of some of your edge peers.

I think it started, but Matthew for sure, has an opinion on this. I think you see us leading in this opportunity already with all the fourth that we have. This is the reason why you saw developer town go up by EUR 1 million in the first quarter alone. We are continuing to up nice margin for these products but we are not at all restricting growth, just the opposite, I would say. So this is not a -- there's no restriction on leaning into this opportunity.

I think that, Jim, at risk of kind of saying as being critical. I think that people don't understand the difference in our business model versus the hyperscalers. The hyperscalers business is to buy a server -- and then to lease that server back, ideally for 5x or more of what they paid for it. And so if they don't have servers to lease then they can't grow their revenue.

And so their CapEx has to invest ahead of whatever that demand is that's out there. We focus on very different things. So the thing to watch for us is when you see us publish a blog post, how we figured out how to get more utilization across our fleet of GPUs or how to get the more models loaded quickly across GPUs. That's real IP that we're inventing internally -- and the metaphor to think about is let's put a time when I was in college, I remember, is a new thing called the web was starting and so we needed to have a web server.

And so we literally from gateway, I remember ordering a box that came with like calprints on the outside of it. We bought a gateway server and we plugged it in because there was no idea of virtualization. And then VMware came along. And after that, you had doctor and containers, and that was sort of the journey that everyone went on. We're still at the stage with GPUs of buying the physical server and needing to use that for most of the industry. And so across most hyperscalers, you're seeing utilization rates of their GPUs that are in the single digits, whereas we're slowly getting our GPU utilization to approach what our CPU utilization is, which is up in the 70% to 80% range.

And so as we do that, we can actually continue to -- with the fleet we have get service the requests that come in and invest behind demand as opposed to investing ahead of demand. And so because our business model is different, that's allowing us to continue to keep up with the inference demand and also do a lot of experiments and trials and things to capture developer mind share that is very different than, again, what your business is, is essentially leasing a physical server that you bought and trying to get at least 5 turns out of that lease.

Yes. Makes sense, Matthew and Thomas. Maybe just following up on the security side. Look, you guys have been aggressive about displacing the legacy hardware guys across security, firewalls, VPN and so forth. Are you seeing any compression in enterprise sales cycles for large-scale Zero Trust deployments or macro or macro approval is still kind of elongated. And really, the crux of it is the supply chain issues, the component issues causing more customers on the enterprise side to start to evaluate more of the cloud opportunities out there for protecting your environments?

I mean the hardware companies seem to have 9 lives, and I don't know how many more they have to use. But it's -- I think that as you see that there are vulnerabilities in hardware, which some of the Palo Alto had issues with, as you see, again, supply chain shortages, especially around memory right now, I do think all of those things are pushing more and more people to evaluate that they need at least some part of the cloud as part of their infrastructure.

But again, I have been impressed by how long the hardware players have continued to operate and hold out. So I'm not willing to call that this is the time that it's going to be a complete change. But I do think that there's -- I do think that the cloud continues to show its resilience and that all of these things are tailwinds behind our business and other cloud-native businesses that are out there.

Your next question comes from the line of Gabriela Borges of Goldman Sachs.

Matthew and Thomas, I wanted to get your observation on how this fleet mix may be changing between GPUs and CPUs. And spatially Matthew, you're just talking about how GPU utilization is approaching CPU utilization. Are you also finding that there are some AI inference that you can now route to CP? And Thomas, I imagine that has an implication on the unit economics of how you're serving up the AI sense market.

Yes. I mean so we want to make sure that for customers we're just abstracting what the most optimum silicon behind the scenes. And so for some models, has worked right. For others, we want to make sure that GPUs are available. As we deploy server, those services today have -- they sort of come with all of the various parts, and we -- and so it's not -- you deploy a GPU server, you play a CPU server, -- when we deploy a server now, it's got a CPU. It's got a GPU. It's got a certain amount of memory.

It's got a certain amount of storage, it's got a certain amount of capacity in the network. And those are all resources that we're constantly trying to both balance and then create opportunities around. And so I think what we're -- what's different about us is we're not renting each 100, but we'll have each 100s across our network. And then we're trying to match the workload that makes the most sense with the silicon that's behind that.

And with what a customer is paying for. So if you're paying us more, then we're going to give you a faster and better experience than if you're paying us less. And so I think that, that has always -- like some of as always been just a giant scheduler -- and so what we're effectively doing is dispatching those jobs across it and then being able to drive your question earlier about, is our cost going up because our developers are using more AI tools.

The answer is, yes, at the margins, but much less than we see from our peers, because we can use that giant schedule or to essentially run those tasks to anywhere we have excess capacity across our network and we can prioritize those based on what is more important or less important the data point intra quota on Entropic announcing managed agents.

I just love to pick your brain on where you think that type of infrastructure intersecting with the LLM creates opportunity and/or risk for their Cloudflare business model? .

I mean, like I think that we're -- so without talking about specific customers, I would say that what we're seeing from the major AI labs is, first of all, that we're partnered with all of them and get access to all of their lives. And that they see our infrastructure is critical to being able to deliver this. And so they are actually looking for partners to be able to have that infrastructure run on.

And so like there are the examples of in the case of -- we launch something called Dynamic Workers, which allows you to very quickly stand up something, which is significantly more efficient than a container. Containers are too slow and too heavy that to actually be able to respond to these incredibly fast agentic workloads. And so what AICs are doing is they're looking at this and they're seeing the opportunity. And so to give you a sense of -- I'm naming them, one of the large AI studios in just the last 15 days went from essentially 0 dynamic workers to over 1 million dynamic orders running across the platform.

And we're seeing almost everyone excited about the the underlying tools and technologies that we build. So I see what anthropic is doing as being very positive to the infrastructure that we see, we see that as an opportunity to deliver incredible value across that. And we see ourselves as very differentiated in the space and able to provide significantly better performance and at a significantly lower cost than anyone else who's playing in the space today.

Your next question comes from the line of Fatima Boolani of Citi.

Note along those lines with respect to the agentic edge and agenetic AI systems moving into production. I want to harken back to you on the announcements that you made at Egypt leak. It's very clear that there is a feel rewiring that you're working on to natively power agents to basically do their best work to put it simplistically, -- and a lot of the stuff seems like it is a frontier monetization opportunity for you.

But I'm wondering if you can speak to what sort of halo the rise of production agentic systems could have on the more flagship established parts of your business, i.e., the revenue and monetization halo to Act 1, Act 2 and certainly Act 3 products? And then I have a follow-up for Thomas, if I may.

Yes. I think for the Act 3 products, the developer platform products, I think it's the most obvious, which is that we're just seeing more code being created. It needs an efficient place to run it needs to have a set of primitives that can act as fast and as a femoral as agents need to act. So if you imagine that you're a company that's building something that's going to plan a vacation for someone.

What you really need is the ability to essentially spin up an extremely lightweight sandbox that writes code that assembles kind of all of the different parts of what that vacation may entail, puts it together, pull it back and then below the sandbox up very, very quickly. To do that with any other platform is extremely expensive and then slow.

And we've made it with some of the things that we've announced during Agents week and otherwise, we've made that just drop dead simple. And so that's driving a ton of use. And again, I think that the way to see that is the fact that we added 1 million developers to our platform last -- just the last quarter, almost a May as we did in all of go last year is extraordinary. And if you look at growth across the Workers platform, more than 3/4 of that is from new customers and the growth rates are pretty extraordinary across that.

That's also, again, to some extent, putting pressure on gross margin because those are less optimized for gross margin products, but we are confident that we can continue to drive GP utilization and all those things that we'll be able to get more out of that.

I think to the other part of your question, just as more -- every time an agent does something, if you think about it, you just -- if you type something into that GPT or any other things, like to search the number of sites that gets searched the amount of traffic that gets generated, if I'm looking for a digital camera as a human, I might visit 5 websites, if I really care about it. My agent is going to visit 5,000.

And so that's going to just drive significantly more usage which is the biggest driver of kind of our Act 1 revenue. And again, I think unlike some of the pure-play CDNs that are out there, that the agents aren't going to go watch real runs at the Super Bowl. They're going to drive things that actually drive real traffic to real e-commerce sites. And that's where Cloudflare and the huge pop the Internet that sits in behind us, is really valuable.

And then for Act 2, again, as we talked about already, I think being able to very narrowly define what data an agent has access to and what they don't. We're just seeing more and more of that usage, especially in the self-service category, which there really isn't another sort of SASE Zero Trust self-serve competitor out there with any sort of scale. .

And so that's with things like Open Claw driving a lot of usage there. And what we found time and time again is, as hobbyists or individuals adopt technology, they inevitably start to bring that technology more and more to work. And that's what we're seeing as we win more of the enterprise accounts across Act 2.

Thomas, just on the pool of fund. You're sort of in year 3 in earnest of having this motion with your customers nature. I was wondering if you have any comments or observations on pending renewal cycle from maybe your earliest vintage of pool of fund adopters? And maybe what trends you are seeing from a renewal and expansion and expansion of usage vectors as you reengage with some of these customers that are coming back to the well, so to speak.

Well, as we are now in our I would say, 6 quarter of pool of funds, it becomes a much more standard tool in our go-to-market motion. Folks are familiar in how to deploy the tool when to deploy it when it makes sense. So I think we get efficiency in the process. From a renewal perspective, you heard this in Matthew's prepared remarks, we had our highest ever renewal rate in the last quarter, and that includes all the pool of funds deals that were up for renewal.

So I would say, the hypothesis that this would be a tool that not only allows us to work expansion really well, but also becomes a very sticky tool from a customer engagement perspective has turned out to be true.

And your last question comes from the line of Shaul Eyal of TD Cowen.

Thomas, you mentioned quota-carrying sales capacity continues to accelerate. Could you provide some more color on your expectations to continue to grow capacity relative to productivity? And I have another follow-up for Matthew.

Yes. So when I said we are not touching carrying a sales capacity what goes along with that is where we see significant productivity in the support ratios for these AEs. So the ratio is going to change significantly. -- which means we are freeing up dollars and within the same spend envelope now of dollars, you can deploy more quota carrying capacity towards our market opportunity. And this, of course, then will allow us to continue to drive productivity from a go-to-market perspective.

Got it. Matthew or Thomas, partners increased to 30% of revenue this quarter. What's driving this continue to increase? And how much more channel mix would you expect going forward?

Yes. I'll start and Thomas might have more to add. I think that -- this really started with Mark Anderson laying out certain 2 years ago that we were going to have a motion that really included partners and made sure that they were able to deliver on that. I think that it has been an incredibly successful way for us to sell, especially our Act 2 products, which require a lot more of a consultative sale and a lot more of work making sure that the integration is done extremely, extremely well.

That, I think, will continue going forward. I think the big question is going to be what partners are really able to leverage this new world of Agentic AI in order to just get additional value and scale and velocity. And I think that's what we're evaluating across the partnership world. I think there's going to be a lot of change in that space. But I think that partners will continue to be an extremely important part of our strategy.

And just like, again, I think a lot of our businesses is changing. A lot of our partners' businesses are changing, and we're seeing that the ones that are delivering the most value to customers and are the best at getting success at selling our tools are the ones that are embracing new ways of selling and servicing and making sure that the customers are successful with our tools.

That concludes your Q&A session. I will now turn the conference back over to Matthew Prince for closing remarks.

So again, I just wanted to emphasize, this has been a hard day. We've never done something like this in Cloudflare's history, and we take it extremely seriously, and we know how much it has affected people who have been friends and colleagues. I am confident those people who are leaving us today are going to go on to take what they learn at Cloudflare and help build many more great companies.

In fact, it's amazing to see how many people are already writing in saying anyone who got traded at Cloudflare, we would be happy to interview. We're going to make sure that we take care of those people, but we also want to make sure that we are hiring for the right role. This isn't about us downsizing. This isn't about us saving costs. This is not making sure we have the right people in the right roles to build the future.

Our mission is to help build a better Internet. That's an important mission. It's never been more important as the Internet goes through all of the transitions with AI and agents and Cloudflare is going to lead the way. I'm proud of everything that we're doing. I'm sorry that we had the -- to take the action that we did today, but I believe it's going to make Cloudflare better for the future. Thank you. We'll see you back here next quarter.

This concludes today's conference call. You may now disconnect.

Excellent. So thank you, everyone, for joining us this afternoon. My name is Keith Weiss. I run the U.S. software equity research franchise here at Morgan Stanley. I'm really pleased to have with us from Cloudflare, CFO, Thomas Seifert. Thomas, thank you for joining us again.

Always a pleasure.

So we were on the same stage a year ago. And we were talking about how 2025 was going to be a year of inflection for Cloudflare. And it was a pretty spectacular year, right? There was activities that took place that we weren't really on the bingo card, if you will, of investors. You saw on your largest ever contract, $130 million contract, 7 years in duration. The highest ACV deal you've ever signed at $42.5 million. RPO growing 40% during the year, 30% revenue growth accelerated in each of the last 3 quarters. The net revenue retention rate improved from 111% in Q1 to 120% in Q4. A number of paying customers, up 40% year-on-year, and the list goes on, right? That's, I mean, by every measure, a spectacular year.

So maybe just to start out the conversation what clicked into place? I mean, it's not like you weren't performing well before, but it definitely feels like you're correct, right? There was an inflection that took place. What were the elements that came like really to fruition that enables those types of results?

I would say last year was the result of mainly two things. The first one was the compounding effects of a sales transformation that has been going on for more than 2 years now and all the things falling in place, moving up the enterprise stack, making good progress with large customers, the last -- and the larger the customer cohort for us, that means above $1 million or $5 million of ACV, the more acceleration we saw productivity on the sales force, pool of funds. So the sales transformation really compounding on all the vectors that have been touched. I think that there was one of the big contributors.

And I think the second part is that it becomes clear and clear that we are not selling just a product, but we put a platform in place that helps our customers solve significant and challenging problems. And the innovation around that offering, that platform is really what differentiates us from many in the market. And I think this could not have been clearer last year, but all the events around AI and the products and tools we are putting in place in order to support that transformation. That came together nicely. So knock on wood. I think those are the main two reasons why last year was such a special year.

Got it. So as we head into 2026, are you going to give us another prognostication of an inflection point to 2026? Or maybe a fair question for you. sales transformation and really stretching the benefits of the platform in 2025, any additional elements that we should be looking on and turning on in the story in 2026?

I think we'll make -- we'll continue to make progress around those fronts. And then on top of that, I think we all currently see that AI is re-platforming the Internet in a way how code is written, how people interact. And I think this will be one of their significant tailwinds for Cloudflare in this year and moving forward, we being honest again.

Yes. I want to start and dig in on sort of the product side of the equation, which is always a little bit dicey with the CFO, but we'll give it a shot. I'll be pretty good with the front. So we came out with a framework on Teams software at Morgan Stanley. We called it our best athletes framework right? And the elements of that best athletes framework was in a time of rapid change within the overall environment of rapid innovation. We're looking to companies that are -- have speed, meaning fast innovation. They have strength, meaning good kind of market positioning. They have good execution capabilities. That's the skill and then flexibility, right? The ability to change the business models to match that scenario.

And I think one of the most impressive things that we've seen over the past year is that product innovation really meeting the period, right? And you can see that across multiple parts of the platform. You guys often talk to like Act 1, 2, 3 and 4. What's so interesting though is that it's -- it's not like Act 1 and 2 goes away when at 3 and 4 comes online, so you build upon it. So -- what is it like that enables you guys to be that nimble and to be that sort of fast ahead of the trend? So if we take Act 3, the workers and the developer platform, Quickly, workers came the developer platform of choice underneath a lot of these cogeneration tools under leaving a lot of those platforms. What enables you guys to get into that positioning to be in the right place at the right time when these innovations are taking place.

I always go back and say when investors or -- start to work on us and if you really want to understand what we are and what we do, you have to go back and understand the architecture of the network. This is really the highest competitive amount we have. The network today after 15 years is now in 330 cities, in many cities like San Francisco in multiple locations in 25 countries and interconnected today with about north of 13,000 networks. So that is the infrastructure today of the shelf hardware, every product and every service we run, whether it's application services in Act 1, SASE and Zero Trust in Act 2, whether it is works now in Act 3, and whether it will be what we say pay to scroll products or whatever how we monetize content in Act 4 is running on that one network. Every product and every service we offer on every server in every location, and this is a flywheel that is feeding itself.

So you talked about Workers. Workers was originally not a product that was supposed to be sold. It was a tool that helped us to innovate our products fast enough how we write code and how we can deploy code with the push of the button to region earth, all around the globe. And this flywheel, I think, is based on this infrastructure is allowing us to stay ahead of this innovation game. And now we're in a world of agentic AI flows, where looking back now, it seems the infrastructure of the network was built exactly for that use case. What do you need in order to build an agent? You need a CPU, you need a GPU need API. The CPU orchestrates. The GPU compute. The API fetches the data. And in cloud and this happens all in one place, on one server in a world where now between 20% to 30% of our websites and all traffic is moving through the network. So the insight you have into what market needs and where markets are going, is driving that flywheel to an extraordinary extent.

Got it. It has a lot of visibility by having some of that perspective. And a network that was built for flexibility, right, that one unified network that everything built up, and it's all software that gets built up. There's no bolt-ons as over the decades. But on the other side, you've been able to develop an extreme level of developer affinity. On the most recent conference call, you talked about over 4.5 million human developers. One of the things that really struck me, I was looking at a stack overflow survey and they ask developers like where do they want to develop their applications -- the first 3 on the list is exactly who think would it's Amazon, Microsoft and Google because they're huge and the behemoth. N4 is Cloudflare, right? It's become the place, not what your boss is telling you to develop on, but the place where developers want to go to develop. How do you -- like how do you gain that developer affinity? How do you make yourself such an attractive platform for those developers?

I think a couple of things come together. The tools you've built, the ease of use of the platform and the languages that you put into the answer of the developers to work with it. The cost of the product, the performance of the product. So they all have been coming together. The interesting part now is we talked about 4.5 million human developers on the platform. If you look at the agents on top of that, it's staggering. And if you look at downloads of work TV over the last 2 months, it's going almost vertical. A big part of that is literally agent AI agents downloading, programming themselves still leading when they are done. And -- so this comes together now in a really unique way.

Yes. And this is a really interesting point. And I was talking with the [ Sachit Singh ], who covers a lot of infrastructure names for us. And we're talking after the data bricks presentation. And one of the themes we're starting to hear more and more is the tooling doesn't necessarily have to be built out just for -- or can't just be built out for humans. Now you have to build a set of tooling a set of infrastructure that's very meetable to the agents. So does that change the way that you approach the Workers opportunity? Does that change sort of how -- sort of what you need to build out in there to make yourself be sort of platform of choice, not just for the human developers, but for the agents?

Very much. I mean we started really early, over a year ago or 1.5 years ago to put this boat in place. This is what we called it, around the developer activity. Their person in charge, Aly, actually the woman who developed Workers as a product in order to get our arms around what go-to-market do we need for enterprises, what features do we need? How do we get to the large deals? How do we keep the human developer count growing? And how do we have -- what do we have to do from an offering, from a feature and a performance perspective to address what you said? How do we make ourselves discoverable for agents that are doing development work writing code on our network.

Yes. Got it. And you guys recently acquired Replicate and ASTRO to strengthen that developer platform. Can you talk to us about what these acquisitions brought into the equation and how it fits into that broader vision of workers that we're talking about?

I mean how do we cater to the developer ecosystem in terms of the offerings we have, the products we have, how do we get more dynamics in our go-to-market place and how do we build this marketplace for developer tools and products that you're trying to work with us in our platform. I think both acquisitions squarely fit into that motion and you'll see us do more activity in this direction moving forward.

Got it. And can we just touch briefly on sort of the defensibility? There's AI labs that are focused a lot on the -- I mean, it starts with co-development, agentic co-development, but they're broadening their scope, right? Is there a dividing line that you feel comfortable there's a defensibility of sort of the Workers platform versus where this expanding scope of what we're seeing the AI labs and their focus on that whole development life cycle is going to hit?

I think it's a combination of many things. I think it starts with us being that control plane where agents have to move through whether it's a human interaction or an agent, interaction needs to be secured. It needs to be transported. And the move to our network today with 13,000 interconnects and 20% to 30% of the traffic behind us. So there's that. I think the architecture of the networks and how Workers architected on top of that in terms of no containers, but isolates. We talked about how agents workflows are architected. You need a CPU to orchestrate. You need a GPU to compute. You need an API to get data. In our world, you have to wait for the traffic for the API to deliver data or not. In our architecture of isolate, this wait period can be important, can be really expensive when you work on container platforms.

In our case, the spin up and more importantly, the spin downtime when you're waiting for the data to come back is really fast. It goes to zero. With that, the offering of how we are able to price, not for time, not for capacity, but for a task completed makes us really unique.

So I think the combination of the ecosystem, the tools, the architecture of the network and then the architecture of Workers itself running on that network is a pretty substantial competitive amount that is hard to overcome.

So it comes back to that strength that being in front of 20% of the Internet, but really the infrastructure of being 20% of the Internet with a secure platform should be a durable advantage. And that kind of sort of segues into Act 4 of what you guys have been talking about in terms of the agentic web. Can you just, maybe as an overview, describe to us what's the opportunity for Cloudflare in there? We all understand a lot of agents are being built on workers that's part of a there more and more attentive activity taking place. But what is the -- like the new monetizable opportunity for Cloudflare in this Act 4?

In the near term, it's actually that opportunity and getting ready to monetize a world of agentic AI workflows and content is actually driving business into the older Act first. So today, about I would say, 80% of the 50 top AI native companies is behind our network and that they need to secure. They need to mitigate traffic and request. So that's driving business. We see tailwinds from verticals that we weren't traditionally strong. And especially media content creators, publishers that now want their content to behind a network where they can control how it's monetized. So in the first time, the opportunity of Act 4 is literally driving business in Act 1 and 2 in the first place. And we've seen this as part of the acceleration in the last year. And we'll see more in -- during '26.

So it then moves us into the world, how is content monetized, how do we sit between frontier models and content and give both sides the opportunity to monetize. There will be agentic AI workflows that do business on each other's behalf or on new behalf with micro transactions. So how do we build rails in order to facilitate that, that's why you hear us talking and partnering with pretty much every large payment provider, whether it's the credit card companies, whether it's PayPal or invoice. We talked about putting our own stable coin in place in order to facilitate that. How that is going to play out in detail? I think there are a lot of questions still that need to be answered. But we are making really good progress building an ecosystem and building the rails that need to enable that.

Got it. Excellent. You guys talked about AI traffic already ramping up on the platform. I think you made a comment about in January of 2026, weekly AI agent requests more than doubled on your network. And a lot of investors was a good case in point, had made the connection of like this is a lot of API calls, right, which is kind of Act 1 in application services. There's a lot of network traffic to be secured. Where are we in terms of sort of yes, there's a doubling, but from a very small base -- where are we on that evolution of when this agentic traffic gets to a volume that it's material for your business that it becomes a more material driver of Act 1 and Act 2?

So I think we're in the early innings. Today, I would say, slightly north of 50% of the traffic in our in our network is API-driven, but most of it is human initiated. As you said, in the first 6 weeks of this year, we saw agent-driven traffic really go through the roof double literally over a very short period of time. From how we handle a request, it really doesn't matter whether it's a human request or machine-driven request. And so how that will -- and most of our requests are priced in terms of [indiscernible] always makes a good comparison in T-shirt sizes. So you're the customer you buy an M T-shirt that gives you [ 40 million ] you request. And as you approach the cap, you buy the next size defer. So there is a lag and a step-up before you see revenue materialize.

So it's a lagging, I think, indicator. We are coming from a smaller base, but the slope of the increase is staggering. I think if you ask me 6 months ago, if that is even close to possible, I would have said no, but staggering.

Got it. And the -- if we think about the unit economics, if you will, of the AI and autonomous agent traffic, it sounds like you guys are neutral, right? It doesn't matter to you, whether it's a human initiated or an agent initiated, you still feel comfortable with the economics of that core application services business from that rising traffic spike?

Yes, we are. I mean this is the beauty of -- and the efficiency and the flexibility of that network that it's able to digest incredible amounts of fluctuations and upsides. And I mean, go back to COVID where traffic literally increased 60% quarter-over-quarter and we didn't even see a flinch in our network, neither from a performance nor from a capacity, even one part not from a gross margin perspective. So it's able to digest incredible amounts of swings of traffic. We are highly comfortable that, that is what the network is built for and that we can digest that.

We have good visibility on what the trends are. We guided for this year in the widest range we ever guided at 12% to 15% and the highest on the upside, 15%. Not so much because we need more capacity. It's much more a reflection of some of the supply chain pressures we see from a price perspective, especially on the memory side. But again, it gives you an idea how flexible the network is. One of the things that always has driven this efficiency in the business model based on the architecture is that we are in this unique position to invest behind demand. So you're never in a position we say, I have to invest $100 million in GPU capacity hoping that traffic will come.

If for whatever reason, traffic would explode in across the locations in San Francisco, we'd probably be able to double the capacity across all San Francisco locations, generate revenue in these sites before we even pay for the hardware that we put in place. And that means that you're always in this comfortable position that you can follow demand and you're never really investing ahead of it. That is also why over the last 1.5 years, we spent so much time on really understanding what inference virtuals are, how they vary. Inference task is not like inference tasks and how we optimize our hardware stack for that. So we can abstract the need for the customer to decide I need this capacity you need this GPU, but I think we'll compute it for you find the cheapest, the most performant way to approach this.

And again, it comes back to the architecture of the network. This is why we are so CapEx efficient, and that's why the absorbability of the network to deal with gigantic fluctuations and traffic is so significant.

Got it. I want to switch gears a little bit and talk about network services and in particular, Zero Trust and Cloudflare One. This has been, I think, a building momentum within the Cloudflare story. I think the part that most aligns with sort of the improving sales productivity. But just also one of the more competitive environment. So there's a lot of areas that Cafe is almost sort of a class of one, right, when we're talking about these application services or Workers. But we've talked to several SASE vendors, right, at this conference. How does Cloudflare look to differentiate in that environment where there's like more of a competitive environment? There's more of a sort of direct vision from the end user of the CISO of what they're looking for in these solutions.

Yes. I mean we are really happy with our asset development for a couple of reasons. First of all, from a feature performance perspective, we launched products early and then the feature out over time, and we call this moving up into the right and whatever Gartner-Forrester quadrant, and we are right there now. what we needed most now in over the last year in terms of continuing momentum is building out our go-to-market partner and channel. And that was needed because from a go-to-market perspective, there was not really a lot of value to be delivered for the partners on application services. The products are really easy to install. It's really even for Morgan Stanley would be behind our network within hours, not within days.

So -- but for Zero Trust, we needed to build out a partner channel strategy, which Mark Anderson successfully did. Our partner-driven revenue share in the fourth quarter was the highest ever. It was 29%, which is really a good proxy for our progress on Zero Trust. This is what we mainly use the Channel 4. If you look at our competitors, their general shares in terms of revenue are significantly higher. So we have lots of opportunity to continue to grow. And our win rates are really, really strong. So when we are in consideration the likelihood that we win is really high. So we are happy with the momentum.

And of course, SASE for us is not just a product or a suite of products. It's part of a platform. And as part of this platform, it is a significant value distribution. We have not been tired to read down the wins we had on even very large customers in traditional verticals even over the last earnings calls. So good progress.

Last point. It's probably our strongest margin product because it takes advantage of the infrastructure that has been built with what you would call Act 1, the size of the pipes that transport the data is sized on Act 1 for data moving out to the eyeballs. Now you fill it with traffic coming back with SASE product. So it's a high-margin product. So we have ample of room to compete.

Excellent. So it seems to me that the -- having Mark Anderson engage a partner channel, as well as the evolution of the products, just gives you more credibility as a security vendor, right? It's fundamentally a little bit of a different buyer from Act 1 and step from Act 3. What are the customer conversations today like? Are you still convincing them that Cloudflare is a real security vendor? Or are we past that point and it's really about how broad can we go with the Cloudflare One platform?

I think we have moved far beyond that already. I mean -- and you see that the largest customer cohorts of the fastest customer -- fastest-growing customer cohorts. If you just go back to the earnings calls, most of the customers that we read down there are coming from traditional verticals, critical infrastructure verticals, the progress we are making on with federal customers, not only in the U.S. but also outside with friendly. So I think we are beyond that.

I think the biggest change in the go-to-market transformation, where Mark Anderson really made the difference is much more in terms of ingraining this enterprise DNA into the sales force and into the cloud fare culture. We've been quite open, transparent that we were always a product and innovation driven go-to-market engine, and it needed the DNA change.

And then last but not least, what I said before, combining that with a highly effective channel strategy, those 2 parts we're missing. I think he did an incredible job instilling this into the company. And I think we are in the customer conversations we are beyond what you said. We don't have to convince people anymore that you are a serious player.

Right. So Mark's been on board for about 2 years, engaged sort of a channel strategy. But to your point, created much more of an enterprise sales motion. You put that together with the breadth of the solution portfolio. And I'm assuming it's those two things coming together that enables something like a $42.5 million ACV deal. So maybe we could just touch on what does that look like? Like what is the customer acquiring from Cloudflare when they're putting that much budget annually behind your company, behind your platform?

Well, first of all, I interpret that it's a vote of confidence, right? You commit a lot of dollars to a supplier across a very broad range of products. A pool of fund steel in this case really means that you can buy every product that is on the offering list. You negotiate a rate card for every product that we have. And then you decide how you use -- we talk, of course, to each other, but how you consume. What it does for us is it makes this product adoption expansion rather frictionless, right? Because literally, once you are in our network with the product, every other product and services, is very much a mouse click away. This is all in me.

So having this rate card already negotiated the commitment in place, it makes a rather frictionless process in terms of expanding and expanding across our product portfolio.

Got it. So pool of funds contracts have pretty quickly ramped up. I think you talked about it is about 20% of your ACV. From an investor standpoint, how should we think about the impacts of this new contracting methodology in terms of variability in the business model and potential more volatility in kind of the revenue run rate?

Well, first of all, we should be all excited because it's large deals and committed revenue over a period of time. And it generated noise, especially in the beginning, especially when an existing customer migrated from a traditional contract into a pool of funds contract to make this more visible. Say you're a customer, you're committed to $10 million a year. Now you signed a pool of funds contract for $40 million over -- or $45 million over 2 years. So of course, you will get better unit pricing because you have a significantly larger commitment. So in the first 2 quarters, while you consume as much as you did before, maybe even slightly more because the unit prices are lower, you have to step. So you have a headwind to DNR for a quarter or 2 makes forecasting a little bit more tricky, but then you catch up and you move out of this, you have this Nike swoosh from a revenue perspective. And this is what we have seen last year.

Of course, this is only a headwind that you see when you migrate an existing customer into a pool of funds contract for a new customer that is just DNR expansion moving forward. So it makes my life a little bit at my teams like a little bit more difficult from a accounting building, especially forecasting perspective. But since it's committed dollars over the length of the contract, this is all good news. And of course, both sides have to get used to how you handle that. And I think we are through this now the forecasting has become much more stable. If you look at across all the pool of funds deals that are in play, we are slightly ahead of linearity. So there's good management also from a customer consumption perspective. So -- that's become a rather beneficial instrument to especially grow our large customer cohorts even faster. And one of the reasons why DNR improved so significantly over last year.

Got it. Got it. I want to turn to gross margins. As the CFO, I have to ask you some margin is as well. So you guys are still within that 75% to 77% long-term target model. But we have seen some impacts on the gross margins. And it seems like a lot of this is you shift from free to paid customers. you dig into why that an impact and sort of how -- like what's the duration of that impact? And is there anything else that we should be aware of from the gross margin perspective within that time frame?

Yes. Maybe just for background. So we have hundreds of thousands of paying customers. But we also have millions of free customers. And that is an incredible value to the business model because if you even go back to [indiscernible] the reasons why we have free customers gives us threat intelligence diversity. It allows us to consolidate huge amounts of data that makes us an attractive partner if we go to an ice say, co-locate us and hear the benefits. So we have many free customers.

The cost to serve these customers, we put into sales and marketing there. We always have. That is how we handle it. Now if such a free customer now decides he or she wants to become a paying customer, the cost to serve the customer move from sales and marketing into cost of revenue. so have an impact on margin for the overall P&L. It doesn't really matter, but you have a shift in margin. What we've seen over the last 2 quarters is a significant shift of free customers wanting to give us money. So it's actually really, really positive. And we -- if you look at our paying customer count, it has exploded. Some of them are really micro customers. They might be just $1 above the threshold, but it means that all the costs move from sales to marketing into gross margins on that is accounting-wise it pressures gross margin overall, the unit economics they hold.

In addition to that, so I don't want to be flimsy about this, our developer business, as we talked about it before, is really taking off. The growth rates are significant. Developer product is [indiscernible] is still below corporate average. We've made significant improvements over the last year in terms of utilization, but it's under the average gross margin. So there's slight pressure. However, the overall unit economics for this product are really, really good because significantly cheaper to sell that product. So if you look at across all 4 Acts, they have very different gross margin profiles, but very consistent unit economic margins.

So we -- one of the lessons we learned now. And as we go into our Investor Day later this year, we'll talk more about both sides of that equation. Because if we just signal and continue to signal gross margin performance that would lead to a misallocation of capital. We don't want that. So we have to adjust.

I'm going to try to squeeze in one question from the audience.

On the AI front -- on the AI front, it sounds like everyone's focused on what the AI is going to [indiscernible].

A couple of ways. And I think if you go back to our plus from -- last week, we wrote a big piece of software in a week with one developer whatever, less than $1,000 of tokens. I think that gives you an idea.

I think it monetizes in a couple of ways. First of all, it just increases traffic through the network in a very gigantic way the traffic continues to be secured, needs to be moved, needs to be controlled. So we become this control plane for that traffic that is going to explode. These agents will do business with each other. We will find ways as part of 4 to monetize that. So clearly -- and then the biggest advantage, I think, of the developer growth that I've seen so far is it gives us insight into how inference tasks are going to develop moving forward. So the variety of things that are happening is so impressive, but it gives us this unique opportunity for get dollars for a moment to optimize our hardware stack, better than anybody else in that direction of how we run workloads on our network. That's the reason why our CapEx is so low.

So I think there are there are more than one benefit to what you described. I think the flywheel around this is going to be a significant tailwind to our business. It's one of the reasons why we think agent AI is such a multiplier to our infrastructure.

Outstanding. It's a great note to end on. Thomas, thank you so much for joining here.

Thank you, Keith. Always. Thank you.

Hello, and welcome to the Cloudflare Fourth Quarter 2025 Earnings Call. [Operator Instructions] I would now like to turn the conference over to Phil Winslow. You may begin.

Thank you for joining us today to discuss Cloudflare's financial results for the fourth quarter of 2025. With me on the call, we have Matthew Prince, Co-Founder and CEO; Michelle Zatlyn, Co-Founder and President; and Thomas Seifert, CFO. By now, everyone should have access to our earnings announcement. This announcement as well as our supplemental financial information may be found on our Investor Relations website.

As a reminder, we will be making forward-looking statements during today's discussion, including, but not limited to, our customers, vendors and partners, operations and future financial performance, our anticipated product losses and the timing and market potential of those products, our anticipated future financial and operating performance and our expectations regarding future macroeconomic conditions. These statements and other comments are not guarantees of future performance. and are subject to risks and uncertainty, much of which is beyond our control. Our actual results may differ significantly from those projected or suggested in any of our forward-looking statements.

These forward-looking statements apply as of today, and you should not rely on them as representing our views in the future. We undertake no obligation to update these statements after this call. For a more complete discussion of the risks and uncertainties that could impact our future operating results and financial condition, please see our filings with the SEC as well as in today's earnings press release. Unless otherwise noted, all financial numbers we talk about today other than revenue will be on an adjusted non-GAAP basis.

You may find a reconciliation of GAAP to non-GAAP financial measures that are included in our earnings release on our Investor Relations website. For historical periods, a GAAP to non-GAAP reconciliation can be found in our supplemental financial information referenced a few moments ago. We would also like to inform you that we will be participating in [indiscernible] Silicon Slopes Conference on February 26 and Morgan Stanley's Technology, Media and Telecom Conference on March 3.

Now I'd like to turn the call over to Matthew.

Thank you, Phil. We had a terrific quarter. We achieved $614.5 million in revenue, up 34% year-over-year. We now have 4,298 customers paying us more than $100,000 per year, up 23% year-over-year. Revenue from these large customers grew 42% year-over-year, contributing 73% of our revenue in the quarter, up from 69% in the fourth quarter last year. Our dollar-based net retention was 120%, up 1% quarter-over-quarter and 9% year-over-year. Our gross margin was 74.9%, just below our long-term target range of 75% to 77%. We delivered operating profit of $89.6 million, representing an operating margin of 14.6% and and we generated strong free cash flow of $99.4 million during the quarter, again exceeding expectations. We blew away our previous record for new ACV in the quarter. with strong year-over-year and quarter-over-quarter acceleration.

In Q4, new ACV book grew nearly 50% year-over-year, making it not only a record quarter in absolute ACP dollars but also the fastest growth rate we've delivered since 2021. We've talked for the last 2 years about our journey from product-led growth to true enterprise sales. I'm proud to report our go-to-market engine is firing on all cylinders. Some highlights.

Global sales productivity increased year-over-year for the eighth consecutive quarter, surpassing our all-time high productivity set back in Q4 of 2021, and we believe there's still significant headroom Year-over-year growth in net sales capacity increased at the fastest pace since 2022. Our sales team achieved the highest quota attainment we've seen in the last 4 years. For the fifth consecutive quarter, we added a record number of customers that spent over $1 million per year. And after starting the year, signing our largest total contract value deal in Cloudflare's history of $130 million over 5 years, this quarter, we closed the largest annual contract value deal in our history, $42.5 million per year. This momentum today is the cumulative result of 2 years of effort to retool our go-to-market team to match our already world-class engineering and product teams. We've long been a place where the best engineers came to earn their strides.

The team, Mark Anderson has built deserves credit for making Cloudflare destination where the best salespeople now know they can come and win. Speaking of winning, let me highlight some of our customer wins in the quarter. a leading AI company expanded their relationship with Cloudflare signing a 2-year $85 million pool of funds contract for our full platform selecting [indiscernible] their single long-term infrastructure provider with 100% traffic allocation. Following a rigorous RFP, they selected Cloudlet over major hyperscalers not just for our unified stack and rapid innovation, but also for our strategic neutrality. This win underscores a growing trend the most sophisticated AI companies are choosing Cloudflare as the mission-critical independent platform to connect, protect and build the future of the AI-driven Internet.

Another leading AI company expanded their relationship with Cloudflare signing a 1-year $5.4 million contract for our Workers' developer platform and application services. what's most compelling about this win is that it was a classic build versus buy scenario against the hyperscalers in an industry where being first matters, our ready-to-deploy developer platform provided the agility and speed to market they couldn't find elsewhere.

With Cloudflare, this customer is now able to manage heavy global traffic with [indiscernible] availability. This deal is a testament to our shift from being just a vendor to instead being a strategic co-innovation partner for the world's most sophisticated AI companies. A Fortune 100 technology company expanded their relationship with Cloudflare signing a 3-year $5.8 million contract, representing a notable upsell from their initial engagement with us in mid-2025.

As a leader in AI, this customer operates under a strict mandate for global resiliency requiring a multi-vendor architecture to ensure 0 downtime for their application performance. We beat out the competition not on price but rather on performance and engineering innovation, a Fortune 500 technology company expanded our relationship with clouding 2-year $45 million pool of funds contract. While this customer leverages hyperscalers for their primary infrastructure they found those providers unable to manage the complexity of global third-party applications at scale.

[indiscernible] Cloudflare to displace a longtime incumbent, not only for the extensibility of our workers' developer platform and our superior application services but for our ability to fill critical gaps in the hyperscaler model, underscoring [indiscernible] role as a necessary fabric for secure performance multi-cloud operations. A major international consumer goods company expanded their relationship with Cloudflare, signing a 3.5-year $6.6 million contract for application services and our Workers' developer platform.

For years, this company relied on a fragmented architecture of multiple legacy vendors, which created significant operational complexity and gaps in their security. Despite an incumbent offering significant discounts to keep the business, this customer chose Cloudflare because of our next-generation architecture and workers' developer platform, which allowed them to customize their edge traffic. We didn't just replace 3 incumbents we replaced a legacy mindset proving once again that when you give developers the best tools, the entire enterprise follows.

A European Global 2000 technology company expanded their relationship with Cloudflare, signing a 3-year $5.8 million pool of funds contract to provide seamless access to our entire platform. We signed our first deal with this customer back in February after quickly realizing the power of Cloudflare's platform, they came back to us looking to move from a small variable commitment to a deep strategic partnership.

Unlike their legacy incumbents, our combination of best-of-breed security and our workers' developer platform enables sophisticated automation to manage their global infrastructure and greater flexibility to innovate at scale. It's early days with this customer, and we're already in discussions regarding AI call control. A large United States government entity expanded their relationship with Cloudflare, signing a 2.5-year $2.2 million contract for our Zero Trust portfolio, including Access, Gateway, DLP, [indiscernible] and e-mail security. We solve complex compliance requirements that the first-generation Zero Trust incumbent we displaced could only handle manually and demonstrated our ability to address their most sophisticated use cases that neither a hyperscaler nor the incumbent could solve.

After seeing Cloudflare action, their security team told us they were "shock" by the simplicity, automation capabilities and power of our platform. A U.S. media company signed a 3-year $3.1 million contract for AI crawl control, along with application services and workers. This customer was facing a massive increase in AI scraping, which was crushing their network and driving up infrastructure costs. They chose Cloudflare to gain visibility into which AI models are consuming their data, allowing them to protect and eventually monetize their unique content.

By leveraging Cloudflare Workers to replace years of complex technical debt from an incumbent, they were able to migrate massive Internet properties into production in just 2 weeks. This deal proves that as AI accelerates, Cloudflare is the partner of choice for companies looking to protect their IP while improving performance, reducing operational costs and enhancing their security postures. What's the trend across all these wins? I think first, vendors are once again being sorted between [indiscernible] haves and must-haves.

The quantitative metrics and qualitative wins this quarter again demonstrates that Cloudflare is squarely in the must-have bucket as a nonnegotiable linchpin of the modern technology stack. Second, we are seeing the shift to AI and agents drive more demand for cloud player services. What we're witnessing is a fundamental replatforming of the Internet. AI is driving a paradigm shift in how software is both created and consumed and that is turning out to be the biggest tailwind for Cloudflare's network and workers' developer platform.

If you look at the last 30-plus years of the Internet and software ecosystem, they were built for human consumption, people and seats and clicks now the genetic Internet is emerging, and we can already see its trends. If humans looked at 5 sites when they were making a decision, agents might look at 5,000. If humans had to fall back on generalized software and interfaces, agents allow for infinite custom ability of every software application for every need. If humans follow a common circadian rhythm to work, agents never need to sleep. Agents in other words, are the ultimate infrastructure multiplier.

In turn, they are reshaping the very economics of software. The industry is transitioning from a business model defined by seat licenses to one where the winners are those providing the compute, connectivity and rails and guardrails for these new digital workers at scale. Cloudflare was built for this moment. We are uniquely architected to capture value on both sides of the Agentic interactions. That means we win when AI applications are built on cloud are workers, but we also win just from the increased usage of all of our products and agenetic Internet drives.

Let me explain. When the cost of generating code drops to near 0, the volume of new applications explode. It's not a coincidence that most so-called Vibe coating platforms are either built on Cloudflare Workers or have us as their preferred deployment target. We exited 2025 with more than 4.5 million human developers active on our platform. It's a lot more if we count their agents. Developers are using workers to run autonomous logic across our global network containers for sandboxes and AI gateway to manage influence with cashing rate limiting and observability. AI usage is even driving adoption of our Zero Trust platform to ensure that data is compartmentalized and access granted in limited and controlled ways.

And it's not just established developers, I'm proud that we're ranked the #1 cloud platform in the latest stack overflow survey for developers learning to code. But the opportunity extends beyond just the developers building on Workers. We're seeing a genetic workloads generate an order of magnitude more outbound request to the web than traditional user-driven applications.

Over the month of January alone, the number of weekly requests generated by AI agents more than doubled across the Cloudflare network. This is driving increased demand for our whole platform. This is where Cloudflare scale becomes our moat with more than 20% of the web already sitting behind Cloudflare's network, we are effectively the global control plane for the genic Internet. That's creating a number of new growth opportunities, both with our traditional business as well as what we've begun calling AT4, helping invent the future business model of the Internet.

If AI agents are the new users of the Internet, Cloudflare is the platform they run on and the network they pass through. This creates a virtuous flywheel more agents drive more code execution on our workers' development platform, which in turn drives more demand for Cloudflare's performance, security and networking services at scale this compounds powerfully.

Given the opportunity we have ahead and how effectively and efficiently our team is executing on it, I wouldn't trade places with any CEO of any other company, public or private. But just to keep things grounded. Let me turn it over to Thomas to walk through our financials. Thomas, take it away.

Thank you, Matthew, and thank you to everyone for joining us. Entering the year, we outlined the specific factors that gave us confidence to invest in reaccelerating growth over the course of 2025. We are proud to report robust fourth quarter and full year 2025 results that validate the strategy and demonstrate our ability to execute our commitments, driving accelerating growth innovating rapidly and delivering financial and operational excellence.

We achieved total revenue for the fourth quarter of $64.5 million accelerating for the third consecutive quarter to 34% year-over-year, providing clear evidence of the continued momentum in our business. From a geographic perspective, the U.S. represented 49% of revenue and increased 31% year-over-year. EMEA represented 27% of revenue and increased 31% year-over-year. APAC represented 16% of revenue and increased 50% year-over-year.

Turning to our customer metrics. In the fourth quarter, we had approximately 332,000 paying customers, representing a record addition of nearly 37,000 paying customers sequentially and an increase of 40% year-over-year driven again by an uptick in customers, including those graduating from the free tier to small paid accounts, particularly for our developer platform products. We ended the quarter with roughly 4,300 large customers, representing an increase of 23% year-over-year.

Revenue contribution from large customers was 73% of revenue during the quarter up from 69% in the fourth quarter last year. For full year 2025, revenue from large customers represented 69% of total revenue compared with 67% in 2024 and 64% in 2023. We, again, saw particular strength in our largest customer cohort. We ended the year with 269 customers that spent over $1 million with us representing a 55% increase year-over-year and a record addition of $961 million plus customers in 2025.

For some perspective, we only had $961 million plus customers in total, just 2.5 years ago. Now that is what Cloudflare is adding in a single year and is still just getting started moving upmarket into the enterprise. This continued significant expansion with our largest customers drove an acceleration in our dollar-based net retention rate to 120% in Q4, up 1% sequentially and 9% year-over-year. Moving to gross margin. Fourth quarter gross margin was 74.9%, representing a decrease of 40 basis points sequentially and a decrease of 270 basis points year-over-year.

During the fourth quarter, paid versus free customer traffic again increased significantly both year-over-year and quarter-to-quarter, resulting in the highest allocation of network expenses to cost of goods sold versus sales and marketing ever. The underlying economics of our network driven by its inherent scalability and efficiency remained unchanged. Our Workers' developer platform also delivered another quarter of outsized growth. while the relative revenue contribution across our 4 acts can impact near-term gross margin, the unit economic margin of our business remains very consistent. Network CapEx represented 13% of revenue in the fourth quarter. We expect network CapEx to be 12% to 15% of revenue for full year 2026.

Turning to operating expenses. Fourth quarter operating expenses as a percentage of revenue decreased by 3% year-over-year to 60%. Our total number of employees increased 21% year-over-year, bringing our total headcount to approximately 5,200 at the end of the quarter. Sales and marketing expenses were $214.4 million for the quarter. Sales and marketing as a percentage of revenue decreased to 35% from 36% in the same quarter last year.

Research and development expenses were $94.9 million in the quarter. R&D as a percentage of revenue decreased to 15% from 16% in the same quarter last year. General and administrative expenses were $61.2 million for the quarter. G&A as a percentage of revenue remained consistent at 10% compared to the same quarter last year. Operating income was $89.6 million, an increase of 33% year-over-year compared to $67.2 million in the same period last year. Fourth quarter operating margin was 14.6%, consistent year-over-year. These results highlight our continued commitment to operational excellence.

Turning to net income and the balance sheet. Our net income in the quarter was $106.8 million or a diluted net income per share of $0.28. Free cash flow was $99.4 million in the quarter or 16% of revenue compared to $47.8 million or 10% of revenue in the same period last year. We are comfortable with consensus free cash flow estimates for 2026. I -- we ended the fourth quarter with $4.1 billion in cash, cash equivalents and available-for-sale securities. Remaining performance obligations, or RPO, came in at $2.49 billion. representing an increase of 16% sequentially and 48% year-over-year. Current RPO was 63% of total RPO and grew 34% year-over-year.

Moving to guidance for the first quarter and full year 2026. As a management team, we've always taken a disciplined data-driven approach to scaling Cloudflare and remain committed to creating significant long-term shareholder value with our focus on delivering durable revenue growth while maintaining the strong unit economics of our business. As we transition into 2026 with visibility into multiple independent growth vectors that give us confidence in our ability to continue to execute against this winning formula.

For the first quarter, we expect revenue in the range of $620 million to $621 million, representing an increase of 29% to 30% year-over-year. We expect operating income in the range of $70 million to $71 million, and we expect an effective tax rate of 20%. We expect diluted net income per share of $0.23, assuming approximately 377 million shares outstanding. For the full year 2026, we expect revenue in the range of $2.785 billion to $2.795 billion, representing an increase of 28% to 29% year-over-year. We anticipate revenue to be weighted at 46% in the first half of the fiscal year.

We expect operating income for the full year in the range of $378 million to $382 million. We expect an effective tax rate of 20%. We expect diluted net income per share over the period to be in the range of $1.11 to $1.12. We expect approximately 377 million shares outstanding based on the assumption of a cash settlement of our 2026 convertible notes.

In closing, the strength in our fourth quarter and full year results reinforce our conviction in our strategy and our ability to deliver exceptional long-term value for our shareholders and customers as we continue to pursue the massive opportunity ahead of us. And with that, operator, please poll for questions.

[Operator Instructions] Your first question comes from Matt Hedberg with RBC.

Congrats really on a stellar year. It's been quite a year of acceleration. I wanted to kind of wrap 2 questions in one, if I could. It's obviously been a volatile start to 2026 from a tech perspective. Obviously, the hyperscalers are spending more on CapEx and being punished for that. And at the same time, SaaS vendors are being punished for AI disruption. And I think part of that is we all hear the news around AI agents like open claw and things like Cloudflare. So I guess [indiscernible] question is, in the short term, how do these agents impact cloud floor? But then also, is there a longer-term perspective on that? And maybe as a related question, how are you guys providing the picks and shovels for this agent [indiscernible] without absorbing all the GPU costs that some of the hyperscalers are?

Yes. Thanks, Matt. As I said on the call, there's not another CEO that I would trade places with. I think that we're perfectly positioned for this moment, and the team is extremely excited for it. Obviously, there's there -- I think there's kind of -- oftentimes the market, strong reactions in both in both directions. And I don't think every SaaS company is dead. But I do see just the rate at which entrepreneurs and developers are able to build new disruptive ways of going on, what have been legacy incumbents in a way that I've never seen before. And what I'm proud of is a lot of them are building that on top of Cloudflare. I think that on the other side, of the equation. You've got a bunch of the agents of the world that are interacting with the Internet and they're interacting with it at a volume that we've just never seen before. And that's just driving more need for what are classically Cloudflare services. So the fact that more than 20% of the Internet sits behind us means that the agents have to interact with us, which means we have a seat at the table in defining exactly what the rules and the rails and the guardrails of the future of [indiscernible] is going to look like; and b, and we are sitting in the middle of that. And so it's been amazing across our business. We've seen, obviously, an enormous amount of strength from our Cloudflare Workers platform, but our Zero Trust platform is getting more and more adoption because it's built from a sense of a genetic AI from the beginning, our Stanford application services platforms are getting more adoption because people need to be able to keep up with all this new traffic that's coming to the Internet. And as we're thinking about really, what's the future business model of the Internet going to be and Cloudflare's in a position to help define that. And so that's been great. In terms of the CapEx spend, I think we're -- I think we're in a different business in the hyperscalers. The hyperscalers are in the business of buying machines and then renting them back so that they basically get 5x what they paid for them over the course of their useful life. And again, it's a great business for them, but it's not the business that [indiscernible] the business of getting work done. And so what we are constantly doing is having research teams inside of Cloudflare figure out how you can run AI workloads significantly more efficiently. The hyperscale is actually have no incentive to do that. They don't want AI workloads to be more efficient because that just means you have to lease your machines from them. Whereas we -- because we only charge you for the actual work that's getting done, that means that we're just getting oftentimes as much as 10x the amount of work off of the same GPU that you might get with the hyperscaler. That advantage is part of how we're able to just bring much more out of the CapEx that we spend than others are. Our CapEx has ticked up a little bit, and I think that that's in response to the fact that we've seen an increase in terms of workers, but it's nowhere close to what we're seeing from the hyperscalers. And we continue to think that we will be able to invest behind the demand. And so I hope you'll -- I'm sitting in the Alps right now at the Olympics with Michelle actually. And so I hope you'll forgive it. But I do feel at times like we're kind of goldilocks where we have a business model, which is -- which I think is going to sustain whatever AI brings. AI agents aren't going to replace the massive global network that we have. But we've also done it in a way which is much more capital efficient than some of the hyperscalers. And again, I feel very fortunate to be running this company right now and incredibly proud of our team.

The next question comes from Keith Weiss with Morgan Stanley.

Congratulations on an amazing quarter and a really amazing year that you guys put together. I think maybe to continue on the discussion you're just having, Matthew, a lot of what's happening with how agents are using the Internet, how Internet traffic is evolving, it's interesting because it involves the application services that you guys have had on board for a while, but also makes use of a lot of the newer capabilities coming on with workers and what you're doing there. Can you talk to us a little bit more about that how the -- you're seeing the evolution in that Internet traffic and where does it go next? And where do some of those like newer acquisitions like human native and [indiscernible] where do they fit into that vision?

Yes. So we're always on the lookout for really smart teams that have done great things to help bring them on board and incorporate them into Cloudflare. And I think both with [indiscernible] they perfectly fit that model. In human native case, they're really helping us think through what is the next business model of the Internet going to look like. It's going to move, I think, away from advertisement that's going to move away from subscriptions. It's going to move to something else and human native team who came out of Google and we are just extraordinary and thinking about what that future business model looks like. I think that you're going to see extraordinary things from them and they fit right in a cloud sowed we're excited to have them. In terms of Astros one of these incredible developer platforms and frameworks, which underlies a lot of what is building kind of the next generation of the web. And it is powering a lot of the [indiscernible] systems and targets that people are developing new applications for. And again, it just made a ton of sense. -- for them to sit in. What I think is maybe what I think our role is at some level is both to help invent what the future looks like. but then also to take the Internet that we've had, the Internet that we've built over the last 40 years and bring it along with that. And so what we're always trying to do is both work with great companies like entropic on their MCP protocols, work with everyone that's in the space trying to define what that future looks like. But then because our application services sit in front of people and 1 of the things that people don't understand is, there's a lot different than than what people think of sort of just traditional CDNs or other things like that, is that we're actually able to rewrite the content that flows through us as it flows through. So if it turns out that agents are better at speaking, I don't know. Latin than they are speaking English. We can literally rewrite the content that's behind [indiscernible] rather than being in English. Now that's not going to be what agents are good at, but they are going to be better probably at speaking code than they are going to be maybe speaking on other things that we might invent. So I think that what we're able to do and part of the reason we think that our legacy business is going to be incredibly durable is that it's going to be able to automatically bring along all of the rest of the Internet that all rate sits behind us into whatever comes next. And I think we're going to figure that out. So one of the things I'm thinking a lot about is what happens to small businesses in a genetic commerce world. There's a lot of ways where agents could be very consolidating and actually put a lot of pressure on small businesses. And so I think us in combinating with great companies that we're working with, like a Shopify or a Visa or PayPal or MasterCard, we've got to figure out how do we make sure that we bring all of these small business along, give them the right tools. And that's exactly the sort of thing that we're thinking about as we think about -- and it's not going to require you to have to go in and rebuild things. We want to make it 1 click simple where as soon as we figure out this is what really works, you push a button in that just whatever you had as your old doping marketplace, that just comes along with it and gets to support whatever agents are going to be providing in the future. I don't know exactly what all those things are going to look like, but we've got an incredible team. We're making smart acquisitions to augment that team and great partnerships with terrific companies try to think about what that future is. But one way or another, we're at the center of that future.

The next question comes from Jonathan Ho with William Blair.

And hope you are enjoying the Olympics. It seems like most of the coating platforms are either built on cloud layer workers or have Cloudflare as their preferred deployment target. Can you give us maybe some examples of how companies are using cloud layer workers for their AI projects? And why are these companies maybe picking cloud flare versus alternative platforms on hyperscalers?

Yes. I mean -- one of my favorite things is to sit down with [indiscernible] who used to work run platform at Cloudflare that I want to get closer to customers and say she's actually running our go-to-market effort around our developer tools. And she just had story after story of amazing examples. And so -- some of the ones you shared with me recently, there's a [indiscernible] pharmaceutical company that literally built a 5 coding platform on cloud, where their internal developers are using Workers AI and durable objects to build AI-assisted tools, which again, I think, directly proves that some of the largest enterprise customers are trying to figure out how do we stay on top of that? How do we give our teams the tools that we need, and they're turning to cloud layer because it's so easy to incorporate everything that we have and go from idea to application faster than you can anywhere else. -- and in a way that allows rapid prototyping that doesn't blow the budget, which is something that every company is trying to figure out. Another publicly traded technology company is migrating their plug-in sandbox infrastructure to cloud containers for secure isolated execution of code at scale, which let their customers then prompt deployments directly to their system, but do it in a way which is secure because one of the things that's really scary sometimes about deploying AI tools, especially to customer-facing applications is there can be a lot of damage that they do with one of these -- these agents goes rod or something goes wrong, the way that we've architected sandboxes allows them to -- and containers allows them to do this secure isolated code deployment. And again, it all comes as part of the toolkit of Claude workers, which is allowing them to go really quickly. And another example, a leading financial services company has partnered with us to launch an official MCT server designed to allow AI agents like [indiscernible] Cursor or OpenAI to interact directly with the company's payment services. The whole thing is built on cloud for workers, and this allows merchants to manage commerce tasks, such as creating invoices, checking transactions, processing and payments using natural language command and using things that are running on cloud flare. So again, it's so fun to get the download from Ali, where she's just listing a big company after a big company after a big company in addition to all of the startups and everyone else is building on the platform. And I think what they like about us is, first, you get a complete toolkit. Second, that toolkit has been architected in a modern way to build exactly what you need for agents and AI applications. And then third, you get it in a way that can scale up infinitely if it becomes wildly popular and can scale down instantly to 0. So you don't blow the budget if somebody is not actually using the system. That's very different than the hyperscalers, which in order to be able to get access to a GPU at a hyperscaler, anything close to a competitive price, you also have to commit leasing that server for an entire year, which, again, if the project that you're leasing it for doesn't go well, that's out of your budget. And so I think the fact that we've built, again, a complete toolbox which is a modern architecture, how developers want to build applications and then deliver it in an incredibly cost-effective way. That's what is driving so much success from our workers' developer platform.

The next question comes from Gabriela Borges with Goldman Sachs.

Matthew and Thomas, I wanted to ask you a little bit more about the large deal that you closed this quarter and it reminds me a little bit of the $100 million-plus deal from last year. Give us a sense on how the pipeline is trending for these types of deals? What sort of engagement across the act that you're able to convert when you have these larger deals? And what are you learning from the 2 sort of large deals that you've closed so far, how do you apply those to get the conversion rate up for whatever in the pipeline?

Yes, I'll start, Gabriela, and then Thomas might have a little bit more to add. What I'd say is we see incredibly strong pipeline. And I think where that pipeline comes from is typically customers that are already using us us getting to know them, often embedding developers directly with their teams to understand what their use cases and needs are. and then figuring out where there are opportunities for us to be able to do something for them that maybe they had budgeted it with a hyperscale or may be budgeted somewhere else and they're all of a sudden able to realize that they do it with [indiscernible], they can do it faster. They can do it in a way which is much more scalable and they can do it in a way which is much, much more cost effective. And so I think that the work that we're doing to really embed with customers is driving success there. And again, we're still not to a point where we're going to be doing a $100 million deal a quarter, but we will get to that point. And I think we've seen an enormous total addressable market for the Cloudflare Workers platform. And I think that will shift more and more spend away from what people are using the hyperscalers for. So I feel very confident. And again, I think with the success has really been how do we take all these amazing tools that we have, but how do we build those deep relationships with customers in a sophisticated go-to-market organization. And I think you'll just see more and more success from that.

The next question comes from Fatima Boolani with Citi.

Matthew and Thomas, this is for both of you, I'd love to hear you both opine on this. The pool of funds contracting and engagement vehicle has seen a tremendous amount of success you've only compounded the credibility and trust with that sales motion. So I was hoping you could give us a sense of what the next phase and atom of funds traction is for the business? And Thomas, any sense of how we should think about higher variability, more stage oriented revenue creating more volatility in the model or revenue forecasting rather? And how we should think about that and internalize that as we think about the the model over the next year or 2?

Yes. Fatima, I'll start and then Thomas can answer the second question. So I think what [indiscernible] really represents is customer trust and their commitment to knowing that they may not know exactly what features they're going to use. They may not know exactly how much they're going to use. But they know that they believe that Cloudflare is going to provide a series of tools and features that are going to be critical to their business, and they know that they want to have access to any of those tools regardless of what needs come up. And so I think that the -- your question was sort of what's the next -- what's our next trick. I think we've got a long way to go with this trick. It still represents a relatively low percentage of revenue, although growing quickly. And so I think that that's going to be something as we build deeper relationships with customers as we make sure that they understand everything that we have to offer that we'll be able to deliver across that broader set of speakers. And Thomas, I'll turn it over to you to talk about severity.

Yes. Maybe some more color on pool of funds in the fourth quarter was about 20% of ACV in the fourth quarter booked around mid-teens for all 2025 with a higher share of pool of funds, we have a higher share of variable revenue in the overall number. So there is some volatility that you might expect for that, and we are trying to compensate for that as we guide. I think it's fair to say as revenue from all of us still has grown and continues to grow. We can expect that sequentially bigger Q4 and sequentially smaller Q1, I think that as compared to just linear revenue recognition deals.

The next question comes from Gray Powell with BTIG.

Great. Congratulations on the really strong results. So I know there's a lot of excitement around the newer products, particularly developer services, we've also been picking up good data points just around an increased need for things within core application services. And I know you hit on this in the prepared remarks, but could you talk about -- can you just talk more about what you're seeing there, particularly as customers look to deploy AI. How do you see the material pickup in Act 1 relative to prior quarters? Just how should we think about the growth profile there?

Yes. I think that -- so Gray, first of all, thank you for asking about some of the other products, the teams that work hard on them every day are proud of building the best products that they have in the world. And even though they aren't the ones that always get the most attention, they still form a lot of the foundation of what we do at Cloudflare and they're still incredibly important to us. I think in this particular case, the first place that we saw just demand was actually from a lot of the AI companies, where the companies would say to us we can't continue to operate our systems unless we can have the security and ability to deal with the load which Cloudflare provides by default. -- every time you run a query against an AI company, it's pretty expensive to deal with those queries. And so being able to sort out who's a [indiscernible] not a human, which is something we're the best in the world at. is really important for the companies, and that's driven actually just a lot of those initial relationships that are there. What really took off in Q4, though, was where we saw other companies, media companies, e-commerce companies, companies that were just doing more traditional things online, seeing such an enormous uptick in how agents were interacting with their systems. I mean we have -- if any of you have used a tool like ChatGPT or grok or cloud. And you just watch how many different things it is looking at for every query that you send out, this is an order of magnitude increase in the volume of queries that are coming to the Internet. And so the people who are providing what is that Internet that they're clearing against. They need ways to do that in a way which is efficient and able to continue to scale. And Cloudflare is -- and again, those application services functions that we have, the kind of Act 1 products that we have, are really critical of being able to deliver that. But it goes beyond that. If you look at something like the new agents that people are running on their own machines [indiscernible] the amazing thing is that people are waking up very quickly. We're sort of speed running all of the security challenges that are out there, where all of a sudden they -- I've just given my agent access to everything in my life what could go wrong. People are very quickly figuring out a lot to go wrong as you got to put controls in place. And that's exactly where our active or Zero Trust products come into play, where we've actually seen a real uptick even in a self-service business of the Zero Trust products. Now that's not going to deliver huge revenue from people that are just self-service home customers. But as we see those reps as we become more familiar in the market for those products that we're able to provide, that's just going to help us sell even to a large [indiscernible] business. So again, across the AI spectrum, I know that AI is putting pressure on some companies that are out there. it's not putting pressure on class ware. We are seeing it as nothing but a tailwind for us. both for our developer tools and kind of the Act 4 stuff that we're working on, but actually for even our legacy products like Application Services at Zero Trust as well.

The next question comes from Shrenik Kothari with Robert Baird.

Congrats on the great execution and really excited what the opportunity is ahead. So Matthew, just double-clicking into [indiscernible] in light of the wins the media company signing that 3.1 million contract for AI crawl control. So as you're engaging with publishers, can you share early feedback around adoption towards this opt out controls to block scraping, but also the evolution of a structured marketplace model here. And I had a quick follow-up.

Yes. So I think this is an exciting area. It's an area that I think we've been leaders in. And we're sort of in this -- we're in this interesting position because we're not an AI company. We're not building a foundational model. We're also not a media and content company. but 80% of the AI companies use us. So we know them. We have deep relationships with them and then a huge percentage of the Internet users as well and increasingly more and more of the world's publishers use us. And so I think we've been sort of that neutral honest broker between the 2 sides that can come together and say, okay, like in order for this to all work, the Internet needs to have a business model like people who create content deserve to get paid. And one of the things that actually surprised me to some extent, which might be relevant to a lot of you listening in, is we've actually been getting called not just from like Associated Press and BBC and New York Times. We've been getting called that increasingly from banks where their research teams are saying, we're actually seeing fewer people subscribe to and read our research because the AI companies, the people are just turning the [indiscernible] companies that are stopping all the data down and taking that intellectual property. Again, I think journalists get deserved to get paid, but so do research analysts. And so all deserve to be able to monetize the work that you're creating, and we're going to have to create some new business to support that because AI bots don't click on ads, and it's going to be something which is is different going forward. The reaction from the content creator side has been just overwhelmingly positive. And we come back to something pretty simple, which is just if you create content, it should be up to you who gets access to it and who doesn't, and we can provide the tools to do that. On the AI company side, they also -- again, nobody wants to pay for something that they were getting for free. But I think that they understand that we're a fair broker. And when we walk them through what happens if we don't create some healthy ecosystem here, they say, we get it. We just want to make sure that everyone is treated fairly. And so I think that's a big piece of it. We've seen a lot of other companies start to try and kind of follow in this trend. Microsoft and Amazon have announced content marketplaces. And they may be successful, but what we're hearing from both the AI companies and from the content creators is that because Cloudflare is that trusted neutral third party that we can be that honest broker between them that they would rather us be the one that figure out what that future business model looks like as opposed to 1 of the hyperscalers, which is they're creating their own foundational model themselves and might have a very different incentives. So I think 2026 will be the time that we start really talking about what this future business model looks like and how that is going to impact us financially. But we are mostly focused on something which is really simple, which is believe the Internet is important. We believe AI is incredible. But in order for AI to be even more incredible, the Internet needs to be even more robust and full of great original content. And in order for that to happen, there has to be a way for those content creators to get paid by AI companies. And that's what we're trying to figure out.

The next question comes from Adam Borg with Stifel.

It's great to see the continued go-to-market maturation. And Matthew, so when maybe thinking about the channel mix, I think that picks up to over 28.5%. So it's a good ongoing momentum there. If you think about the new fiscal year, maybe to provide a quick update on the channel priorities especially in the context of driving [indiscernible]

Yes. So I'll start and then Thomas may have a little more to add. I think one of the things that was really important to Mark Anderson was that in order to really be able to scale and get leverage in a true enterprise sales team, you need the channel first. And so I think that's something that we really embarked on 2 years ago. I think we had the willingness to do it. I think we had the margin structure to be able to support it. I think that we were thought of well, and we're a good partner generally to channel partners and have been what I think was missing for a long time for us was, first, we -- our products for a long time were too simple. It was like you -- tagline our site used to be give us 5 minutes and we'll supercharge your website, like there's not a lot of value the value-added reseller can add if they're doing that today, when you look at the large [indiscernible] resellers, when you look at the giant systems integrators, they're seeing that there's an enormous amount of complexity. There are places where there's integration with other products. There's also, because of Workers to the developer platform and ability for them to actually create their own intellectual property and then sell that over and over again in specific verticals or specific industries and that's really taken off. I think we started to [indiscernible] out really 2 years ago. I think what we've really focused on over the course of the last -- of 2025 was how do we now make it easier for channel partners. How do we actually publish a price list where they can quote deals without having to check with our team and doing all of that work for us. And that was complicated work but we did that in the second half of 2025. And you're going to see the fruit of that through 2026. So we just had [indiscernible] kickoff a couple of weeks ago in Las Vegas. And it was just extraordinary to talk to and listen to partners who, again, I think for a long time, we're like, hey, we want to work with you, but you don't quite understand us for now they're saying, you've got the team that understands us you've made our jobs simpler, and we're going to go out there and just sell the hell out of Cloudflare products. And I think that is how you become again, a true enterprise sales company, and I think Mark Anderson and the rest of his team deserve an enormous amount of credit in teaching me and the rest of the organization how to do that. And a big part of that is about channel and partners.

Your last question comes from the line of Jackson Ader with KeyBanc Capital Markets.

Matthew, thank you for recognizing the plate of research analysts. I'll make sure I pass that along.

You definitely deserve to eat, like, definitely -- it's important. AI should not take your job, Jackson.

It's common. Yes, we'll see. My question

though is actually just following up on the channel discussion that you just went through. And I'm curious one of the only down thinking about about a channel presence is that it does make you a little less nimble, right? Like you could -- if you wanted to add sales heads, you could do that a little faster than maybe trying to supercharge the channel. And so I'm curious how you guys are thinking about hiring direct salespeople, if you see further demand signals like if we're going to continue to accelerate. How nimble can you be like really lean [indiscernible] the go-to-market motion?

Yes, I'll start and Thomas will say something smarter than I did. But I think that the channel makes you more nimble, not less nimble. We are going to continue to hire great people internally. I think we are hiring and ramping at a rate, which is very, very healthy. But then the channel gives you leverage to be able to do that and execute as well. And so I don't think that that's -- I don't see it as a constraint on our nimbleness. I actually see it as a way of offering leverage to our already very successful inside sales and field sales teams that then use the channel to be able to support themselves. And again, I think it's been -- we've seen -- I think it's still early days, but we've seen through the growth, just real success there. And I don't think we'll ever get to 90% plus of business going through the channel like the hardware companies of all some of the software companies of old, even companies like Zscaler and Net scope that have gone there, I think we will be less than that. But I do think that we can do a lot more with the channel. And I think we've got a lot of what had been the things that have held that back, cleared out of the way. And I think that will actually make us faster and more nimble rather than slow us down.

The only thing I would add is, for us, this is not an either or from a capital allocation perspective, but how we stack both vectors on top of each other, so you can accelerate both prospectus independently, and we've been really successful at this. With the productivity metrics that we shared, we are full speed ahead on on adding sales capacity, and we do this early in the year just to get a head start.

This concludes the question-and-answer session. I'll turn the call to Matthew Prince for closing remarks.

I just wanted to thank everyone in the Cloud Star team for an incredible 2025. And there's a ton of work that goes into having a quarter like the one that we just did, I'm proud of everyone who helped achieve it. We got 2026 ahead of us. And again, as I said, I couldn't be more excited to be the CEO of Cloudflare and we're [indiscernible] places with anyone. Thank you all. See you back to your next quarter.

This concludes today's conference call. Thank you for joining. You may now disconnect.

Thank you for standing by. My name is Greg, and I will be your conference operator today. At this time, I would like to welcome everyone to today's Cloudflare Q3 2025 Earnings Call. [Operator Instructions]

I would now like to turn the call over to Phil Winslow. Phil?

Thank you for joining us today to discuss Cloudflare's financial results for the third quarter of 2025. With me on the call, we have Matthew Prince, Co-Founder and CEO; Michelle Zatlyn, Co-Founder and President; and Thomas Seifert, CFO.

By now, everyone should have access to our earnings announcement. This announcement as well as our supplemental financial information may be found on our Investor Relations website.

As a reminder, we will be making forward-looking statements during today's discussion, including, but not limited to, our customers, vendors and partners, operations and future financial performance, our anticipated product launches and the timing and market potential of those products, our anticipated future financial and operating performance and our expectations regarding future macroeconomic conditions. These statements and other comments are not guarantees of future performance and are subject to risks and uncertainties, much of which is beyond our control.

Our actual results may differ significantly from those projected or suggested in any of our forward-looking statements. These forward-looking statements apply as of today, and you should not rely on them as representing our views in the future. We undertake no obligation to update these statements after this call. For a more complete discussion of the risks and uncertainties that could impact our future operating results and financial condition, please see our filings with the SEC as well as in today's earnings press release.

Unless otherwise noted, all numbers we talk about today, other than revenue, will be on an adjusted non-GAAP basis. You may find a reconciliation of GAAP to non-GAAP financial measures that are included in our earnings release on our Investor Relations website. For historical periods, a GAAP to non-GAAP reconciliation can be found in the supplemental financial information referenced a few moments ago.

We would also like to inform you that we will be participating in RBC's Global Technology, Internet, Media and Telecommunications Conference on November 18 and Needham's 6th Annual Tech Week on November 24.

Now with that, I'd like to turn the call over to Matthew.

Thank you, Phil. We had an extremely strong Q3. We achieved revenue of $562 million, up 30.7% year-over-year. Great companies innovate and execute, and I think we owe our reacceleration of revenue growth to doing both of these things very well. We now have 4,009 large customers, those that pay us more than $100,000 per year, a 23% increase year-over-year. Revenue contribution from large customers grew 42% year-over-year, contributing in total 73% of our revenue during the quarter, up from 67% in the third quarter last year.

Our dollar-based net retention was 119%, up 5 percentage points quarter-over-quarter. Our gross margin was 75.3%, within our long-term target range of 75% to 77%. We delivered an operating profit of $85.9 million, representing an operating margin of 15.3%, and we generated strong free cash flow of $75 million during the quarter, again exceeding expectations.

Our go-to-market transformation, evolving from purely product-led growth to true enterprise sales continue to track along. Growth in net capacity of our sales force grew at its fastest pace year-over-year in more than 2 years. Sales productivity increased year-over-year for the seventh consecutive quarter. Close rates ticked up notably both year-over-year and quarter-to-quarter. Bookings from partner-initiated opportunities doubled year-over-year. Gross retention levels increased year-over-year and quarter-to-quarter. And new pipeline attainment again exceeded our expectations. Across the company, the team is firing on all cylinders.

One bit of disappointing news is that CJ Desai is going to be leaving Cloudflare. CJ called me some time ago to talk about an opportunity he's been approached to be the CEO of an exceptional public technology company. He was torn because he loved his team, the work and the mission at Cloudflare. But since his first job in technology over 25 years ago, he dreamed one day of being the CEO of a great public company. We talked through the opportunity, his career goals and what's great and not so great about being a public company CEO. In the end, while I'm sad to see him go, I'm excited for him to get to helm his own ship.

I want to give CJ an opportunity to say something on this call, in some ways, as practice for his many earnings calls to come. CJ?

Thank you, Matthew. This was an extremely hard decision for me as I love the team and mission of Cloudflare, and I see incredible opportunities ahead. I really appreciate the support as I figured out what was right for me. This job at Cloudflare is the coolest Product & Engineering job in tech today. And I will help ensure whoever feels the seat next will be world-class. I'm incredibly bullish, as you know, on Cloudflare's future. I'll miss you all, but will always be among your biggest fans.

Thanks, CJ. I appreciate how you brought a customer-first focus to Cloudflare's already powerful innovation engine. That's made us a better company able to win bigger deals. It's now part of our DNA that you deserve credit for having helped shape. And while I'm bummed you're leaving, I'm proud that Cloudflare is a place that has trained the leaders of other great technology companies. You're our second product leader in a row to be recruited away to be CEO somewhere awesome. We can't say yet where you're going, but they're lucky to have you, and I have no doubt you'll bring some of Cloudflare's relentless culture of innovation to them.

With that out of the way, let's talk about some of our wins in the quarter. A Global 2000 digital media platform expanded its relationship with Cloudflare, signing a 3-year $22.8 million pool of funds contract for application services and workers. This contract marks the culmination of a powerful comeback story. We actually lost this customer to a competitor in 2016, but the Internet and Cloudflare evolved. We earned their trust back in 2023, starting with our Zero Trust portfolio.

During 8 months of testing before signing this deal, our world-class security, unmatched product breadth and powerful Workers platform ran circles around the incumbent. But that's not the whole story. The decisive factor of the win was AI. This customer looked at the landscape and correctly identified Cloudflare is the only company building the essential platform to protect and manage content for the emerging AI-driven web. This strategic win established us as the customer's clear forward-looking partner and creates a direct on-ramp for Pay Per Crawl, which could transform Cloudflare from a vendor they pay for services into a powerful revenue generator for their business. We and they believe that this is what the future looks like.

A leading European technology company expanded its relationship with Cloudflare, signing a 5-year $34.3 million contract, representing an upsell of $6.8 million for Workers platform and application services. This customer is fully redesigning their architecture to move their front end on to Workers and Durable Objects. The decision to commit to a 5-year term underscores the customer's view of Cloudflare as a critical long-term strategic partner.

A rapidly growing media platform expanded its relationship with Cloudflare, signing a 3-year $15 million contract for Workers and Application Performance. This customer was experiencing significant egress fees, high latency for its global customer base and vendor lock-in with a hyperscale public cloud provider. Moving to Cloudflare will enable data to be processed and served closer to their end users, delivering superior performance and eliminating egress fees. With our unified platform, this customer will be able to drive down their total cost of ownership by more than 30%.

A Fortune 500 financial technology company expanded its relationship with Cloudflare, signing a 2-year $16.1 million pool of funds contract with an upsell of $4.6 million for application services and workers. As a textbook land-and-expand journey across 3 apps, this customer started with Cloudflare's application services in 2022, expanded with our Zero Trust platform in 2023 and has been adding a number of products from our Workers platform over the last 2 years. Another Workers deal is already underway for AI use cases.

A Global 2000 European pharmaceutical company expanded its relationship with Cloudflare, signing a 3-year $12.4 million contract with an upsell of $4.5 million. This is a great example of platform adoption as the customer is utilizing products from our first 3 apps, application services, SASE and Developer. This customer views Cloudflare as a critical strategic partner choosing to displace services from 2 hyperscale public clouds and multiple-point solution providers because according to them, "It's so much easier to build on Cloudflare."

A U.S. cabinet-level agency expanded its relationship with Cloudflare, signing a 2-year contract exceeding $20 million for our complete FedRAMP portfolio. The agency is standardizing its network and security platform on Cloudflare, displacing over a dozen legacy point solutions and generating more than $10 million in annual cost savings. We are seeing more traction than ever before across U.S. government as it looks to modernize its digital infrastructure.

A Fortune 100 financial services company signed a 3-year $4 million contract for Magic Transit and Advanced Magic Firewall. Recent outages, capacity limitations and a lack of automation features with 2 legacy incumbents left this customer with DDoS vulnerabilities at their network layer in a time when we're seeing new record-breaking DDoS attacks every few weeks, like the nearly 30-terabit per second attack we mitigated earlier this month. Cloudflare won because our fundamental architecture advantage gives us literally 4x the capacity of all our scrubbing center-based competition combined. As the Internet gets scary and scarier, customers are realizing Cloudflare is the only network engineered to survive.

A global industrial company signed a 3-year $2.2 million contract for a complete SASE portfolio, including Access, Gateway, Browser Isolation, CASB, DLP, Magic WAN and Magic Firewall to consolidate and modernize their security stack. We're displacing a first-generation Zero Trust vendor as well as a legacy on-premise VPN provider, which were expensive and difficult to maintain across their global operations. This customer chose Cloudflare for the operational simplicity of our unified platform that delivers both superior performance and significant cost reduction.

A global web infrastructure platform expanded its relationship with Cloudflare, signing a 14-month $1.2 million contract for AI Crawl Control and Bot Management. This customer is experiencing a massive surge in AI scrapers and malicious bots hitting their origin servers, inflating costs without revenue conversion and obscuring visibility into legitimate traffic. They selected Cloudflare for our innovative best-of-class bot blocking capabilities in addition to seamless expedited deployment by our deep platform integration. We're already exploring a much larger opportunity with this customer for Pay Per Crawl.

We talked last quarter about how the rise of AI would impact media companies. Cloudflare has emerged as a strategic partner to these firms as they work through what the new business model of the Internet will be. But it goes beyond just media. Businesses of all shapes will be transformed by the rise of AI. I don't think people yet appreciate how AI is another massive information consumption platform shift, just as we move from consuming information via a browser on a desktop to social media and then to apps on mobile devices, AI is another information consumption platform shift. It changes where and how we will consume and interact with information.

With the last 3 platform shifts, the business model of the Internet remains the same: create content, generate traffic and then sell things, subscriptions or ads. With AI, for the first time in a long time, the fundamental business model is going to change. Human eyeball traffic is unlikely to be the currency of the Internet's future. We already can see glimpses of that future. It's represented in SciFi. When George Jetson asks his helpful robot Rosie for a recipe for cookies, the response isn't 10 blue links to hunt through. It's a recipe for cookies. Most of us are increasingly living in some version of that future now with tools like ChatGPT, and it seems inevitable that more and more commerce will be facilitated by AI-powered agents working on our behalf.

As that happens, new questions will arise. What happens to small businesses? What happens to brands? Brands, of course, are just shortcuts for humans to be able to assess quality and value. What do they mean in the world of agentic commerce? I don't know what the future business model of the Internet will look like, who the winners and losers will be, but I do believe Cloudflare will help shape it. We estimate 80% of the leading AI companies already rely on us. A huge percentage of the Internet sits behind us. The agents of the future will inherently have to pass through our network and abide by its rules. And as they do, we will help set the protocols, guardrails and business rules for the Agentic Internet of the future.

And we'll make sure the tools to participate in that future are available to all businesses, large and small. It's what we've always done. Again, we don't know exactly what the future will look like, but I believe Cloudflare will be one of the key players helping shape it. What we are playing for is a world with as many AI companies, media companies and businesses, large and small, competing fairly to best serve customers anywhere and everywhere they and their agents transact.

I'm really excited for that future, and I'm optimistic about it. But to bring it back to the present, let me hand it off to Thomas to walk through this quarter's financials. Thomas, take it away.

Thank you, Matthew, and thank you to everyone for joining us. We are pleased with our strong third quarter results that underscore how our strategy for delivering continued innovation and accelerating growth while also maintaining a relentless focus on operational excellence is working.

Revenue growth accelerated for the second consecutive quarter to 31% year-over-year, providing clear evidence of the momentum building in our business. We complemented this robust growth with a highly balanced operating plan, investing significantly in our innovation pipeline and expanding our go-to-market capacity while simultaneously remaining committed to the strong unit economics of our business to drive operating leverage and deliver compounding shareholder value.

Turning to revenue. Total revenue for the third quarter increased 31% year-over-year to $562 million. From a geographic perspective, the U.S. represented 50% of revenue and increased 31% year-over-year, which is up nearly 10 percentage points sequentially. Growth in the U.S. region was primarily driven by strength with partners, our workers developer platform and large customers, including pool of funds. EMEA represented 27% of revenue and increased 26% year-over-year. APAC represented 15% of revenue and increased 43% year-over-year.

Turning to our customer metrics. In the third quarter, we had approximately 296,000 paying customers, representing a record net addition of nearly 30,000 paying customers sequentially and an increase of 33% year-over-year, driven by an uptick in customers, including those graduating from the free tier to small paid accounts for developer platform products around our AI Week and birthday week. We ended the quarter with more than 4,000 large customers, representing an increase of 23% year-over-year.

Revenue contribution from large customers increased to 73% of revenue during the quarter, up from 67% in the third quarter last year. We again saw particular strength in our largest customer cohorts. For the fourth consecutive quarter, we added a record number of our largest customers year-over-year, those that spend over $1 million and $5 million with Cloudflare annually. Accelerating sequential and year-over-year revenue growth from both of these cohorts served as a significant tailwind to our expansion business. As a result, our dollar-based net retention rate accelerated to 119% during the third quarter, up 5% sequentially and 9% year-over-year.

Moving to gross margin. Third quarter gross margin was 75.3%, remaining within our long-term target range of 75% to 77% and representing a decrease of 100 basis points sequentially and a decrease of 350 basis points year-over-year. During the third quarter, paid versus free customer traffic again increased both year-over-year and quarter-to-quarter, resulting in a higher allocation of expenses to cost of goods sold from sales and marketing.

Our Workers developer platform continues to deliver outsized growth with the world's most innovative companies increasingly adopting Workers for running AI inference tasks as well as building AI agents and full stack applications. While the relative revenue contribution across our 4 Acts can impact near-term gross margin, the unit economic margin of our business remains very consistent. Network CapEx represented 14% of revenue in the third quarter. We expect network CapEx to be approximately 13% of revenue for full year 2025.

Turning to operating expenses. Third quarter operating expenses as a percentage of revenue decreased by 4% year-over-year to 16%. Our total number of employees increased 16% year-over-year, bringing our total headcount to roughly 4,800 at the end of the quarter. Sales and marketing expenses were $201.2 million for the quarter. Sales and marketing as a percentage of revenue decreased to 36% from 37% in the same quarter last year. Research and development expenses were $82.5 million in the quarter. R&D as a percentage of revenue decreased to 15% from 16% in the same quarter last year.

General and administrative expenses were $53.5 million for the quarter. G&A as a percentage of revenue remained consistent at 10% compared to the same quarter last year. Operating income was $85.9 million, an increase of 35% year-over-year compared to $63.5 million in the same period last year. Third quarter operating margin was 15.3%, an increase of 50 basis points year-over-year.

Turning to net income and the balance sheet. Our net income in the quarter was $102.6 million or diluted net income per share of $0.27. Free cash flow was $75 million in the quarter or 13% of revenue compared to $45.3 million or 11% of revenue in the same period last year. We are comfortable with consensus free cash flow estimates for the fourth quarter of fiscal 2025. We ended the third quarter with $4 billion in cash, cash equivalents and available-for-sale securities. Remaining performance obligations, or RPO, came in at $2.143 billion, representing an increase of 8% sequentially and 43% year-over-year. Current RPO was 64% of total RPO.

Moving to guidance for the fourth quarter and full year 2025. For the fourth quarter, we expect revenue in the range of $588.5 million to $589.5 million, representing an increase of 28% year-over-year. We expect operating income in the range of $83 million to $84 million, and we expect an effective tax rate of 20%. We expect diluted net income per share of $0.27, assuming approximately 377 million shares outstanding.

For the full year 2025, we expect revenue in the range of $2.142 billion to $2.143 billion, representing an increase of 28% year-over-year. We expect operating income for the full year in the range of $297 million to $298 million, and we expect an effective tax rate of 20%. We expect diluted net income per share over that period to be $0.91, assuming approximately 370 million shares outstanding.

In closing, the strength of our third quarter results confirms that our strategy to deliver continued innovation with accelerating growth and strong unit economics is driving significant and measurable value. At the beginning of the year, we committed to reaccelerating revenue growth over the course of 2025 on the way to our goal of achieving $5 billion in annualized revenue by the fourth quarter of 2028.

Our performance over the last 2 quarters demonstrates that we are effectively executing against both of these objectives. In fact, we expect to reach a $3 billion annualized revenue run rate in the fourth quarter of 2026 on our journey to $5 billion and beyond. This trajectory reinforces our conviction in our strategy and our ability to deliver exceptional long-term value for our shareholders and customers.

Before opening it up for questions, I would also like to extend my personal thanks and congratulations to CJ. The processes, discipline and leadership bench she established at Cloudflare will enable our innovation engine to continue to scale well beyond his tenure. All of us at Cloudflare with CJ continued success in his next chapter. And with that, operator, please poll for questions.

[Operator Instructions] And our first question today comes from the line of Matt Hedberg with RBC Capital Markets.

Congrats on the results. And CJ, we look forward to hearing about your future role. Matthew, there were a lot of strong metrics this quarter, but 43% RPO growth that accelerated. I think that was the highest RPO growth that you guys have reported since 2022, certainly stood out. I'm wondering if you could provide a bit deeper dive into what drove that acceleration this quarter.

Yes. I'll start, and I think Thomas can probably add to it as well. I think we try to be a place that says what we do and do what we say. And so I think the real thing that's happening is we are transforming from being a product-led growth company to being a true enterprise sales company. So you're seeing the average tickets tick up. You're seeing the large deals tick up. And that's driving just success in taking what have always been exceptional products and getting them in the hands of customers. And so I think our sales team deserves a lot of credit for really just driving great execution.

What I would add is we are -- I think the RPO growth points to primarily 2 drivers, the customer quality and the platform expansion. We are seeing exceptional strength with our large customer cohorts, specifically those that spend more than $1 million or $5 million with us, both delivered record growth this quarter. And in addition to that strength is increased consumption of our large pool of fund customers, demonstrating I think, the increasing strategic importance of our platform for those large enterprises globally. And in addition to that, our Workers platform, the developed platform, including Workers AI is just providing to be a significant new vector for long-term commitment and with that growth.

That's great. Actually, could I double-click just Thomas, you mentioned the pool of funds, and I know you mentioned in your prepared remarks. But specifically, like how is that showing up in the results? You introduced that several, I think, years ago at this point now. But how is that driving some of this as well?

The share of pool of funds deal this quarter was again up. It's now low double digits of total ACV. And we are seeing now across our pool of Funds contracts an extremely balanced consumption of these contracts. On average, we're slightly ahead and that delivered to the strong performance in the quarter. So if you have a platform like ours with more than 55 revenue-contributing products now, you need a vehicle that allows frictionless adoption and consumption of these products. And I think the sales team and the organization at Cloudflare has become quite good at deploying these contracts and driving consumptions with customers.

The other thing that I'd add is I think where we saw downward pressure on things like dollar-based net retention as we rolled out pool of funds. As those pools are now getting consumed, you can see our dollar-based net retention is ticking back up. And so I think pool funds will show up in RPO, pool of funds as it initially puts downward pressure on things like dollar-based net retention, but you can see that, that's now ticking up again. And so just to reiterate what Thomas said, these are an indication of customers trusting us as a strategic vendor, making larger, bigger bets on us, and it is an undoubtedly positive sign for us as a strategic vendor to more and more large customers.

And our next question comes from the line of Adam Borg with Stifel.

Maybe for Matthew, on the sales productivity gains, it's been great to see that continue. Are we at a point now where these gains are beginning to flatten out? Or is there still room for this to continue to trend higher in the coming quarters?

I think that we think that these will continue -- that the productivity will continue to tick up in coming quarters. I think that the caliber of the team that we're bringing on, their ability to sell much larger deals, all of which contribute to having a much higher productivity from the sales team. And so we think that there is still headroom there. And then I think importantly, in addition to that headroom, we've turned the corner starting last quarter on having the ramped rep capacity also ticking up again. So I think we've gotten through what was a period of time where we really needed to revamp the sales team, and we're -- and now we're seeing the benefits of that coming out the other side.

That's great to hear. And maybe just as my follow-up, it was really interesting to see a few weeks back the integration with Oracle OCI that was announced. Maybe talk a little bit about what advantages does it provide to those OCI customers?

Yes. So we're really excited to work with Oracle. They've been a terrific partner for us over the years. They evaluated Cloudflare's products and realized that we were really the best of breed for what they could offer to their customers. And so Cloudflare will be natively available within Oracle's OCI platform, including across hybrid, multi-cloud and OCI hosted workloads, which gives us access to a large pool of customers and gives Oracle's customers access to Cloudflare's world-class tools.

I think one of the things that we're particularly aligned on is that we and Oracle both see the future as a multi-cloud future, where customers are going to have many different cloud providers. And what they need is one consistent interface where they can apply security rules, have consistent network performance. And Cloudflare is the best in the world at doing that. And so I think the fact that we have been able to work with Oracle, integrate our products directly into Oracle and Oracle's customers are going to be able to enjoy the benefits of that. That's great for us, but it's also great for Oracle, and we're excited to have them as an even more deeply integrated partner.

And our next question comes from the line of Gabriela Borges with Goldman Sachs.

Congrats on the quarter. Matthew and Thomas, I wanted to revisit your comment from earlier in the year about doubling your network capacity this year. So my question is, do you think that you're capacity constrained in Workers? To what extent are the capacity decisions that you're making this year essentially dictating a range of outcomes on what Workers revenue could be next year? And I know you have some really interesting thoughts on fungibility of workloads between CPUs, older gen GPUs and newer gen GPUs. So I would love to hear your comments there as well.

Sure. I don't think we're capacity constrained because of somewhat the nature of how we've architected Cloudflare and the philosophy of how we make CapEx and network investments. We always have tried to invest behind demand, not ahead of demand. And the thing that allows us to do that is that what we are selling is not a particular box in a particular place or a fraction of a particular box in a particular place. What we're selling is the ability to get work done across our network. And so Cloudflare itself is effectively a giant scheduler where we can move workloads to wherever we have capacity anywhere in the world. And the nature of the network is that it's always somewhere it's the middle of the night, and there's always excess capacity there.

Now that's not ideal, but the good news is that for some of our smaller customers or low-end customers or free customers, we can move them to places across the network that has that free capacity, still gives them a great performance. but then reserve the capacity that we have as close as possible to our largest customers. As we see that growth, that then means that we can invest behind it and be able to just make sure that we're getting the most utilization possible.

The other thing that I think is unique about us is that certainly versus the hyperscalers, the primary business of the hyperscaler is to essentially rent you a server or a fraction of a server, and they try to effectively get whatever they pay for the server back 5x over the life of the server. That's their business. Whereas we're about, again, getting work done for our customers. We're selling something different, which is a sort of level of abstraction up from that. What that means is that we believe it's our job, not our customers' job to make the utilization rates as high as possible, make our systems as efficient as possible.

And so it's been remarkable to see over the last 15 years, how our team has been able to squeeze as much as possible out of the CPU capacity that we have, where we can run that CPU capacity at 70% to 80% utilization and get more out of every CapEx dollar we spend. But what's fascinating is we're sort of speed running the last 15 years now with GPUs, where we're figuring out how to make GPUs multi-tenant, how to make them load and unload models more quickly and driving the utilization of GPUs up substantially. And so that is still well below what we have with CPUs, but we see no reason that we can't get GPUs also up to that 70%, 80% utilization.

And that, again, just means that every CapEx dollar that we spend, both can be behind the demand that we see. And then secondly, that we'll get more out of it, more effective value out of it for the services that we're delivering our customers versus some of the legacy hyperscaler models.

The additional point I would make is in addition to what Matthew said is that the supply chain within Cloudflare is so optimized to a large degree because we use off-the-shelf equipment and parts that we can deploy hardware, especially in Tier 1 cities and generate revenue even before we start to pay for the equipment. So not only do we have the flexibility that Matthew described really well at length, our reaction time to deploy hardware where we need it is really, really fast.

That makes sense. The follow-up is on competition for Cloudflare in the enterprise for securing those inference workloads and winning those inference workloads in particular. Matthew, I would love to hear you comment how do you think competition is evolving in the enterprise as you build out some of the breadth and depth of your functionality? And on the flip side, are you seeing anything new from newer platforms, newer cloud platforms that are AI native or inference focused?

I think that the primary competition for inference workloads continues to be the hyperscalers. And it continues to be the model of do you want to do this work yourself and have to optimize yourself or do you want to hand it off to Cloudflare. And I think in the cases where we're in the conversation, we're able to show that there's just a much better TCO, total cost of ownership, a much lower cost, much better performance when we manage that for you. And so there's kind of a standard way people do things, which is the hyperscaler way. We're having to teach them that there is a different way that's out there. But the primary competition still comes from the hyperscalers. And I think that we are finding, though, that once somebody learns that there's a better way that Cloudflare is very, very sticky, and we keep those customers over the long term.

And our next question comes from the line of Shaul Eyal with TD Cowen.

Congrats on the quarterly results. So many new product announcements in recent weeks during Cloudflare Connect and Birthday Week. Specifically, Matthew, I wanted to ask about NET Dollar. We have received many questions about this product. It could become a meaningful long-term growth driver. How should we think about the regulatory framework around it? And what has been maybe the early reception kind of out there?

And maybe along these lines, my follow-up will be maybe a word about AI gatekeeper. I know you started discussing it more vocally last quarter. Lots has changed over the past few months. You've indicated some initial activity, some contract wins around the guardrails from publishers and AI companies. So can you talk to us about what has changed in recent months? And is there anyone else out there emerging with a similar offering?

So let's start with NET Dollar. So as we have really interacted with AI companies, but also the merchants and media companies and the real long tail of the Internet, much of which sits behind us. What we realized was that as we move into a world of agentic commerce, we're going to need a currency to pay for the commerce that is done between agents that is really designed specifically for that task. And that's the spirit with which we started the NET Dollar project.

Now we're not -- we're unlikely to do it entirely ourselves for some of the regulatory reasons that you're familiar with, but there are lots of opportunities. And if you think about someone like Stephanie Cohen on our team, who is very familiar with the challenges of working in the financial services space, I think we're approaching it in a thoughtful way and are confident that we can execute in a way that is both going to help facilitate agent-to-agent commerce and be something that it fits well within any of the regulatory regimes that we have both in the U.S. and around the rest of the world.

At the same time, that is only one of our bets in this area. And I think a little bit the way that we're thinking about this is that we want to be the Babel fish of AI, sort of the universal translator, whether you're using MCP, the Anthropic protocol or Google's version of it or Microsoft's version of it, Cloudflare supports all of those. And so I think in addition to the excitement that we've seen around NET Dollar, I am equally excited about the partnerships that we're doing with Coinbase around X402, with Visa, Mastercard, American Express, around how you can create agent-to-agent payments. And I think that Cloudflare is a network, and what you want networks to be able to do is facilitate the ability for connection to happen and do it regardless of what makes sense.

So we think there are potentially some advantages to what we're building with NET Dollar, but we're not all in on any one of these things. We want to make sure that we support everything, and we can meet both customers and merchants and media companies, small, large, everything in between, wherever it is that they exist. And I think that's something that is unique about our approach. It's actually very similar to the answer to the previous question. We really do believe in multi-cloud and that we can be the facilitator of that.

We also believe that there are going to be multiple different ways to pay. There are going to be multiple different agentic protocols, and they are going to be hopefully many, many, many AI companies interacting with many media and businesses to create a more frictionless and AI-powered future of commerce. And I think that we see ourselves in the center of that.

In terms of sort of gatekeeper, so we have a product that's called AI Gateway. I don't think that's what you're asking about. I think you're asking about the product around us thinking about how do we help media companies figure out a new business model for the future. I think that, yes, I think that's going just extremely well. Like the number of media companies that are signed up and engaged is powerful. We're hearing from them about how the deals that they are able to do with AI companies have gotten markedly better, and we are getting a lot of praise for that. There will be others that compete with us in this space.

But I think one of the things that has really set us apart is -- and this is thanks to our over time, just significant investment in public policy and the side of the house that maybe doesn't always get as much attention. But I think we have been thought leaders in thinking about what does the future business model of the Internet look like. And that is getting us into a number of different conversations. And as we have done that, it's been clear that it's not just traditional media companies. But frankly, at banks, the research departments they're a little nervous because they're seeing ticks down in the amount of research that people are paying for because the AI companies are slowing that up.

So that's open conversations with financial services companies. We're seeing challenges with brands that are worried about what does a brand mean in the future of Agenta commerce. We're seeing challenges from small businesses. And I think one of the things that I am passionate about is how do we make sure that as this new paradigm, as this new platform emerges, how do we make sure that everybody has a fair shot to be able to participate in it. And so we will continue to do what we always have done, which is make our tools available to everyone, large and small.

And our next question comes from the line of Fatima Boolani.

Matthew, I wanted to ask you about the AI native ecosystem. It is embryonic, but on an absolute tear, there's so much capital flowing into the space, and you have taken a very active interest in bringing these AI natives on to the Workers and Workers AI platform. So what I wanted to ask you specifically was, can you help us think about the AI native exposure that you have today in the business? Anything that we should worry about from a concentration standpoint at this present time? And then maybe at a higher level, some of the engagement that you are seeing from a pool of funds perspective, how much of that is drawing in more AI-native eyeballs specifically because of the differentiation that you provide from an architectural standpoint at the edge for AI inferencing?

Yes. So I think that even though we're excited about AI and AI inference, it is still a relatively de minimis portion of our overall revenue, growing fast, but not -- I don't see any current concentration risk that's there. And what we're seeing is actually sometimes it's not the inference products that initially get interest from the AI native companies. It's actually the security products. And the reason why is the cost of AI, every query can be so high that making sure that you don't have fraudulent queries running through your system is critical in order to make sure that you can continue to operate cost effectively. And so many of the AI companies, we estimate that about 80% of AI companies use us in one way or another. But a lot of the times, that's using us for actually securing some of our -- really our Act 1 products. And then we are working on getting more and more of them to use the inference products as well.

In terms of what we can do that others can't do, I think you're absolutely right that being able to be close to users is important for a latency perspective. And that's -- and when you have human computer interaction, especially with something that is seems almost alive when you're interacting with it. Every millisecond counts because it breaks that illusion if things slow down, especially as you get to things like voice communication and other things that need to have kind of a natural rhythm to them. And so I think we're well positioned for that.

But in addition to that, because of the fact that we are taking the responsibility for driving utilization, and we're better at that than most customers are on their own, we can often, in addition to giving better performance, also give a lower cost of functioning. And again, I think that, that keeps us in a pretty healthy space. And so I have no doubt that there's going to be kind of ups and downs in AI over the coming months and years.

But it's clear to me that there is something very, very real here that it is going to be transformative that a lot of inference will run on your handset or your driverless car directly there, but that if it can't run there, it needs to run somewhere else, the next best place for it to run is in the network. And Cloudflare is the only network that gives you that capability on a global basis today. And I think that, that's going to continue to allow us to win workloads regardless of what happens to AI generally.

One comment I want to make, just to make sure we have no misunderstanding. When we say de minimis, we mean that no customer is bigger than 2% of revenue.

And our next question comes from the line of Mark Murphy with JPMorgan.

So Matthew, we noticed that Cloudflare is upgrading its security to be quantum safe so that data stays protected even when quantum computers eventually arrive, whenever that's going to be. I'm wondering if you can just describe the work you're doing? And do you think this is more of a long-term science project that won't matter for, say, 5 to 10 years? Or do you think it's something that could have some implications in the medium term? And then I have a quick follow-up.

Yes. I have sort of mixed feelings on quantum, where I think there's a sort of a lot of fear, uncertainty and doubt in the marketplace where people are going to say on quantum changes everything and it's going to be apocalyptic. That is not my opinion. I think quantum changes some interesting things. I think it's likely that you'll have more efficient package delivery that you'll be less likely to be delayed on your flight. And it does -- it will cause problems for some of the older generations of cryptography. But that's a very solvable problem.

And I think the thing that is unique about Cloudflare is that we have the scale on the content side to help figure out what the right solution is. And so we have partnered in the past and are continuing to partner with Google, who has scale on the eyeball side with Chrome browsers to be able to figure out what is a future-proof version of cryptography that will stand up even as we eventually have powerful quantum computers. And so we've worked together. We helped submit the data that went back to the Internet Engineering Task Force, the IETF and to NIST to standardize some of the new protocols that have been released.

We have rolled that out across our entire network for every customer, whether they pay us or not, because we believe that everybody should have the foundational levels of the best of security at no cost. And as you all upgrade your browsers on your phones and your laptops, all of them now are supporting post-quantum cryptography. The reason it's important to do now, even if we don't think that there are going to be quantum computers that can factor giant numbers, which is what you need to do in order for it to affect cryptography is the risk of storing the data. So if you just hoovered up a bunch of Internet data and then held on to it, you could, in the future, replay that and decrypt it.

And so for most of our customers, like it's no big deal. Like if your credit card number today gets compromised 10 years from now, it doesn't really matter because your credit card number has probably changed several times in that period of time. But for some of our customers, including the U.S. government cabinet level agency that we did a renewal with, this is incredibly important. And the reason it's important to do it broadly is you need to make sure that you're doing it in a way which is still fast, isn't burning a ton of battery life on phones. And what we can do in partnership with organizations like Google is actually roll out real-world tests and prove that it's possible and cost effective to do it. So I think -- I don't think science project is the right thing. I do think forward leaning is the right thing. And I think it is an example of how Cloudflare is always trying to live up to our mission of helping build a better Internet.

Yes, that is pretty fascinating. Just as a very quick follow-up. You mentioned egress fees I think last call and again, this call, I should say, the elimination of egress fees. It feels like you're winning some real business that's partly tied to that. Can you just touch on the economics that, that would unlock for the customer by removing those fees?

Yes. I mean, so egress fees are the cost that -- when we talk about them, the cost that hyperscalers charge you every time your data leaves their system. And hyperscalers have been notorious at keeping egress fees high. And it's actually one of the places where there's the most leverage because at scale, bandwidth becomes extremely inexpensive and really gets close to being free. And that's a longer conversation than we probably have time for but that's just the fact. And yet, even as the hyperscalers bandwidth costs have dropped and dropped and dropped, they've not passed those savings on to customers.

The reason they don't pass those savings on to customers is because they don't want the data to leave. They like to hoard all of a customer's data. And so what we believe is that it's the right thing to let customers take their data wherever. We believe that a multi-cloud universe is the right universe. And so products like R2, which is our object store, allow customers to take their data, their heavy object data that they usually have to pay a lot for if they have to move it around and move it onto our network where they can move it anywhere they need and be able to access it.

And so I think that what we're trying to do is say customers should be able to use whatever combination of clouds makes sense for them and that Cloudflare is the network that connects them all together and gives them the controls they need in order to do so safely, securely, efficiently, reliably and quickly.

And our next question comes from the line of Jackson Ader with KeyBanc Capital Markets.

The first one, Matthew, certainly losing CJ is a real loss, even if it is a great opportunity. And I'm just curious, whether you feel comfortable with the bench that he leaves behind and how you kind of ensure maybe that he either is or will be replaceable?

Yes. So first of all, I mean, CJ is terrific. And we can't talk about where he's going, obviously, but they are lucky to have him. And while we are bummed that he's leaving, I'm also really proud and excited for him. And this has definitely been a career goal of his for a long time. And so I'm glad that he's found his way to what I think is an outstanding technology company that he'll be leaving.

What's the good news for us is that when CJ came in, he actually didn't make many changes or even hire all that many people to Cloudflare. He looked around our engineering team and said, these are exceptional engineers. These are exceptional product leaders. What they need is someone who can come in and really focus them on being customer obsessed. And CJ has done a great job of implicating that customer obsession into our product and engineering team. And that's something that we're not going to forget.

But there isn't -- I don't think that there's a significant flight risk of people fleeing because they came for CJ and now CJ has gone. They came for Cloudflare. CJ was great as a member of Cloudflare. And I think that, that bench continues to be strong. And then both CJ as well as our team, now that we've got the news out of the way, I think that this is the most exciting job if you're a product or engineering leader anywhere in the world working in tech, you get to help build the future. You get to help invent what the business model of the future of the Internet is going to be. There is nothing that is more exciting than that.

And so I think we'll have -- CJ is one of a kind, but we will have no trouble finding someone else who is world-class. And I think the thing that will be the real legacy of CJ is that you will have taught us how important in product and engineering being customer-obsessed is and whoever comes next, I bet that's going to be something that you'll say, wow, that person has that quality as well.

Okay. That makes sense. And then a real quick follow-up. Matthew, you mentioned earlier the move from a product-led growth company to more of an enterprise company. The other side of that coin is now seasonality matters, right? We're heading into a fourth quarter. If you're selling more enterprises, that makes December all that much more important than prior December. So I'm curious about how the pipeline has built through the year and what you're kind of expecting as you're going into a more enterprise-ready fourth quarter than is typical for Cloudflare.

Let me get started, and Matthew can jump in. We gave the guidance we gave for the fourth quarter in light of what we are seeing. So we are very encouraged by the pipeline buildup for the fourth quarter, but guidance and foreshadowing what is going to come is more than just pipeline. We look at sales productivity. We look at the net sales capacity we add. We look at the motion and velocity in the developer business and all the indicators that we are currently seeing are reflected in how we guide for the quarter. So pipeline is encouraging, but there are more factors contributing to the picture we have of what is in front of us. And you heard from both of us that we are quite optimistic in how we look at the future.

Yes. And we've had -- we have lived with some level of seasonality for quite some time. You can see that historically, Q4 performs more than earlier quarters. But having gotten pipeline review meetings, I think we feel very good about where the coverage is for what we have in Q4. And again, it's just part of the journey of being more and more of a true enterprise sales company.

And our final question today comes from the line of Mike Cikos with Needham.

Matthew, first for you. I just wanted to see, can you please provide an update on the trends you're seeing in the SASE market specifically? Would just love to get an update on traction you're seeing out there as well as how the competitive landscape is changing, if at all, from where we were a couple of months ago.

Yes. I mean I think our SASE product when we're in consideration is performing extremely well. We don't see significant changes in the competitive landscape. We think we are very competitive. I think the biggest change for us has been just a kind of the proverbial forehead slapping moment of just looking at Netskope's S1 seeing that 95% of their sales are through channel, seeing the same thing for a while, we thought that, that was sort of an aberration for Zscaler and realizing the way that you sell these products is through partners.

And so we are doubling down on that. You see that we're seeing a significant uptick in the partner-led opportunities. We have always, I think, had good kind of willingness to partner, but we haven't always made it as easy as possible to partner with us. I think we're cleaning that up and doing a good job getting that in shape. And that, I think, will be the big unlock for those products to be able to be sold. And those are critically important products for us because their gross margins are so high.

And so as we're seeing strength in parts of our business like Workers, the best way to balance that out is also sell SASE as well because, again, it's something that has just extraordinarily high gross margins to it. Whereas products like Workers, the gross margins are good and will get better and better over time. But because they are newer products, they're not as optimized in that space. So I think this is a really opportune time, and I'm excited about sort of the partner-first strategy, especially with our SASE products.

And with that, I will now turn the call back over to CEO and Co-Founder, Matthew Prince, for closing comments. Matthew?

I just want to thank our entire team for an incredible quarter. It takes a ton of effort of people executing, of people innovating and us being able to deliver results like this. Thank you so much and we'll see you back here again next quarter.

Thanks, Matthew. And that concludes today's conference call. You may now disconnect. Have a great day, everyone.

All right. We are going to go ahead and kick off the Cloudflare session at day 2, Goldman Sachs Communacopia and Technology Conference. Thanks, everyone, for joining us. And thank you to Thomas, CFO. It's very good to have you back.

Pleasure being here.

So Thomas, I know from the outside, we can see the metrics this year. And what's really interesting about this year is that you're actually poised to accelerate as we go through the year, probably consistent with what you said a year ago. And so putting some of the sales productivity conversation to the site just for now. What stands out to you about this year? What are your observations on what's driving or what's going better and has more momentum this year versus last year?

I think what has always made this business model at Cloudflare special was that growth was driven by many vectors, many independent vectors. And I think we see a lot of goodness coming together in the first half of this year that is driving that. So if you take aside the sales productivity gains, the contributions from adding net sales capacity, it fall in a couple of buckets. We see extraordinary strength in our largest customer cohort. So customers north of $1 million and north of $5 million of ACV are growing the fastest.

We see strength across pretty much the complete product portfolio. Well, I'm sure we will talk about Act 3, Act 2, our Zero Trust, our Cloudflare One products are ramping very, very nicely. And we see a reemergence, so to speak, of Act 1, mainly driven by all the work and announcements we have been doing around what we call Act 4, pay per crawl, I'm sure we'll come to that. So it's a broad acceleration in -- across many different vectors that is making that happen. And we are, of course, pleased by the strength we see in the first half, and we guided for continued growth in the second half.

Yes, fantastic. Well, let's jump into Act 3 because there are so many products we can talk about. So Cloudflare talked about how your share within AI-native companies is higher than your share across the ecosystem as a whole. Maybe share with us your observations on that AI native cohort. How are you seeing your footprint trend within them? And how does that cohort compare to your classic cohort?

I think you have to spend some time talking about Workers first. So the idea was to give a product to developers that would be without lots of competition in terms of performance, in terms of latencies, in terms of ability to scale. And it is the basis of everything that Cloudflare is built on. Workers enables us to scale and have the pace of innovation we have. So having now a product platform in place that is rather complete as a platform. It offers unique interfaces. It has storage capability, it has database, it has CPU, it has GPU, enables now a world of -- a class of developers to develop products and us that is quite unique. Our developer count now is north of 3 million developers. It has been accelerating since the second half of last year based on all the AI work.

More interestingly enough, we see we are now at a point where because of the latency and the performance differential in the platform, you see non-native -- non-digital native companies build on us. We just -- the other day within the company talked about an industrial customer that is building an agentic AI workflow on Cloudflare in a vertical that you would not necessarily associate with us first. But it's a highly latency-sensitive application, and we are built for that.

So while the initial traction was very much digital native, we see now based on the completeness of the platform in terms of features and performance that it expands into our non-digital native customers. And we think there's a lot more to come if you think about agentic AI workflows in the future, that will be how seamlessly and how latency -- with low latency, you can interact with those agentic AI, so will be key, and we are built for that.

Maybe on this point on features and functionality in Workers, bring us up to speed on how you think about competition because Matthew has consistently been cognizant of the hyperscalers as a competitive cohort. But what about some of the newer companies like Vercel or some of what we're seeing in developer tools with prototyping? How do you think about your structural differentiation versus some of the newer players?

When investors start to work with us and then get the first time to take a closer look at us, I would say you have to start at the architecture of the network. This is the true competitive mode of Cloudflare, this idea of off-the-shelf hardware and the software stack that allows every product on every server and every location completely agnostic to the hardware it runs on. This is what makes this differentiation.

And this is a global network. So every code you deploy as a developer, as a customer is instantly available all around the globe. So it's not a geographic architecture like some of the hyperscalers. It is because of how it consolidates so much demand from a bandwidth and traffic perspective, more distributed than anybody else. So in a world where latency and performance becomes key, where we all cannot change the speed of light, so how close you get to the eyeballs is going to be key, this network wins.

And I think we see that differentiation play out across all verticals against the hyperscalers, but especially against -- it's hard to build a network like this and be competitive from a latency performance and the reach perspective.

It's interesting that you brought up an industrial company as being a new customer in Workers. I know that one of your areas of focus with Aly Cabral has been how do you expand the developer motion top down and bottom up in some of the enterprise customers that you have today who may already have existing tools in their stack. Maybe give us a little bit of an update on how that effort is going and how you feel about your traction displacing or augmenting some of the existing developer stack?

It's going really well. And if there is -- if you want to nag on us internally, why didn't we come up with the speedboat and the idea earlier, but it's not only about the idea, the product needs to be ready for it and the market needs to be ready for it. So Aly runs the speedboat. She was -- she switched site. She was -- she developed workers on the engineering product side before and now runs what we call a speedboat that is literally looking -- they have a list of ideal customers and applications that can be targeted with Workers and Workers AI applications.

And we learned from our large deal that we announced in the first quarter that the best way to win a $100 million customer is to help the customer save $100 million. And because of the architecture of the network, we can do this with significant gross margin. And there are a lot of applications out there of this size where you can help customers save significant spend or significant spend reduction on their hyperscaler bills and still do great things. So they are a dedicated effort. It's a self-contained team, independent of the rest of the go-to-market motion, and they are just working on their target list of customers.

And they're making good progress. It's super interesting discussions because we learn a lot also about features that are need to be built out in order to enable that. Lifting an application with this spend is not easy. So how do you enable the customer, how do you take risk out of the migration is key, but good progress. It doesn't mean we will announce 9-figure deal every quarter, but there are a lot of these deals out there.

Well, I won't hold you to announcing a 9-figure deal every quarter, but maybe give us some qualitative commentary on the pipeline. I think there was a comment earlier...

It's a longer list of targets that you would think, and they're chipping away on it. I don't want to give guidance here, but it's a huge significant opportunity set.

There is a nuance here where we've spent time at the Analyst Day, for example, understanding why Cloudflare Workers is better for inference workloads in particular. But it sounds like you're also describing lift and shift for other types of workloads. So what are the types of workloads where you're able to realize this sort of order of magnitude...

There -- the $100 million deal is a good example. It wasn't just Workers AI. It was plain vanilla Workers, where there are many apps out there that can be targeted. It was surprising to us to see how big that universe is. And your persona changes a bit. If you enable a customer or your business partner to go to his or her CFO to save $100 million, that is a lot easier discussion getting big deals through a pipeline. So quite encouraging what we learned. And as I said, really excited about the opportunity set that is there for us.

So part and parcel with the demand signals that you're talking about on the Workers' side comes the CapEx build? So we're seeing a step-up in CapEx this year in the order of magnitude of about 50%. And you've talked about how you wouldn't make that kind of investment if you didn't have the demand signals. What are the demand signals telling you about the durability of that CapEx increase? Should we be thinking about a similar size step-up for next year or help guide our expectations?

So it's a step-up in terms of dollars. It's -- we are still within the CapEx ratio that we guided to. So we talk about 12% to 14% this year, 12% to 13% of CapEx in relation to revenue. We feel quite confident about this. The levers we have to model what we need are quite significant. It's -- Cloudflare has always been a business where we invest behind the demand signal. We never invest ahead of it. It's a playbook that served us really well on the CPU side. We, on average, run slightly north of 80% CPU utilization. And when we started Cloudflare, it was pretty much an Intel-only CPU environment, and then it became flavors of Intel CPUs and then became AMD CPUs and they become ARM CPUs. And today, that software stack that runs on it is completely agnostic to the hardware that it is running on.

We deploy the same playbook on the GPU side. We started off with one flavor and one company, and now it's diversifying into many flavors from one company, and now it's diversifying into many flavors for many GPU suppliers. Our big levers are getting the hardware stack right. Inference task is not like inference tasks. The need for what GPU compute capability you need differs. So you have to optimize your hardware stack that is allowing us to be efficient. Getting utilization up is really important. We are today better than anybody in the world when it comes to GPU utilization because the architecture, how we run our workloads on isolates, no wait time, no spin-up time is super-efficient. But we still have tens of percentage points of room to improve to our CPU performance. So that is impacting it.

And then there is how much capacity do we deploy? And what helps us on the CPU side is what we do on the GPU side. We have a baseline of capacity that we feel comfortable with what we understand in terms of customer commitment. And then we have certain arrangements in place that allow us to outsource spikes in GPU capacity to third-party vendors. That might -- it's CapEx efficient. It's a bit more expensive. But as soon as we get more comfortable with our baseline, we increase that and we in-source. But that means always we are, to a certain degree, under provisioned in terms of the capacity we have visibility to. And that allows us to run our CapEx investment highly efficient. We think there's plenty of room to stay in this range.

The other thing that helps us is you really have to understand how the network works. So we -- Cloudflare is built on Act 1 products, product like firewall routing, DDoS mitigation, where you move traffic to the eyeballs. Our second generation of products, what we call Act 2 is all the Zero Trust and Cloudflare One products. They pretty much transport data back from the eyeballs to wherever the data center is. And they take advantage of an infrastructure that is already paid for by the other products. So every dollar -- revenue dollar in Zero Trust is literally at margins that are north of 90% for us, and it literally comes at 0 CapEx. So every dollar we have opens up allocation degrees of freedom for us to move CapEx spend to GPUs. So many vectors that you can influence many that you can work on, and that makes us highly comfortable that we are going to stay CapEx efficient for a while.

Yes. There's a question buried in here on unit economics a little bit because I think there are some question marks for the industry as a whole as to what the unit economics look like on GPU heavy investments. So I guess 2 parts for you. It sounds like you're already comfortable that you're benchmarking better than some of your peers on GPU utilization. Maybe just share with us how you think about that benchmarking. And then the second part is, how would you compare and contrast the unit economics of the GPU-based workloads versus the CPU-based workloads?

So on the infrastructure layer, margins are lower than on the application layer for sure. But as we have seen in the Googles and Amazons of this world, there are huge benefits from an economies of scale perspective when it comes to leveraging OpEx, when it comes to leveraging R&D spend and others. So we think there's plenty of room for us to stay in the band of unit economics we need.

And the second is we are not a point solution provider. We play across a wide variety of products. Today, 64, I think, is the latest count of revenue-generating products with very different unit economics and especially the unit economics on our strongest growing product is providing us all the degrees of freedom we need in order to lean in when it comes to Act 3 and Act 4. That's why despite all the growth you have seen over the last years, all the product launches, the different apps, the explosion in traffic there, the gross margin has been rather stable within 2 or 3 percentage points over the last 9 years.

Absolutely. Well, let's talk about Act 4. The industry moves so quickly when it comes to working through some of these new developments. So tell us a little bit about your observations on how the conversations are going between the independent publishing sites and the LLMs.

Yes. Maybe just for those who are not that familiar with all the Acts we have. Act 4 is a product that we launched in June of this year. It's paid to crawl. It's pretty much a tool that allows the content generators to decide who is allowed to call their websites and under what conditions, for free, paid, paid for how long. And we think this is in a world where the economic structure of the Internet is changing from a click and link-based model to where pages are crawled today for free to a large extent, allows us to help level the playing field. So we announced this in June, all the major publishers around the globe are signing up for it or have signed up as customers. Everybody deals with the product in a different way, but it gives them as an individual company or as a group of companies, a higher bargaining power when it comes to how you monetize content. So that part of the equation is going really well.

In parallel to that, we, of course, have discussions with the other side of that equation, that is the companies, especially the large language models. And I think they also understand that their ability to train to a significant extent depends on high-value content. And I think they're ready to monetize for that. And we've seen announcements just in the last couple of days of payments that are made and compensation that is made. They are just nervous about having a level playing field on their side, too, and not one LLM or one company having an unfair or having an advantage at all.

So I think conversations on both sides are progressing. They are progressing really well on the content creator side. And we are making good progress. There is a lot of work that still needs to be done to figure that out, lots of interest that need to get aligned. But we think after the first couple of months, the interest it has generated, the momentum we see customers coming to us, and we should talk about that in a minute, shows us that this business model has legs. There are a lot of things that need to get figured out in terms of economics and transaction models, but this business model will have legs.

Maybe let's stay on the business model for a moment. So I fully appreciate that. We're very early in this discussion. Matthew made a comment, I think it was about a month ago on $10 billion being the opportunity associated with, I believe it was ad revenue tied to independent publishing websites. Maybe give us your perspective on how you think about that, and what...

Yes, I'm not sure -- so I didn't hear the $10 billion from him. I mean, if you look at their digital ad volume in the U.S. alone, that is $300 billion a year, right? So the free web is anywhere $30 billion to $50 billion just in the U.S. on that part. So that somehow needs to be replaced or moved from one bucket to the other. So it's a big opportunity. If you just look at the amount of fees that Spotify distributes for music, I think there's a lot of dollars on the table, so to speak, and we'll figure out a business model that takes that into account. It's not an easy feat, but it has legs. That is what we have seen.

We have seen so much momentum on publishers normally and content creators was not normally a vertical that was strong on us, right? We never considered us a CDN that moved just data -- a byte of data from point A to point B. This allows us to generate interest in a vertical from a completely different perspective on the fat tail of that distribution, big publishers and content providers.

But the interesting part for me as a CFO is Cloudflare came into existence consolidating the long tail of a market where people just didn't have $500,000 to spend on a firewall, right? This is how we consolidated significant traffic on our network. We have an opportunity to do this again in the long tail of the market with content providers and diversified content, high-quality content is an important part of us training and developing LLMs further. So we think we have a very unique opportunity that is -- that comes again to us repeating what we've done in Act 1.

It's actually really interesting because when we started this conversation, I think you alluded to Act 4 impacting Act 1. So are you already seeing a change in your win rate for classic CDN, WAF, DDoS?

We have significant shifts from very large legacy installations that move to us in non-digital native companies. There are large shifts underway, especially in Europe, some big wins that we also talked about in the earnings call. That is one. We talked -- if you listen carefully to our earnings call, we talked about gross margin being impacted by shifts in free traffic to paid traffic. That is one topic we see where customers that are free customers on our website that might have a super interesting, I don't know, travel blog post on Africa, all of a sudden wanting to become a paid customer, so they have an opportunity to monetize content.

So you see already shifts in free-to-paid traffic. That reminded us again how important free customers are for our started initiative going into the year with a dedicated team focusing on free customers was a big part of our Act 1 a long time ago. So yes, Act 4 is impacting Act 1 and not only in one part of the market, but across the whole range from free to established Fortune 10 companies.

Yes, really good color there. Okay. So I wanted to ask you one question about SASE. And my favorite data point from 2Q was actually about the customer that essentially said no to Cloudflare 18 months ago and then came back, reevaluated the portfolio and you won that customer. So maybe level set us, how do you feel about being best-in-breed in SASE today? And are you seeing more of those types of deals that perhaps wouldn't have been in your addressable market 18 months ago?

Yes, very much. So you have to keep in mind how we started to build what we call our Cloudflare One Zero Trust product from the ground up, targeting the small and medium-sized customers and then trying to feature complete and go up into the right on a Gartner Quadrant over time. We are at a point now where we have feature parity across most of the SASE components. There are 1 or 2 where there is still room for improvement, especially when it comes to DLP, but in general, you are there. So you can address also large legacy installations. So that is very helpful.

We have seen significant traction in the financial service and financial vertical over the last 2 quarters because of that. You know this vertical as good as I do. There are some marquee banks. And if you win them, you get approval across the whole stack, and we've been really successful. This is also to a certain extent, the success of CJ joining us. There are customers who have been successful betting on him in the past. And even if a feature is missing, they will understand if he commits to it, that it will come. So we have made significant progress.

What is helping us also in a world of agentic AI that even on the Zero Trust side, latency is becoming a significant decision driver. That is where we excel. It's performance, latency and ease of installment. So the customer you mentioned that came back after 18 months, what we saw was that the competitor was only 30% rolled out. And it was a result of hard to install, hard to get to performance and really bad latency compared to our product. So this coming together now sees momentum in really difficult verticals like critical infrastructure and financial services. So we are quite pleased with the progress we've made.

So I think throughout this conversation, we've been pairing the product excellence with some of the changes you've been making in go-to-market. And what I found really interesting coming into this year was you had already succeeded in making a lot of the block and tackle changes that you wanted to make under Mark Anderson, such that you had a higher number of ramped sales executives at the enterprise level coming into this year. You've also talked about increasing hiring this year now that you're where you want to be from a productivity per rep standpoint. That should set you up really nicely to have more ramped sales reps next year, which then increases the amount of opportunity you can go after. Maybe just level set us how should we think about ramped sales reps impacting reacceleration and the pace of Cloudflare.

They are an important part of that equation. When we went into the sales transformation more than 2 years ago now, we set ourselves certain goals that we wanted to see. This idea at Cloudflare that you never invest ahead of demand is true for everything we do. So we wanted to see certain progress on productivity levers for reps before we get comfortable to open the hiring struggle. And we talked about it on our earnings calls on the third and fourth quarter, we've got comfortable that the productivity was where it needed to be. There's still room for improvement, but they're diminishing returns now in terms of the levels we have achieved. That allowed us to get comfortable to increase hiring.

And we said that last year, much of the sales growth was productivity driven. This year, it will switch from productivity-driven sales growth to net sales at capacity. And this is what we see, what we've seen a bit in the first half. We'll see a lot more in the second half. And then every sales rep you add has a ramp-up period, right? So until they come to full productivity. So that impact will spill over into next year, too.

And I'm curious to what extent, I know how thoughtfully you think about AI internal tools for your organization. To what extent are you seeing AI accelerate the productivity ramp for your sales reps?

So we were quite careful in the beginning, how we allow people to interact, what LLMs we use, where the data needs to reside. The first lessons learned were that every time we opened for a productivity tool, the results were significantly more impactful than we thought. So we have a broad initiative across the company. We have a dedicated VP who does even tutorials for every employee on how to use AI to automate his or her workflow. So it's not only limited to go-to-market.

We have significant tools now deployed that help us from a productivity perspective in terms of first contact, reducing the numbers of PDRs you need to support an AE. One of my biggest challenges is building a budget for next year that incorporates it all, right? So where can we -- where can -- and I'm sure I'm not alone, I must be -- every other CFO must think about that, that the benchmarks of yesterday are going to be disrupted. What used to be, I don't know, 8% of revenue for G&A might be half of it moving forward, and we want to be ahead of that game. So we have a vast deployment.

And it's not so much that we want to save the money, we want to reallocate it. Our biggest engine is our innovation engine. So every dollar we can put into more products and more innovation is going to be helpful. And our -- we are way ahead in terms of experimenting with it. It's a broad-based effort, and it's not only limited to go-to-market at this point.

Perfect. I want to close here on pool of funds. I also want to -- share with us your observations on how the consumption patterns within the pool of funds deals are tracking relative to your originals.

Yes. So maybe for the audience, pool of funds was a concept that we started to deploy last year because the amount of products we have now 64, as I said, is harder to sell. So how do you get customers comfortable and how do you find a way to make this process of consuming as much as possible from the platform and as fast as possible as frictionless as possible. So pool of funds for us is we negotiate with the customer contract, and it has a rate cut literally for every product we have. And there's a commitment to consume, I make this up now, $100 million over 5 years on a rate card that could mean any product.

So when we started this, we -- it was quite successful. But that also means that all of a sudden, every salesperson in the company wants to a pool of funds, and it's a complicated tool. It's not good for every situation. So we -- there was some learning we had to do, but we've become better at it.

And the second part was once you have a customer sign up for it, you have some conversations, they tell you where they want to start, but there's a lot of variability in this process. So we had no idea how fast they would ramp, how fast they would consume. And there was an uncertainty in the forecasting process because of that.

Today, I think that noise is lifting and is disappearing. If I were to click on my phone here, I could see across all pool of funds where we are at this moment in time in terms of consumption, who is ahead, who is behind. You would see that on average, across all of pool funds deals, we are pretty much on target, yes, and where we need. That's hard work. But it helps us, I think, to be more precise in how we look at the business moving forward, and it's helpful that the noise is disappearing. So good progress.

Very good. Well, thank you so much for all your comments. Please join me in thanking Thomas for his time.

Thank you.