Microsoft Stock

Greetings, and welcome to the Microsoft Fiscal Year 2027 First Quarter Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded. It is now my pleasure to introduce Jonathan Neilson, Vice President of Investor Relations.

Good afternoon, and thank you for joining us today. On the call with me are Satya Nadella, Chairman and Chief Executive Officer; Amy Hood, Chief Financial Officer; Alice Jolla, Chief Accounting Officer; and Brian DeFoe, Deputy General Counsel and Corporate Secretary.

On the Microsoft Investor Relations website, you can find our earnings press release and financial summary slide deck, which is intended to supplement our prepared remarks during today's call and provides the reconciliation of differences between GAAP and non-GAAP financial measures. More detailed outlook slides will be available on the Microsoft Investor Relations website when we provide outlook commentary on today's call.

On this call, we will discuss certain non-GAAP items. The non-GAAP financial measures provided should not be considered as a substitute for or superior to the measures of financial performance prepared in accordance with GAAP. They are included as additional clarifying items to aid investors in further understanding the company's first quarter performance in addition to the impact these items and events have on the financial results. All growth comparisons we make on the call today relate to the corresponding period of last year, unless otherwise noted. We will also provide growth rates in constant currency when available as a framework for assessing how our underlying business performed, excluding the effect of foreign currency rate fluctuations. Where growth rates are the same in constant currency, we will refer to the growth rate only.

We will post our prepared remarks to our website immediately following the call until the complete transcript is available. Today's call is being webcast live and recorded. If you ask a question, it will be included in our live transmission, in the transcript and in any future use of the recording. You can replay the call and view the transcript on the Microsoft Investor Relations website.

During this call, we will be making forward-looking statements, which are predictions, projections or other statements about future events. These statements are based on current expectations and assumptions that are subject to risks and uncertainties. Actual results could materially differ because of factors discussed in today's earnings press release, in the comments made during this conference call and in the Risk Factors section of our Form 10-K, Forms 10-Q and other reports and filings with the Securities and Exchange Commission. We do not undertake any duty to update any forward-looking statements.

And with that, I'll turn the call over to Satya.

Thank you very much, Jonathan. It was a fast start to our fiscal year. Microsoft Cloud revenue surpassed $63 billion with growth accelerating to 30% year-over-year. We are focused on 3 things: designing AI to empower people, ensuring every organization has sovereignty over its own learning and building trust with controls that ensure agent safety and security.

Now let's talk about how we are making progress on those goals as we build the new tech stack, starting with Copilot. We announced our biggest update to Copilot to date this quarter, bringing together Chat, Cowork, Autopilot and code into one unified experience. It's one product experience across commercial and consumer and comes with all of Office built in. We think of Copilot as the new OS for work. Customers are increasingly turning to Copilot to get real work done. Automated tasks has nearly doubled since July. Daily users are up nearly 50% quarter-over-quarter and over 5x year-over-year. User satisfaction improved nearly 20% quarter-over-quarter to another all-time high. And 90% of Copilot queries use Auto Mode, which dynamically routes tasks across models optimizing for quality, speed and cost. These improvements are translating into larger purchases.

The number of customers with more than 50,000 Copilot seats increased over 8x year-over-year, including Capgemini, HCL Tech, Hitachi Group, Truist Financial and U.K. Revenue and Customs Department. And just last week, Volkswagen announced plans to deploy Copilot to more than 330,000 employees, one of the largest deployments to date in Europe. We're also seeing strong momentum with E7, which brings together Copilot, E5, Entra and Agent365. Thousands of customers in every industry have purchased the suite to date, including Amadeus, BlackRock, Fiserv and New South Wales Health Department and ZEISS. Our business model is expanding from per seat subscriptions to include consumption, expanding our TAM and ARPU. The subscription brings increasingly capable everyday intelligence to every user.

Typically, IT funds this, consolidating frontier advances for everyone across the organization. And as agents take on higher-value work, consumption opens up business funded demand with spending tied directly to the value these agents deliver. The number of Copilot customers actively consuming credits through usage-based billing grew over 75% over the quarter. When it comes to software development, all major coding agents have shown up to work on GitHub. Agent-generated PRs have increased 6x over the past 6 months and agents now open more than 1/3 of all PRs on the platform. We are very hard at work on reliability as agentic activity on GitHub sores. At GitHub Universe this morning, we shared how we are expanding capacity and rebuilding the core Git architecture for the Agentic wave.

Also at Universe, we announced broad availability of Copilot Fusion, which is all about bringing multi-model orchestration to software development. It delivers frontier level quality up to 40% faster and at up to 77% lower cost. Credit consumption from agentic features on GitHub has more than doubled since usage-based billing went into effect. And we're also helping developers secure their code and turn AI-generated code into high-quality software. GitHub Advanced Security revenue increased over 50% year-over-year as customers use its AI capabilities to scan entire repos and fix vulnerabilities in real time.

In the business application space, we are extending this vision of Copilot as the new OS for work, bringing the business logic and capabilities of Dynamics 365 directly into Copilot and helping every organization stand up their own infinite SaaS factory. This quarter, we introduced 30 new Copilot skills across Dynamics 365 which enables customers to access CRM and ERP capabilities directly in the flow of work. In recruiting, more than 10 million LinkedIn members use our LinkedIn AI-powered job search daily and 25,000 customers are using our agentic hiring solutions. Recruiters using hiring Assistant are finding interview quality applicants 33% faster while reviewing 83% fewer profiles to find qualified match. In health care, we automated 30 million patient encounters this quarter, up 70% year-over-year, helping clinicians spend less time on administrative work and more time with patients.

10 providers, including Advocate Health, Mass General Brigham, Mercy, Intermountain Health and others have used our solutions to document over 1 million encounters each since the beginning of the calendar year. And in science, our FE teams are working with customers to use discovery to accelerate scientific progress. A great example of Saint-Gobain, which used Discovery to accelerate the development of low-carbon construction materials. Underneath our first-party agentic experience is a common platform. It starts with foundry, which is becoming an increasingly important driver of our growth.

Foundry helps organizations build learning loops grounded in their business context and deploy agents safely at scale. We now have over 120,000 foundry customers with revenue more than tripling year-over-year. Foundry's model choice and feature set, coupled with its deep integration across our stack is driving differentiation. Azure OpenAI directly benefited in the last quarter with big advances in OpenAI models. And when you couple this with our model choice built into foundry, customers are able to build solutions optimizing across quality, cost and performance. In fact, the number of customers building agentic solutions with multiple models increased 5x year-over-year and revenue from open source models on foundry increased 10x.

We also continue to make rapid progress with our first-party models. Our goal is to advance the cost-to- outcome frontier, which means delivering capabilities comparable to leading models within 3 to 6 months of their release at a fraction of the training cost, thereby allowing us to transfer those cost benefits to customers. And you see this in our products today. MAI Cyber now powers over 40% of the scans in our Agentic M-Code scanner at 18% lower cost per scan. MAI Transcribe is the most accurate and cheapest transcription model in the world, supporting Dragon Copilots multilingual workflow. MAI Code Flash has 3 million users in GitHub Copilot last month alone, helping developers get more out of their AI credits.

Microsoft Decision One is our new model for structured decision-making, outperforming both LLMs and other decision models in latency and quality. Over 4,000 customers, including Atos Group, Equinor and RBI have already used it in the first month since its release. And we recently completed pretraining MAIbase 2, which outperforms other Open weight models we tested with 4 to 12x greater training compute efficiency. This base model will be used in the upcoming MAI models across thinking, code and cyber. Beyond models, we are making the harness behind our own Copilot experience available to customers in foundry. It gives customers a Microsoft managed way to build agents that reason, use tools, complete complex long-running work while preserving their choice of model. Another critical layer is data. We now have 44,000 paid fabric customers, up more than 50% year-over-year. We are bringing Foundry and Fabric closer together so you can build your own agents right next to data.

Over 19,000 customers now use Foundry and Fabric as enterprises connect and operate agents over real-time operational, analytical and unstructured data. Customers are also rapidly deploying our AI optimized databases, including PostgreSQL, SQL DB, Hyperscale, Cosmos DB to give agents fast, secure access to the data and context they need. PostgreSQL growth accelerated for the fourth consecutive quarter, and the number of PostgreSQL customers also using Foundry accelerated to 90% year-over-year. We also have one of the most important databases for any enterprise, which is the knowledge underneath Microsoft 365 spanning people, their relationships, documents, e-mails, meetings, chats, projects and other artifacts. Work IQ makes this context available to every Agentic workload.

And with Microsoft IQ, we are bringing together Work IQ, Fabric IQ and Foundry IQ as a unified context for Agentic systems. Nearly 60,000 customers are accessing context across Microsoft IQ, up 25% quarter-over-quarter. As enterprises deploy more agents, they also need to govern them at scale. With Agent 365, we provide a platform for AI observability, enabling customers to manage, secure and optimize the performance and cost of their agents. Over 50 million agents are now in the registry, up over 30% quarter-over-quarter. And finally, as agents gain greater access to enterprise data and systems, security becomes even more critical. Our approach is to separate intelligence from authority, enforcing identity, permissions, monitoring and containment outside the model itself.

Last week, we expanded perception with new capabilities to help customers secure and contain agentic insiders, including expanded agent discovery across local and cloud environments, dynamic permissions that enforce least privileged access for each task and continuous risk assessment. We also introduced Teams of Guardian agents that continuously monitor agent activity to help prevent unsafe actions before they occur. And we are seeing encouraging feedback on perception overall from preview customers, including KPMG, Nationwide Building Society, Northern Trust, Stellantis.

More broadly, we are helping customers secure and govern AI use across the enterprise. Customers have used Microsoft Purview to audit over 7.6 billion agent activities to date, up 5x quarter-over-quarter. Underlying everything I've talked about so far is our infrastructure. We are building against a growing book of business with a fungible fleet that allows us to dynamically allocate capacity to the highest value opportunities across compute data, foundry and Copilot. Accelerating demand led to us adding 24 new data centers across 5 continents this quarter. And we are bringing capacity online faster, reducing doctor lifetimes by about 20% quarter-over-quarter for new GB300s in our biggest regions, which make up the majority of the GPU deployments.

At the same time, we continue to get more from the infrastructure we already have by optimizing across the data center design, silicon systems, software and models. For example, we increased the throughput for Copilot workloads by over 50% quarter-over-quarter. We continue to scale our fleet with the latest innovations from NVIDIA AMD alongside our own Maia systems. We have docked our first NVIDIA Vera Rubin GPUs and are very excited about it. Maia 200 is already delivering up to 1.7x more tokens per watt across the AI experiences in Teams and Copilot, and Maia 300 will drive further gains, delivering 40% more tokens per watt compared to the leading GPU-based systems. Our growth extends beyond GPUs. We continue to see strong adoption of our cloud-native cobalt CPUs, improving price performance for general-purpose cloud workloads. All up, Azure growth was broad-based across both enterprises and small and medium business segments and spanning both core cloud services and AI infrastructure.

Finally, let me turn to devices and consumer. In Windows, we are executing on our vision to bring hybrid intelligence to every desk and every home, making every PC the secure agent box. As more capable models run locally, we see an opportunity to make more intelligence effectively unmetered. We are working with the broader hardware ecosystem to bring these capabilities to a new generation of PCs across AMD's Ryzen AI 400 Series, Intel Core Ultra Series 3, Qualcomm Snapdragon X2 series. And earlier this month, we also announced new PCs powered by NVIDIA's RTX Spark and DGX Station for Windows.

With these new devices, you can run last year's frontier models on your laptop and last quarter's model at your desk. We are bringing together Windows and Agent 365, so agents can work with secure boundaries on device. We introduced Microsoft Execution Containers, a policy-driven execution layer that governs and contains agents across Windows and other platforms integrated with Agent 365 for governance and observability. Agents like OpenAI [indiscernible], Anthropic Cloud Code, Meta Muse are adopting MXC and customers like Meta are starting to deploy it across their enterprise environments. In search and advertising, Bing and Edge once again took share.

In LinkedIn, engagement continues to deepen across our more than 1.4 billion members with time spent viewing content in the feed up 15% year-over-year. And finally, in gaming, we remain focused on building great games and reaching players wherever they choose to play across console, PC, mobile and cloud. Minecraft delivered a record quarter, reaching all-time highs in both engagement and revenue, while World of Warcraft had its largest beta in franchise history. We entered the holiday season with one of our strongest first- and third-party lineups in years, including Modern Warfare 4, which launched last week to the best reviews for a Call of Duty game in more than a decade; and Grand Theft Auto 6, which is coming to Xbox next month.

In closing, based on the demand we see, the strength of our products and the efficiency we are driving across the stack, I've never been more confident in the long-term returns for our investments. With that, let me turn it over to Amy.

Thank you, Satya, and good afternoon, everyone. Our results exceeded expectations across revenue, operating income and earnings per share, driven by strong demand across the Microsoft Cloud. This quarter, revenue was $92.3 billion, up 19%. Gross margin dollars increased 17% and gross margin percentage was 68%. Operating income increased 21% and 22% in constant currency. Earnings per share was $4.89, an increase of 18% and 19% in constant currency when adjusted for the impact for our investment in OpenAI. And FX was roughly in line with guidance. Operating expenses increased 6%, driven by continued investments in R&D compute capacity and talent. Operating margins increased slightly year-over-year to 50%, driven by strong revenue growth, platform efficiency gains and focused investments across our portfolio, partially offset by sales mix shift to Azure.

When adjusted for the impact for our investments in OpenAI, other income and expense was negative $234 million, slightly lower than expected. Capital expenditures were $45.2 billion as some third-party capacity deliveries we expected in Q1 were delayed. Roughly 3/4 of our CapEx was for short-lived assets, primarily CPUs and GPUs, including higher memory and storage costs as we continue to invest in the infrastructure required for our customers to build and scale AI applications. The remaining spend was for long-lived assets. This quarter, finance leases were $7.4 billion. As a reminder, we recognize the full principal value of finance leases and CapEx at the time of lease commencement. And cash paid for PP&E was $39.4 billion. Cash flow from operations was $58 billion, up 29%, driven by strong cloud billings and collections.

And free cash flow was $18.6 billion, reflecting higher capital expenditures. And finally, we returned $10.2 billion to shareholders through dividends and share repurchases. Now to our segment results reported under the updated segment structure announced in September. Revenue from Agents and Infra was $76.8 billion and grew 25%, reflecting strength across the connected layers of our platform. Commercial bookings decreased 12% and 11% in constant currency, driven by the impact from large Azure commitments in the prior year, partially offset by healthy growth in our commercial sales motions, including a large Azure commitment this quarter.

Commercial remaining performance obligation grew 73% to $677 million, with a smaller sequential decline relative to historical seasonality due to the large Azure commitment this quarter. Our RPO has a weighted average duration of 2.3 years and roughly 30% will be recognized in revenue in the next 12 months, up 29% year-over-year. The remaining portion recognized beyond the next 12 months increased 98%. Now to Microsoft Cloud. Revenue was $63.9 billion and growth accelerated to 30% and 31% in constant currency with sequential dollars added nearly doubling year-over-year, reflecting strong demand across Azure and M365 Commercial Cloud.

Azure revenue was ahead of expectations and increased 48%, even with the 3P capacity delays noted earlier. driven by foundry and improved CPU monetization. Foundry revenue more than tripled year-over-year with upside driven by efficiency gains across our GPU fleet that allowed us to allocate additional capacity to customers as well as higher revenue per token that we benefited from in the quarter. Across our CPU fleet, engineering work enabled us to identify and monetize underutilized capacity this quarter, helping us serve more of the growing demand with our existing capacity. The timing of engineering efficiencies like these as well as 3P GPU lease deliveries, our own capacity build-outs and ARPU changes from model mix can create quarterly variability in the year-over-year growth rate.

M365 Commercial Cloud revenue increased 18% and 19% in constant currency when adjusted for the prior year comparable that benefited from 1 point of in-period revenue recognition. On a reported basis, revenue growth was 17% and 18% in constant currency. Results were ahead of expectations, driven by stronger GitHub Copilot adoption and usage, partially offset by weaker sales execution in our M365 E3 and stand-alone SKUs. We continue to evolve our sales motions to balance those more traditional SKU per seat rhythms with the newer premium E7, M365 Copilot and usage-based productivity offerings. Year-over-year growth was driven by both ARPU and seats.

ARPU growth was led by our premium seat-based offerings, including M365 Copilot, E5 and E7, which has already reached millions of paid seats with additional contribution from usage-based offerings, primarily GitHub Copilot. And Copilot revenue growth accelerated quarter-over-quarter with continued adoption across M365, GitHub and Security. Paid M 365 commercial seats grew 5% with some impact from the weaker sales execution noted earlier. Growth was driven by installed base expansion across all customer segments, though primarily in our SMB and frontline worker offerings.

M365 Consumer Cloud revenue increased 17% and 16% in constant currency, down sequentially as we begin to lap the benefit from the price increase in the prior year. Productivity and server licensing revenue decreased 3% and 2% in constant currency, primarily driven by a decrease in transactional purchasing as we lap the benefit from product launches in the prior year, partially offset by a greater mix of contracts that carry higher in-period revenue recognition. Industry Solutions Cloud revenue increased 10% with growth in Dynamics 365 and LinkedIn Talent Solutions, though booking trends continued to moderate. And in Frontier and Support Services, revenue increased 4% and 5% in constant currency, driven by support.

Microsoft Cloud gross margin percentage declined year-over-year to 66%, driven by sales mix shift to Azure, partially offset by product level margin improvements, primarily in Azure and for this segment. Gross margin dollars increased 20% and 21% in constant currency. Operating expenses increased 10%, driven by the R&D investments noted earlier. Operating income grew 24% and 25% in constant currency, and operating margins decreased slightly year-over-year to 53% as strong revenue growth, platform efficiency gains and focused investments across our portfolio were more than offset by sales mix shift to Azure.

Now to Devices and Consumer. Revenue was $15.5 billion and declined 3%. Search and advertising revenue ex TAC increased 12% and was ahead of expectations, driven by higher volume and rate across LinkedIn, Edge and Bing. Xbox revenue decreased 2% and Xbox Content and Services revenue decreased 2% and 1% in constant currency, driven by declines in third-party content and Xbox Game Pass, partially offset by growth in first-party content. And Windows OEM and devices revenue decreased 22%. Windows OEM revenue was below expectations and decreased 27% and 28% in constant currency, driven by the Windows 10 end of support demand in the prior year, inventory drawdowns by OEM and channel partners and weaker PC market demand.

Inventory levels remain elevated. Segment gross margin dollars were relatively unchanged year-over-year and operating expenses decreased 3%, driven by Xbox, partially offset by the continued investments in R&D noted earlier. Operating income increased 2% and 3% in constant currency, and operating margins increased year-over-year to 35%, driven by lower amortization of intangible assets from the Activision acquisition, partially offset by the impact from a lower mix of Windows OEM revenue.

Now moving to the outlook for our second quarter, which unless specifically noted otherwise, is on a U.S. dollar basis. Based on current rates, we expect FX to have no meaningful impact on total revenue, COGS or operating expense growth or on revenue growth in either segment. Starting with segment guidance. We expect revenue from Agents and Infra of USD 80.1 billion to USD 80.7 billion or growth of 24% to 25%. In commercial bookings, we expect healthy growth in our commercial sales motions, though results will be impacted by the significant Azure commitments signed in the prior year. Timing related to significant Azure commitments like these will result in material volatility in bookings, making quarterly growth rates less indicative of underlying sales execution.

In Azure, we expect revenue growth of 49% to 50% in constant currency with continued efficiency gains across our CPU and GPU fleet, helping to offset some 3P GPU-as-a-Service contracts that have been delayed to H2. Customer demand continues to exceed available supply. And looking beyond Q2, year-over-year growth will continue to reflect the timing and scale of engineering efficiencies delivered in the quarter and new capacity adds as well as a strong prior year comparable that included accelerating growth in the second half. In M365 Commercial Cloud, we expect growth of approximately 19% in constant currency with FX expected to reduce reported growth by less than 1 point. We expect continued momentum across our premium per seat offerings and usage-based offerings, along with some improvement in execution across E3 and our stand-alone SKUs.

Growth will be driven primarily by ARPU with additional contribution from seats, and we continue to expect revenue growth to accelerate through the year. M365 Consumer Cloud revenue should grow in the low teens, down sequentially as we lap the price impact noted earlier. Productivity and server licensing revenue should decline in the mid- to high single digits, down sequentially due to a lower expected mix of contracts that carry higher in-period revenue recognition. Quarterly growth can vary materially based on contract mix, product launch timing and other transactional purchasing trends.

In Industry Solutions Cloud, we expect revenue growth to be in the mid- to high single digits with continued impact from the bookings trends noted earlier. And in Frontier and Support Services, we expect revenue to grow in the low to mid-single digits. Microsoft Cloud gross margin percentage should be roughly 65%, down year-over-year as continued improvement in product level margins is more than offset by sales mix shift to Azure. In Devices and Consumer, we expect revenue to be USD 15.9 billion to USD 16.4 billion.

Search and advertising revenue ex TAC growth should be in the low double digits, driven by consistent trends in rate and volume with overall search and advertising revenue expected to grow at a similar rate. And in Xbox Content & Services, we expect revenue growth to be in the mid- to high single digits, up sequentially, driven by the timing of content launches across both third and first party. With content launch patterns through the year, we expect growth to be the strongest in the second quarter with some deceleration in Q3 and further moderation in Q4.

Hardware revenue in Q2 should decline year-over-year. Windows OEM and devices revenue should decline in the mid-30s. Growth will continue to be impacted by elevated inventory levels, PC market demand and a prior year comparable that benefited from Windows 10 end of support. As in prior quarters, the range of potential outcomes remains wider than normal. We expect these trends to continue through the rest of the fiscal year. Therefore, at a total company level, revenue should be between USD 96 billion and USD 97.1 billion or growth of 18% to 19%. We expect COGS of USD 32.7 billion to USD 32.9 billion or growth of 26% to 27% and operating expense of USD 18.5 billion to USD 18.6 billion or growth of 9%, driven by continued investment in R&D compute capacity and talent. Our cost guidance across both COGS and OpEx includes an estimate for the recently announced employee voluntary exit offer.

Operating margins should be down slightly year-over-year. Excluding any impact from our investments in OpenAI, other income expense is expected to be roughly negative $300 million as interest income will be more than offset by interest expense, which includes the interest payments related to data center finance leases. And we expect our Q2 effective tax rate to be approximately 20% Next, capital expenditures. In Q2, we expect CapEx spend will be over $50 billion.

Therefore, for the calendar year, spend should be approximately $170 billion due to the 3P capacity delays noted earlier. In the second half of the fiscal year, we currently expect CapEx spend to accelerate primarily from CPUs, inclusive of higher memory and storage costs and GPUs as well as the 3P capacity deliveries that shifted to H2. And as we continue to improve the efficiencies of our infrastructure, each dollar of CapEx has the potential to support more revenue and higher returns over time. Given our strong business performance, we expect to continue to remain free cash flow positive in FY '27. To complement cash generated by the business and preserve financial flexibility, we will access additional sources of funding when we believe it is attractive to do so.

In closing, we are off to a strong start to FY '27 with accelerated growth across the Microsoft Cloud. Product improvements, along with efficiency gains across our data center infrastructure and our first-party apps and agents are increasing the value and improving the outcomes we deliver to our customers. Delivering that value to our customers positions us for durable long-term growth. With that, let's go to Q&A, Jonathan.

Thanks, Amy. We'll now move over to Q&A. [Operator Instructions] Operator, can you please repeat your instructions?

[Operator Instructions] And our first question comes from the line of Adam Wood with Morgan Stanley.

Alongside lower-than-expected CapEx and those better cloud gross margins sequentially. Can I ask, as the incremental AI capacity comes online, what do you see as likely to be the more important driver of returns over the next few years, improving utilization and unit cost within Azure or your ability to monetize more layers of each workload across Fabric, Foundry and Copilot.

Thanks for the question, Adam. I'd say all of the above. But I think the way we are coming at this is we have one pool, and we want to get the pool to be as efficient and fungible, right? That's sort of the place where if I sort of think about all of the physics of how we're building out the data centers, all of the operational rigor with which we can bring up the capacity. And then, of course, how do we run the workloads that are optimized on that infrastructure. And once you do that, then it's just a function of how we allocate it by looking at the demand signal. And even there, even in our new segment reporting, I think you kind of see it, which is copilot Copilot state full database underneath it is Work IQ. That gets used in foundry, which is where people will build, let's say, a multi-agent system for supply chain that spans both unstructured and structured.

Of course, it uses a post-Cris database in Azure. And so -- and then, by the way, all the agentic work is registered in the Agent 365 and perception manages the security of it, right? So to us, at least we look at all of this as the same infrastructure. Of course, we report to you in a variety of different products. But your core question of where is the margin going to come from? Where is the leverage is going to come from? It's not going to come from any -- it's going to come from every one of them, and it compounds when we work them across.

I think that's -- and Adam, I'm sorry if we missed the front half of your question. I think we at least caught the back half.

The next question comes from the line of Karl Keirstead with UBS.

Amy and Satya, I'm wondering if you could unpack what I thought was one of the key metrics from this call, and that is the guidance for Azure growth of 49% to 50% constant currency in December. I can imagine there's a number of variables, the capacity you're putting up, the allocation to third-party customers, probably OpenAI strength. But perhaps you could comment a little bit on what some of the key drivers are given how good that number seems at least to me.

Thanks, Karl, for the question. Maybe I'll start by talking a little bit about Q1, and I think that will help to think through Q2 and even going forward. And it's not unrelated really to Adam's question at some level as well. In Q1, I would say the engineering teams have really delivered very, very good work in terms of improving efficiency at the platform layer. That is not just a statement around the efficiency of GPUs. It's a statement around the efficiency of every layer, how the app utilizes the platform and how foundry utilizes the platform as well as the Azure platform itself. Those efficiencies they're sort of ground out every week and you have good ideas, a lot work, some don't, and you stay at it.

So they're not maybe as predictable in terms of how much impact we'll get in quarter. Q1, they did a very good job. And you saw some of that result in us being able to put more allocation into foundry in the quarter. That's a great thing because we're -- as I said in the call, we still have demand that exceeds supply. So where we can have our engineering teams deliver those efficiencies, we do really try to put it into foundry when we can because it's so important, as Satya talked about, into how customers broadly are building what I would call against the new stack. Secondly, and you mentioned it, token pricing will create a little bit of variability. We had higher token pricing than we expected in the quarter.

Listen, high-quality models that we're able to take advantage of and see good margins on do benefit us in the quarter, and that did happen in Q1. And I think also in that number in Q1 was the efficiencies I talked about applied not just to our GPU fleet. But I think people sort of underestimated when the example Satya gave for Adam's question, that's a GPU, CPU storage memory workload. I mean it takes all of the pieces. And so our ability to also create additional capacity within our existing fleet gave us some room to be able to convert that actually pretty quickly as well in Q1.

So in Q2, we've got those efficiency gains, which is great, and those will continue into Q2. We need to find some more, which we're counting on enable to get to that guidance that I gave because I have faith in the team's work in Q1 continuing into Q2. But we also need that work to be able to help us a little bit because we had some supply we were counting on Q2 move out into H2. And so I think -- and to your point, there's also some unknown on some token pricing through the quarter, that stuff can change. So I think I feel good in terms of the efficiency gains that we saw when we're able and when I'm able to see those and put them in, I feel good about guiding to that. But as other things are less predictable. I mean I hope that's helpful on the levers, Karl.

The next question comes from the line of Mark Moerdler with Bernstein Research.

Congratulations on the solid quarter and the solid guidance. It's really nice to see how well you guys are executing. I'd like to get a sense on the drivers of the AI strength in Azure being from enterprise versus digitally native in general or AI native. With inferencing starting to ramp and Microsoft the go-to for enterprises, how should we think, Amy, about the mix of enterprise customer workloads within overall workloads at this point? And Satya, with the strong growth of the Azure AI platform offering, can you give us color on what's driving that? Is it, again, enterprise? Is it digitally native? Obviously, it's a mix of both. But any color that shows the ramp of that enterprise, I think, would be valuable.

Yes. Thanks for the question, Mark. Maybe I'll start and maybe you can add to it. I was even here in Europe just last week, and it was great to see the qualities of the Agentic systems, right, that are just broadly getting deployed, right? When Unilever deploys what is their automated order taking and trade coupon dispute management stuff, all as a fully agentic touchless system. Rolls-Royce deploys a system to really manage all of their maintenance or whatever their jet engines with 30,000 parts, a pretty complex supply chain, both planning, execution, inventory management system, all built on foundry Agentic systems. RBI, which is an Austrian bank, showed me 20-plus business processes end-to-end. And they were talking about like whether it's in for credit or what have you, how they are able to now do that so much more optimally, right?

So the breadth, the quality, their ability -- oh, by the way, the multimodel nature of it, right? They look -- they start with a frontier model. Many of them, in fact, the Unilever one, they just dropped an MAI model trained using our traces. So even the degree of sophistication of what I think of as model optimization, workload optimization, token optimization was just great to see. So the breadth is starting.

Finally, it does feel like the enterprise workloads. They feel like to me, like the database applications enterprises built in the past that I grew up with is happening now when it comes to AI. And in fact, if anything, some of the shiny object stuff is waning, right, which is not about the next big thing. They're really more about, wow, can I build the system? Can I own the weights? Can I manage my costs? Can I have reliability? What about security? And so in that sense, the enterprise breadth motion is starting. But like everything else, the enterprise stuff is not like a quarter-over-quarter. It's a decade-long thing, and that's what I've been waiting for as opposed to like 4 customers who buy a lot of capacity from you and you can show a few quarters of growth. And that's not a business as far as I'm concerned.

So I want the breadth. We are well positioned on the breadth. And we have large customers who want to buy from us, and we're open for business with it. And as long as the pricing is right, we may sell them some. If not, we will be patient about building the breadth of business.

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

Satya, I want to stay on this topic of breadth because you had a comment in the prepared remarks talking about how MAI can now be comparable to leading models within 3 to 6 months of their release. At the same time, you have OpenAI at 11% of revenue in the quarter. So maybe just take a step back and give us your latest thinking on how you think token mix and token value is going to evolve between the frontier MAI and maybe what some of your more sophisticated enterprise customers are doing from a proprietary AI strategy standpoint?

Yes. So I think -- in fact, I would say that you should think of 3 things perhaps as in the toolbox for any workload. The first is a new model drops and it's got some great performance and you want -- you can eval Max on whatever that matters to you, i.e., the outcome that matters to you by using it, people should use it. We saw that with some of the OpenAI models in the last quarter. They were fantastic models. And we're obviously very thrilled about the progress OpenAI has made with Astra and also what's to come.

And I thought we saw some good demand on that. Then the next thing people do is, hey, can I actually put multiple of the frontier models because these got -- these are very different prices. In fact, this morning, artificial analysis had the -- what I think is the real benchmark, which is the coding agents using our Fusion, and I think we were #2 or #3 there, which is awesome to see, right, which is you can now get frontier performance, 45% faster, 75% cheaper. Like why -- who would want to not take that, right, which is like if you're a coding agent user, you want, in fact, Anthropic and OpenAI, by the way, to compete for your workload, not like once a quarter, but for every task.

And that, to me, is what the value of these routers and auto fusion stuff routing does, and we are very excited about it. And then the last thing is another important one. A lot of customers are saying, "Hey, look, take the Unilever one, right? It's their system. They will ultimately take even the traces after having done even a bunch with frontier models, and they just want to train their own model. So it can be open weight or in this case, it was MAI model. And so I think all of that needs to be as options that are available.

And then depending on pricing, depending on the need of the customer, they will exercise it. And there's a temporality to it. It's never like, oh, I start one place and that's where I stop, right? It's like I may start at the frontier, I may go to an open weight, I may go to an MAI model. I may go back to the frontier for the next drop for a few months, come out. Like that's how I think -- how dynamic these workloads are. And that's where when we think about foundry or when we think about fabric or when we think about Copilot and GitHub Copilot, we really want to build a model system that delivers customers value continuously versus them being stuck in any one thing.

The next question comes from the line of Brent Thill with Jefferies.

Amy, good to see Azure approaching 50%. You have another line item approaching 50%, and that's your operating margin. I think many ask how can you sustain this level of growth with 50% on the top and a 50% on the bottom? Maybe you can discuss the operational improvements you continue to put forward to maintain that high level on the bottom line.

Thanks, Brent. Let me first say, the best solution here is revenue growth. And so I would say I tend to start and focus there, even though I think many people seem to think I start at the bottom of the P&L and work up. The right answer is you always start at the top of the P&L and work down. And so what we've been focused on, and I think really, it's almost inherent in all of the commentary is we changed segments because we're changing how we deliver products and services to customers. We're changing it because we're delivering new and different products with a lot of value in them. And those products are far more integrated than they were before. I think when you see that revenue accelerate at the top line, of course, it does in a pretty healthy way, contribute to positive margin. And part of the positive margin dollars, which I think I need people to also focus on.

The second piece is the real focus on efficiency because efficiency, when we say the word, I think people believe it's cost related. The outcome is cost related. But the real benefit of efficiency is stability, it's security, it's reliability. It's that it delivers the highest quality unit at the lowest possible cost. And so it's operational excellence. And so I think when you think about you can deliver product excellence, which is what we're really focused on, I think we did a very good job on product quality in the past quarter, which I think team should be proud of in lots of places. I think then you have this sort of operational focus on improvement. And then it is a real focus in terms of operating expense to make sure it's going toward building the new thing or providing security and quality.

And I think that's -- we've made, I think, a very active choice to be really focused on those 2 things, especially in this environment and how important it is. And I think when you do those things and you happen to be in a good market with breadth and breadth in customers and a global business, we've seen a good result. This quarter, in particular, when you're delivering a lot of efficiency gains in the quarter and you can deliver revenue growth without spending additional costs, you, of course, see a benefit, which we did. But I feel good about what's behind that number, which is, I think, a focus on the right levers.

Our next question comes from the line of Brad Reback with Stifel.

Great. Sorry about that. I did not hear my name. On M365 consumption, when you all shifted GitHub from a per seat model to a consumption model, it unlocked a ton of revenue growth for you all. As you think about the tools in place now unlock -- on the consumption side for the core M365 customer, do you think the unlocks is fast and without sales effort? Or is it going to take a fair amount of effort to get those meters spinning?

Yes. Thanks for the question. I think on the coating side, it will be faster just because I think people are used to paying on this PayGo as the sort of the general market and how it works. And if anything, hopefully, some of the things that we announced even at GitHub Universe today with Fusion are a more attractive way, right? Because you even -- if you think about it, right, you buy a bunch of credits and you want to use those credits most efficiently. That's the promise of GitHub Copilot Fusion. So we are very optimistic about making progress on that. And then when it comes to Copilot itself, right, we are trying to make sure that we, in some sense, are driving more and more value into the subscription with every model drop and essentially consolidating the frontier gains for broad deployment.

And in many instances, as my remarks said, IT funds this, right, as the thing that you want to have standard issue across the board. And at the same time, you have your pay-as-you-go or consumption meters for places where you need it. And so -- so both of those options are open. And I think that's why the -- with Copilot, Amy referenced, which is we are trying to make sure that customers understand both options. They're able to use the 2 options in the optimal way that benefits them. We don't want them to overbuy one thing and then feel like they're token maxing because that doesn't serve anybody's purpose.

We want to make sure that they understand the value of the subscription and how it's a continually improving value. It's not even static. And at the same time, you can exercise this consumption meter. And so that's a little bit of the degree of sophistication in both in our sales motions, even our customers' understanding that's, I think, coming together. We feel very good about the progress, but a lot more to do.

And I think I would say it's also why we've made the investment into the frontier company here, coupled with changes in our sales motion, which I talked a little bit about, we need both of those things to make sure customers are -- because what I think Satya is saying, we want them to utilize both the Copilot license to its fullest and understand when the consumption meter is delivering value to them. The investment in the FDE organization or the Frontier core as well as new motions within our modern work and Microsoft 365 sales teams are both meant to do that. And I think to Satya's point, it will be a slower move outside of coding. But I also still think we're still seeing encouraging signs, and I feel positive about that.

Operator, we have time for one last question.

The next question will come from the line of Tal Liani with Bank of America.

Just a clarification on something you just said. Did you disclose the Copilot numbers this quarter? But my question is more about the capital commitment or the capital strategy. You said in the prepared remarks that you will approach raising capital if need be in the future. Can you discuss your strategy around it between debt and equity and other sources? How do you intend to raise capital if need be?

Great. I'll take that. Listen, I think what I said is that if it's attractive and makes sense, we're lucky. We have a lot of flexibility in terms of our capacity to raise financing. I've talked about this before. There's a form of financing called leasing, and we've been using that one as a vehicle. There's also -- we're fortunate to have access in the debt markets, given that where our ratings are and sort of we haven't been able to -- and we haven't yet, I would say, tapped that. And then, of course, there's other options.

But I think the way I see it is we're feeling good that we are free cash flow positive. We feel good that the return on every dollar of CapEx we're spending, we're confident in that. And therefore, we feel like we have lots of options to be able to tap the financing markets when they make sense and they're attractive. And we did not disclose, to answer your direct question, an update on the Copilot number this quarter. We talked about momentum in revenue and usage.

Thanks, Tal. That wraps up the Q&A portion of today's earnings call. Thank you for joining us today, and we look forward to speaking with all of you soon. Thank you all.

Thank you. This concludes today's conference. You may disconnect your lines at this time, and we thank you for your participation.

Greetings, and welcome to the Microsoft Fiscal Year 2026 Fourth Quarter Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded. It is now my pleasure to introduce Jonathan Neilson, Vice President of Investor Relations. Please go ahead.

Good afternoon, and thank you for joining us today. On the call with me are Satya Nadella, Chairman and Chief Executive Officer; Amy Hood, Chief Financial Officer; Alice Jolla, Chief Accounting Officer; and Brian DeVeau, Deputy General Counsel and Corporate Secretary. On the Microsoft Investor Relations website, you can find our earnings press release and financial summary slide deck, which is intended to supplement our prepared remarks during today's call and provides a reconciliation of differences between GAAP and non-GAAP financial measures. More detailed outlook slides will be available on the Microsoft Investor Relations website where we provide outlook commentary on today's call.

On this call, we will discuss certain non-GAAP items. The non-GAAP financial measures provided should not be considered as a substitute for or superior to the measures of financial performance prepared in accordance with GAAP. They are included as additional clarifying items to aid investors in further understanding the company's fourth quarter performance in addition to the impact these items and events have on the financial results.

All growth comparisons we make on the call today relate to the corresponding period of last year, unless otherwise noted.

We will also provide growth rates in constant currency when available as a framework for assessing how our underlying businesses performed, excluding the effect of foreign currency rate fluctuations, where growth rates are the same in constant currency, we will refer to the growth rate only.

We will post our prepared remarks to our website immediately following the call until the complete transcript is available.

Today's call is being webcast live and recorded. If you ask a question, it will be included in our live transmission, in the transcript and in any future use of the recording. You can replay the call and view the transcript on the Microsoft Investor Relations website.

During this call, we will be making forward-looking statements, which are predictions, projections or other statements about future events. These statements are based on current expectations and assumptions that are subject to risks and uncertainties. Actual results could materially differ because of factors discussed in today's earnings press release, in the comments made during this conference call and in the Risk Factors section of our Form 10-K, Form 10-Q and other reports and filings with the Securities and Exchange Commission. We do not undertake any duty to update any forward-looking statement.

And with that, I'll turn the call over to Satya.

Thank you very much, Jonathan. It was a very strong close to what was a record fiscal year for us. All up, our annual revenue surpassed $331 billion, up 18%, Microsoft Cloud surpassed $214 billion, up 27% and Azure surpassed $100 billion, up 41%. Going forward, we have 2 goals; first, ensuring AI empowers every person, amplifying their agency and ambition; and second, empowering every organization to build their own continuous learning loop and ensuring that they don't outsource their core IP.

Now let's talk about how we are delivering this across our stack starting with our AI platform and infrastructure. We added 31 new data centers across 5 continents this quarter, bringing the total to 88 this year as we expand our footprint in response to accelerating demand. We are also bringing capacity online faster than ever. Over the last fiscal year, we have reduced Doctor lifetimes for new GPUs in our largest regions by nearly 50%. All up, we added another gigawatt of capacity this quarter and remain on track to roughly double our overall capacity in just 2 years.

We're also getting more from the infrastructure we already have by optimizing across silicon systems and software. For example, we increased the throughput for Copilot workloads 4x since the start of the year. AI sovereignty is increasingly top of mind for our customers, and we are expanding our offerings to meet that need. Just last week, we announced a partnership with Mistral to bring its models to Microsoft sovereign cloud, enabling customers to run them across public, customer-controlled and fully disconnected environments.

We also continue to modernize our fleet with our own silicon innovation alongside the latest from NVIDIA and AMD. Miya 200 continues to scale. It delivers 30% better performance per dollar than the latest generation hardware in our fleet and is now supporting both OpenAI and MAI models, and we will be among the first cloud providers to deploy next-generation rack scale AI infrastructure based on AMD Helios and NVIDIA virurubin.

When it comes to running agents, CPUs are just as important as GPUs. Our Cobalt VMs are powering both our own first-party workloads as well as workloads for customers, including Adobe, Arm, Elastic, OpenAI, Sprinkler and TomTom. And by the end of this month, we expect to have our Cobalt 200 Racks in over 25 data centers around the world as we rapidly expand capacity.

Now let me turn to the end-to-end platform we are building on this infrastructure to run, govern and distribute apps and agents. It starts with model choice. Every customer wants the right model for each task based on quality, latency, cost and compliance. We offer the broadest model catalog in the cloud with over 11,000 models, including the latest from OpenAI, Anthropic, Mistral, XAI as well as our own MAI family.

Since the start of the year, we have seen 5x increase in the number of customers building with models from multiple providers. Levi Strauss and Co, for example, is using models from OpenAI and Anthroopic on foundry as it brings more than 1,000 domain-specific agents into a unified enterprise AI platform.

We are also accelerating our own model development. We announced more than a dozen new models across image, voice, transcription, coding, security, including our first reasoning model MAI Thinking One, all with cost-efficient influence at the core for the enterprise use cases. We are co-designing these models with our silicon, and we are seeing 40% better performance per watt when running MAI models on Maia 200. But more importantly, we are building a new model system where the harness context, memory and action space are separate from any 1 model family, thereby moving the frontier on the cost to outcome curve. And it's not just about cost. It also has the added benefit of business continuity and resilience because every model is substitutable. This is the system we are using in our products with great results. For example, millions of developers have used MAI Code 1 flash on GitHub Copilot achieving higher core acceptance rates and 10% lower median token usage while still having access to frontier capabilities from OpenAI and Anthropic.

In Excel MAI Code 1 flash is delivering comparable quality to GPT-56 for the most common task while operating at significantly lower costs. In security, MAI CyberOne Flash achieves better performance than much larger [indiscernible] model, but at half the cost when combined with our multi-agent security harness. More broadly across our model implementations, we are seeing significant efficiency gains, including 89% reduction of GPU costs in Dynamics 365 with MAI voice to flash, and up to 84% reduced GPU costs in PowerPoint with MAI Image 25. And this system is available to any company to use as part of foundry.

The next layer is the enterprise data in context. The data estate is evolving from primarily supporting apps used by people to supporting agents. Customers are rapidly adopting our AI-optimized databases like CosmosDB and PostgresSQL to give agents fast, secure access to real-time data and context they need for memory and retrieval. PostgresSQL revenue was up 55%, accelerating for the third consecutive quarter. Also, the number of PostgresSQL customers also using foundry increased 80% as customers increasingly choose it as the database for AI workloads. And we are going further with HorizonDB, our new fully managed PostgresSQL service on Azure, which delivers 3x the throughput of self-managed deployments. When it comes to analytics, we now have over 40,000 paid fabric customers, up more than 60% year-over-year, and over 17,000 customers now use foundry and fabric, up 60% year-over-year as enterprises connect agents to real-time operational, analytical and unstructured data in fabric.

This quarter, we also introduced Rafin, the agent-first STK that delivers a back-end as a service for building apps and fabric. More than 2,500 customers have already used Rafin and is now powering the back ends for apps created with Repla 2.

On top of this data estate, we are building an IQ layer that combines data with model capabilities to deliver the right context at the right time. Tens of thousands of customers, including nearly 90% of the Fortune 500 are already grounding their agents in enterprise context with foundry, fabric and work IQ. And this quarter, we introduced Web IQ, which gives agents access to real-world intelligence from across the web. It is already being used by the most popular AI assistants, including chat GPT.

Beyond model choice, data and context, we are building foundry as the complete app and agent stack. It gives agents access to the IQ layer, the tools they use along with durable state and memory, secure sandboxes, rubrics and [indiscernible] and even their own self-improvement loops. We now have 100,000 foundry customers and revenue more than doubled year-over-year.

Telefonica, for example, adopted foundry as the foundation of its corporate agentic platform. with its first wave of agents tackling mission-critical network operations. All up the number of foundry customers at 1 trillion token annualized run rate increased 4x year-over-year. And finally, with Agent 365, we offer a control plane that extends company's existing governance, identity, security and management frameworks to agents they build. Just 2 months in, Agent 365 now has nearly 40 million agents listed across tens of thousands of companies.

Now let me turn to the apps and agents we are building on top of this platform for individuals and organizations. When it comes to knowledge work, we now have over 30 million paid Microsoft 365 Copilot seats, with net seat adds more than doubling quarter-over-quarter. Copilot is evolving rapidly from chat to co-work to auto pilots. Last month, we made cowork generally available, helping customers complete multistep tasks grounded in their work data while meeting enterprise security and compliance requirements. This quarter, we also introduced autopilot's autonomous long-running agents with full enterprise compliance, including always-on personal agent powered by OpenClaw. And this quarter, we are bringing these Copilot experiences together, including code in 1 super app spanning both consumer and commercial experiences. This is a major step forward, and I look forward to sharing more soon.

More broadly, we have steadily been improving the quality and performance of Copilot and have been delighted by the recent customer feedback. Over the last 3 quarters, user satisfaction scores have doubled and are now at an all-time high. And this quarter alone, we cut latency by 25%. These quality improvements, together with continued product innovation, are driving record usage intensity. The number of conversations per user nearly doubled year-over-year. Average weekly engagement is on par with Outlook and Teams. And the time from deployment to what we think of as high usage, meaning monthly active usage about 80% across a customer's user base, has fallen from months to just days over the past year.

The number of customers with more than 50,000 seats increased over 7x year-over-year, and the number of enterprise customers deploying Copilot to the majority of their information workers grew nearly 75% quarter-over-quarter, a signal of how central Copilot has become to their operations.

NHS England, for example, is rolling out Copilot to 505,000 clinicians and staff, the largest health care deployment of its kind after a trial showed it saved employees an average of 43 minutes per day. KPMG is expanding its deployment across its global workforce of more than 276,000 professional, and HSBC committed to 200,000 seats to accelerate its workforce transformation. AstraZeneca, Boeing, Infosys, Coke Inc., Procter & Gamble, Selantis, Tata Consulting Service, University of Pittsburgh Medical Center, Wells Fargo and Wipro, each purchased 60,000 or more. And we have been encouraged by the response to our new E7 suite as customers increasingly go all in on an integrated AI offering that brings together Copilot, E5, Entra and Agent 365.

Just 2 months after launch, hundreds of enterprise customers have already purchased millions of seats. And this quarter, EY deployed E7 to 400,000 employees in our largest win to date. In addition to this, we are also evolving our business model beyond per seat to per seat plus consumption, further expanding our TAM and delivering more customer value.

Earlier this month, we added usage-based billing to co-work with thousands of customers already paying for and actively using it. In biz apps, we have been reinventing Dynamics 365 for an agent-first world. We are exposing over 650,000 MCP actions across sales, finance, supply chain, HR and customer service so that agents can now access business context and take action using the same data models, rules, permission, security guardrails and audit trails as any application user. And we are also moving from seats to seats plus consumption model.

Customer service is at the forefront of this transformation with usage-based credit consumption in this category up 4x quarter-over-quarter with customers like Northern Trust using our tools to drive proactive intelligence.

When it comes to developers, GitHub Copilot now has 50 million users. This quarter, we introduced usage-based billing and have continued to see business and enterprise seat growth and also significant consumption revenue after the new model went into effect. Copilot revenue accelerated over 60% quarter-over-quarter. All up, GitHub now has 225 million users as organizations across every industry, including over 90% of the Fortune 500 choose GitHub for their AI-powered development.

The agentic era is being built on GitHub. Every major coating agent runs on the platform and 1 in 3 pool requests on getup now involves an agent.

In security, we are helping customers both secure their AI deployments and use AI to strengthen their security posture. To date, Purview has audited over 15 billion Copilot interactions to meet compliance obligations, up nearly 360% year-over-year. And earlier this week, we introduced Project Perception, a complete multimodal agentic security system that brings together teams of agents to simulate attacks, investigate threats and drive remediation. As Perception moves beyond private preview, we expect to bring it to customers through a consumption-based offering.

In health care, we are on pace to automate over 100 million patient encounters this calendar year, including 28 million this quarter, up 2x year-over-year. Mass General Brigham rolled out Dragon Copilot to over 4,000 providers, and after a study found AmbientAI reduced burnout by 21%. And in Science, Microsoft Discovery, now broadly available, provides a comprehensive platform for building and governing agentic workflows for science and engineering. Early customers include BHP, GSK Pacific Northwest National Lab.

Across both our high-value agentic experiences and the AI platform and infrastructure, we are focused on helping customers turn AI into measurable outcomes. The most comprehensive and valuable data in the world is inside each of the customer tenants, and therefore, there is a tremendous opportunity to turn customers' workflows, domain knowledge and accumulated judgment into AI systems that learn and improve with every usage.

To help customers capture that opportunity, this month, we launched Microsoft Frontier Core, the largest outcome-driven engineering organization in the industry. We will embed 6,000 industry and engineering experts with customers to co-design co-innovate and continuously improve AI systems at scale. We've been testing this model over the past year, completing over 330 projects across 164 customers, including many of the world's leading companies across industries. For example, our FDE teams worked with Novo Nordisk to build an agent that helps analyze clinical data while meeting its strict compliance requirements. And we partnered with LSAG to embed AI into LSAG workspace, helping finance professionals ask complex questions and quickly finances across structured and unstructured financial content.

Finally, let me talk about devices and consumer. When it comes to Xbox, we are making the necessary decisions required across our content portfolio, platform and operations to reset the business for long-term growth. We have the best IP in the industry and talented studios around the world, and believe we can bring these strengths together and expect to return the business to growth in fiscal 2027.

In Windows, we are investing to ensure that it has the best quality and fundamentals while also ensuring it's the best place to run secure edge AI. We see significant opportunity for Windows to become the offload for unmeeted intelligence, combining powerful on-device compute with enterprise-grade security.

In search and advertising, Bing and Edge have both taken share for 5 straight years. And LinkedIn continues to see strong engagement across the platform with double-digit member growth for the fifth consecutive year. Recruiters at over 20,000 companies are now using our AI-powered solutions to reduce time to hire and improve candidate matching. Seats increased to 140% quarter-over-quarter.

In closing, I'm energized by the opportunities I had. I've never been more confident in Microsoft's opportunity to drive durable long-term growth and ensure the benefits of AI flow broadly. With that, let me turn it over to Amy to walk through our financial results and outlook.

Thank you, Satya, and good afternoon, everyone. This fiscal year, we delivered over $331 billion in revenue, with growth accelerating to 18%, driven by strong demand across both the Azure platform and our first-party AI applications and services. Operating income growth outpaced revenue growth, increasing 21% to more than $155 billion as we invested in long-term growth while continuing to expand operating leverage.

This quarter, revenue was $90 billion, up 18% and 17% in constant currency. Gross margin dollars increased 15% and operating income increased 18%. Earnings per share was $4.74, an increase of 23% when adjusted for the impact from our investment in OpenAI. And FX was roughly in line with guidance.

Several discrete items impact our financial results in the quarter when compared to our forward-looking guidance provided on our April earnings call, resulting in a benefit of $0.27 on diluted earnings per share. These included a $3.2 billion gain from our investment in Anthropic and lower-than-expected expenses related to the voluntary retirement program, which were partially offset by severance expense and impairment charges in Xbox.

When adjusting for these items, we exceeded expectations across revenue, operating income and earnings per share due to strong demand and execution in the quarter.

Company gross margin percentage was 67%, down year-over-year, driven by sales mix shift to Azure as well as continued investments in AI infrastructure and growing product usage, partially offset by ongoing efficiency gains, particularly in Azure and M365 Commercial Cloud.

Operating expenses increased 10%, driven by continued investment in R&D compute capacity, talent and data to support product development across the portfolio.

G&A growth was impacted by a low prior year comparable as well as some of the discrete items mentioned earlier.

Operating margins increased slightly year-over-year to 45%.

Total company headcount declined 2% year-over-year.

When adjusted for the impact of our investments in OpenAI, other income and expense was $2.8 billion, driven by the gain on investment in Anthropic noted earlier.

Capital expenditures were $41 million, including the impact from higher component pricing, as noted in our guide. Roughly 2/3 of our CapEx was for short-lived assets, primarily CPUs and GPUs as customers increasingly build solutions that leverage both AI and non-AI infrastructure. The remaining spend was for long-lived assets. This quarter, total finance leases were $5.6 billion and were primarily for large data center sites. And cash paid for PP&E was $35.8 billion.

Cash flow from operations was $55.4 billion, up 30%, driven by strong cloud billings and collections, partially offset by an increase in operating lease payments. And free cash flow was $19.6 billion, reflecting higher capital expenditures.

And finally, we returned $10.2 billion to shareholders through dividends and share repurchases, bringing our total cash return to shareholders to over $43 billion for the full fiscal year.

Now to our commercial results. Commercial bookings grew 18% when excluding the impact from OpenAI, driven by strong execution in our core annuity sales motions and reflecting broad customer demand across geographies and customer segments. Bookings increased 10% and 11% in constant currency when including Azure commitments from OpenAI.

Commercial remaining performance obligation grew 84% to $678 billion. All sequential commercial RPO growth was driven by commitments from customers outside of Frontier model companies. And RPO increased 25% when excluding OpenAI. RPO, including OpenAI has a weighted average duration of 2.3 years and roughly 30% will be recognized in revenue in the next 12 months, up 37% year-over-year. The remaining portion recognized beyond the next 12 months increased 112%.

Microsoft Cloud revenue was $59.3 billion and grew 27%, reflecting strong demand across Azure and our first-party AI applications and services. And for the full year, our cloud revenue surpassed $214 billion with nearly 90% from customers outside of Frontier model companies.

Microsoft Cloud gross margin percentage was better than expected at 65% and down year-over-year, driven by sales mix shift to Azure as well as continued investments in AI infrastructure and increased product usage, partially offset by ongoing efficiency gains noted earlier.

Now to our segment results. Revenue from Productivity and Business Processes was $37.8 billion and grew 14%. M365 commercial cloud revenue increased 16% on an adjusted basis when normalized for the prior year comparable that benefited from 2 points of in-period revenue recognition. And on a reported basis, revenue growth was 14%. Building on our Copilot momentum from Q3, net paid seat adds more than doubled sequentially, with paid seats now over 30 million. Premium offerings, including Copilot, E5 and early traction in E7 drove ARPU growth this quarter, and paid M365 commercial seats grew 6% year-over-year, with installed base expansion across all customer segments, though primarily in our small and medium business and frontline worker offerings. And M365 commercial products revenue increased 19% ahead of expectations, driven by large long-duration M365 contracts that resulted in higher end period revenue recognition from the Windows Commercial on-premises component.

M365 consumer cloud revenue increased 24% to 22% in constant currency, again driven by ARPU growth. And M365 consumer subscriptions grew 7%. LinkedIn revenue increased 12% and 10% in constant currency, primarily driven by Marketing Solutions.

Dynamics 365 revenue increased 13% and 12% in constant currency against a strong prior year comparable. Bookings growth in ERP remains healthy, while CRM continued to moderate with longer sales cycles. Segment gross margin dollars increased 14% and 13% in constant currency, and gross margin percentage decreased slightly with increased M365 Copilot usage as we continue to invest in product quality and drive further efficiency gains.

Operating expenses increased 11%, primarily driven by the shared R&D investments mentioned earlier. Operating income increased 15% and 14% in constant currency, and operating margins increased year-over-year to 58%.

Next, the Intelligent Cloud segment. Revenue was $39.3 billion and grew 32% and 31% in constant currency. In Azure and other Cloud Services, revenue grew 43% against the prior year that included accelerating growth. Customer demand continues to exceed available capacity.

Revenue growth was ahead of expectations, driven by efficiency gains across our CPU and GPU fleet as well as process improvements to enable earlier delivery of new capacity. That additional in-quarter capacity for Azure was quickly monetized.

Results also benefited from stronger-than-expected GitHub Copilot consumption following the June business model change to align pricing with usage and value.

In our on-premises server business, revenue was relatively unchanged year-over-year and was down 1% in constant currency. Results were ahead of expectations, primarily driven by renewals with higher in-period revenue recognition from the mix of contracts. Segment gross margin dollars increased 24% and gross margin percentage decreased year-over-year, primarily driven by sales mix shift to Azure as well as the continued scaling of our AI infrastructure ahead of growing demand, partially offset by ongoing efficiency gains in Azure.

Segment gross margins were also impacted by growing GitHub Copilot usage, though margins improved through the quarter with the June business model change to usage-based pricing.

Operating expenses increased 10%, driven by shared R&D investments noted earlier.

Operating income grew 31% and operating margins with a strong focus on efficiencies and investment returns were relatively unchanged year-over-year at 41%.

Now to more Personal Computing. Revenue was $12.9 billion and declined 4% and 5% in constant currency. Windows OEM and Devices revenue decreased 7%, and Windows OEM decreased 5%, driven by lower PC market demand and a high prior year comparable that benefited from Windows 10 end of support. Results were ahead of expectations as OEM and channel partners continue to build inventory given increasing component prices. Search advertising revenue ex TAC increased 10% and 9% in constant currency with growth driven by higher revenue per search across Edge and Bing as well as higher volume, the growth was impacted by third-party partnerships. And at Xbox, revenue decreased 10% and 11% in constant currency. Xbox content and services revenue decreased 10% against the prior year comparable that benefited from strong first-party content performance.

Segment gross margin dollars decreased 2% and gross margin percentage increased year-over-year, driven by lower amortization from the Activision acquisition. Operating expenses increased 8% and 7% in constant currency, driven by the continued investments in shared R&D noted earlier as well as impairment charges in Xbox.

Operating income decreased 14% and 15% in constant currency, and operating margins decreased year-over-year to 21%.

Now before I move to outlook, effective at the start of FY '27, we are extending the estimated useful life of our data centers and office buildings from 15 to 25 years, reflecting our operating history and expected use of these assets. The impact of this update is reflected in today's guidance.

This change affects only the timing of future depreciation and is expected to have a minimal benefit to FY '27 operating income. The greater impact is on capital expenditures as more of our future data center leases will shift from finance leases to operating leases as a result of this update. Finance leases are included in capital expenditures while operating leases are not. Outside of this useful life impact, our calendar year 2026 CapEx investment expectations remain unchanged. However, the shift from finance to operating leases adjust our expectation to approximately $175 billion.

Now moving to our outlook. Let me start with some full year commentary for FY '27. First, some reminders. In both M365 commercial products and server products KPIs, we are lapping higher transactional purchasing from the timing of product launches and expect revenue from both to decline in the mid-single digits for the full fiscal year. Growth in Windows OEM and Devices will be impacted by lower PC market demand as higher component cost increased device pricing, a prior year comparable that benefited from Windows 10 end support and elevated inventory levels. As a result, we expect revenue to decline in the high teens for the fiscal year.

Moving to FX. Assuming current rates remain stable, we now expect FX to decrease full year fiscal revenue growth by less than 1 point with no meaningful impact to COGS and operating expense growth. At the company level, with strong commercial momentum, we continue to expect another fiscal year of double-digit revenue and operating income growth.

Operating expenses should grow in the mid- to high single digits, reflecting continued investment in R&D compute capacity, talent and data. And we expect FY '27 capital expenditures will grow year-over-year given demand signals across our portfolio.

Even as we invest to meet growing demand, full fiscal year operating margins should be down less than 1 point. In addition, we expect to remain free cash flow positive in FY '27.

And finally, we expect our FY '27 effective tax rate to be approximately 20%.

Now to the outlook for our first quarter, which unless specifically noted otherwise, is on a U.S. dollar basis. Based on current rates, we expect FX to decrease total revenue growth by less than 1 point with no meaningful impact to COGS or operating expense growth. Within the segments, we expect FX to decrease revenue growth in productivity and business processes by roughly 1 point and Intelligent Cloud by less than 1 point. There is no meaningful impact in more personal computing.

Starting with our commercial business. In commercial bookings, when adjusted for the impact from OpenAI, we expect healthy growth on a growing expiry base driven by strong execution across our core annuity sales motions. As a reminder, the significant OpenAI contracts signed in the prior year will result in some quarterly volatility in both bookings and RPO growth rates.

Microsoft Cloud gross margin percentage should be relatively stable quarter-over-quarter.

Now to segment guidance. In Productivity and Business Processes, we expect revenue of USD 36.7 million to USD 37 billion, a growth of 11% to 12%. In M365 Commercial Cloud, we expect growth of approximately 16% in constant currency on an adjusted basis, which normalizes for the prior year comparable that benefited from 1 point of in-period revenue recognition, or 15% on a reported basis.

Sequential growth from our momentum in Copilot, E5 and E7 is mitigated a bit by the lower ARPU, new seat adds in front-led worker and small and medium business SKUs. With the premium SKU momentum and the increased monetization opportunity from adding usage-based billing products alongside per seat licensing in July, we expect to see acceleration in M365 Commercial Cloud revenue growth through this fiscal year.

M365 Commercial products revenue should grow in the mid-single digits, driven by the timing of long-duration M365 contracts, partially offset by the impact from the prior year comparable noted earlier. M365 consumer cloud revenue should grow in the mid-teens, down sequentially as we lapped the benefit from last year's price increase. Growth will again be driven by ARPU and an increase in subscription volume.

For LinkedIn, we expect revenue growth in the high single digits. And in Dynamics 365, we expect revenue growth to be in the low teens, relatively stable quarter-over-quarter, driven by continued growth in ERP, although impacted by the bookings trends noted earlier.

For Intelligent Cloud, we expect revenue of $40.95 billion to USD 41.25 billion or growth of 33% to 34%. In Azure, we expect revenue growth of approximately 45% in constant currency, and we remain focused on delivering efficiencies that help us bridge the gaps we see as customer demand continues to exceed supply.

Even with the strong close to Q4, we continue to expect H1 growth to accelerate. And as a reminder, year-over-year Azure growth rates can vary quarter-to-quarter based on capacity, timing and contract mix.

In our on-premises server business, we expect revenue to decline in the low to mid-single digits with ongoing customer shift to cloud offerings and the prior year comparable noted earlier. And more Personal Computing, we expect revenue to be USD 12.2 billion to USD 12.7 billion as we continue to lap strong prior year comparables noted earlier, and navigate complex PC market dynamics impacted by component prices and inventory levels.

Windows OEM and Devices revenue should decline in the low 20s, driven by the market dynamics noted earlier. As in prior quarters, the range of potential outcomes remains wider than normal. Search advertising revenue ex tech growth should be in the mid-single digits, down sequentially due to the impact of third-party partnerships. Growth will continue to be driven by consistent trends in revenue per search and volume.

And in Xbox content and services, we expect revenue to decline in the mid-single digits. Hardware revenue should decline year-over-year. Therefore, at the total company level, revenue should be between $89.85 billion and USD 90.95 billion or growth of 16% to 17%, with accelerating commercial growth partially offset by the impact from the PC market dynamics noted earlier.

We expect COGS of $29.6 billion to $29.8 billion or growth of 23% to 24%. And and operating expense of $16.8 million to $16.9 billion, a growth of 7% to 8%, driven by continued investment in R&D compute capacity and talent. Operating margins should be relatively flat year-over-year. excluding any impact from our investments in OpenAI, other income and expense is expected to be roughly negative $100 million as interest income will be more than offset by interest expense, which includes the interest payments related to data center finance leases.

And we expect our Q1 effective tax rate to be approximately 20%.

Next, capital expenditures. We expect CapEx spend will be over $50 billion, including the lease reclassification impact from the useful life update.

In closing, in FY '26, we delivered accelerating revenue and operating income growth while expanding operating margins. Our execution across sales and product engineering strengthened through the second half of the year. As we begin FY '27, we remain focused on delivering products that create meaningful return on investment for our customers, which will result in durable long-term growth for Microsoft and our shareholders. With that, let's go to Q&A, Jonathan.

Thanks, Amy. We'll now move over to Q&A. Out of respect for others on the call, we request that participants please only ask 1 question. Operator, can you please repeat your instructions?

[Operator Instructions] And our first question comes from the line of Karl Keirstead with UBS.

Okay. Great. Satya, maybe I'll start away from the numbers and ask if you could spend a minute and elaborate on your opening comments about model choice and the protection of corporate IP. Maybe I could ask this in 2 parts. First, how material do you think traction could be for open and custom models over the next year or 2, knowing that many enterprises might be initially reticent to use open models? And secondly, how exactly does Microsoft benefit from this shift knowing that you've also got fairly large Frontier Lab exposure?

Thank you, Karl. So the way we are coming at this is, at the end of the day, the goal is to have the firm be in control of their own destiny in terms of what I describe as building their human capital and their token capital, right? So at the end of the day, if a firm is a learning machine, they need their own learning machine. And that's sort of really the goal. And the models are an input not some extraction of the knowledge of the enterprise. But in some sense, you have to really -- at the end of the day, every firm is going to evaluate who are the providers who are helping them with their outcomes and their knowledge creation? I think that, that is now fairly clear, and it's going to become clearer by the day. This is not going to be about come in and take all my knowledge and benefit yourself, whereas I am not getting any output.

So given that direction of travel, we are very, very clear about the architectural sort of design of the platform, which is you've got to keep your harness separate from the model. When the harness will ensure that your memory, your context, all of that is external. That means any given model at any given time is swappable. You should and you can use Frontier models. There's no reason not to. But you also can use multiple of them, right? So if you look at some of the stats I gave. It's a great example of how to use the Frontier models for what they deliver, how to use low-cost models for what they deliver, and in fact, train your own model when you don't want to use any external model itself because after all, you have all the output, you have all the traces, you have all the context. That's really the enterprise design architecture that we are going to evangelize.

We ourselves using it. Copilot is built that way. GitHub Copilot is built that way. Our security Copilot is built that way. And we want to democratize that design pattern so that every enterprise can use it. And within that, there will be a mix of open weights, closed weights. And by the way, one of the things that [indiscernible] talked about is, remember, right, if you look even at the [indiscernible] incident, the biggest thing that we should take away from that is you can't sort of depend on any one model. You will maybe need multiple models to even remediate some challenges that get caused by one model, right? That's the way to think about it, right, which is you can't be subject to the refusal of one model.

So there's a lot more design space here. We talk about the Frontier as if it's one thing. The Frontier is about every firm having a Frontier and the choice, the cost control and the capability that they need in order to be able to control the destiny.

And I think maybe, Karl, just to add a little bit to the end of your question, which is that it's why it's important that the platform is built. And I think Satya mentioned this in his comments. To be able to deliver the right model for the right job on the architecture called Azure. And so given that we continue to see growing demand. No matter what model is chosen or what model family or whether it's run a model of your own, the Azure platform is quite efficient at delivering that. So think about that infrastructure as being pretty fungible.

The next question comes from the line of Brent Thill with Jefferies.

Impressive acceleration in Azure up to [ 43 ] going to mid-40s. I guess the questions around the underlying drivers, what you and Satya are seeing in terms of just what's driving this. And many of the questions around capacity constraints. Are we just still in the same environment? Or is this Microsoft just executing better given the constraints we're all seeing?

Thanks, Brent. First, there are still constraints in the system. I think we've continued to say, I think, now for a number of quarters that demand continues to exceed available supply. And that certainly remains true. You can even see it, I think, in some of the pricing that's occurring in the spot market for assets. So when you think about being able to deliver better, the first thing we focus on, and I tried to talk a little bit about it in my prepared remarks is efficiency. Being able to get more out of everything that we've got in the fleet. That applies to efficiency gains in the CPU fleet. It's going to be efficiency gains in the GPU fleet. We saw a good work this quarter in particular from our engineering teams to make as much of that available as we could. And because of the supply-demand imbalance, we've been talking about, when we can make efficiency gains, they are quickly monetized in quarter. And I think that dynamic certainly impacted the quarter positively.

I would also say some of the process improvements we've made to make sure both CPUs and GPUs just the lead time from how quickly we can get things to simplify it tremendously plugged in was also improved over the past 90 days. And so those improvements, again, are very quickly monetized and we're able to do that. And at the scale that we're operating in terms of across the entire hyperscale fleet, making efficiency improvements that can be quickly monetized as a result and acceleration in the quarter, it's part of also what we expect to see and talk about with Q1.

The next question comes from the line of Mark Moerdler with Bernstein Research.

Congratulations on the solid, really great quarter. Satya, Amy, sentiment around AI remains incredibly volatile with concerns about oversupply coming as well as concerns about component pricing increasing impacting margins. Amy, 2 related questions. How does Microsoft protected cell if there really is overcapacity and overbuilding of data centers or overbuilding of chips, et cetera? And on the flip side of that, how do you manage through the hardware price increases that we're seeing, the component prices and that it doesn't just drive you to either massively drive up the price of your offerings or negatively impact your margins?

Thanks, Mark. The questions are a little bit related, but I'll start with maybe the first. Currently, the situation is obviously that demand exceeds available supply in a sort of relatively extreme moment. But when you start to think about over the duration, I try to remind people, a lot of the expense, especially you see it in CapEx, you've seen our CapEx really pivot toward what I would call and do call short-lived assets, which really, right, that CPUs and GPUs that have relatively shorter lead times. And so if the demand environment changes, you just slow down what is, in fact, the largest component, right, and the driver of COGS. The investment into land and data center builds is actually quite flexible, right? It's a smaller percentage of the overall cost structure. And it can be -- timing can be changed on much of that, especially on the build, or you can stagger the timing of the build-out of, as I was saying, some of the GPUs and CPUs that you plan to put in.

And so when you think about being able to manage through that, hyperscalers have been doing that for quite a long time in terms of having the flexibility and the understanding of manage those changes in demand. And the other thing is that's important, Mark, is you just have an incredibly diverse book of business by geo, by segment, by industry. And I feel like when you look even at our backlog or what we added in RPO this quarter, it is from the breadth of really the Microsoft product portfolio as well as our customer portfolio.

So when you have the ability to late bind some of the more expensive components in the short term, you have a big book of business that's flexible. You have a big first-party app business that also uses the capacity that you're building out in addition to your Azure platform. It does allow us to have a lot more flexibility to manage through those.

When it comes to the pricing question, I think that's really impacting everybody equivalently in so many ways. What we've been trying to do, of course, at this point is to just make sure that we're doing the best efficiency work we can so we can continue to give customers great value. We're reminding people that, frankly, the cloud offers tremendous benefits versus having to make these purchases as servers on-prem yourself where the price increases or even more hard for customers. So the cloud still provides a great ROI in those types of situations. And we're adding this capacity, to your point, but a lot of this, obviously, is also being sold in newer contracts, and we're able to have the pricing reflected the keep value where we -- listen, for the long term, you want to have pricing work for customers and for you. And so we're trying to stay focused on that as well.

And if I just add to Amy's comments. I thought and we captured it well. All of us are reading this [ 1873 ] as the book to be read. And so in my mind, I think you've got to get the product ship right. That's sort of a lot of what we are focused on. You have to get the portfolio right, Amy talked about how -- what we're doing, whether it's in Copilot or the super app, bringing all the form factors or all the way to Azure and the agent-first sort of primitives in Azure, you kind of have to really get that portfolio to all come together. The mix of customers is super important. You have to recognize the breadth, the geo mix, the segment mix, the workload mix and you got to really think about all of those when you're even building capacity. And then you've got to run an efficient railroad.

At the end of the day, you do need to -- Amy talked a little bit about even in the last quarter, how we've improved on the efficiency front. It's not something that will just show up at the end. You have to sort of monotonically work at it. And so we are very focused on all of those. And then we know that there will be ups and downs of what is the cycle here. But the secular shift is clear, and we're very bullish about us coming with the right sort of mix of business and the right margin structure and most importantly, with the right value for our customers.

The next question comes from the line of Adam Wood with Morgan Stanley.

Also congrats on a very strong end of the year. I wanted to maybe just ask about M365 Copilot. Obviously, very strong quarter there with over 30 million paid seats and had strong acceleration. Could you just talk a little bit about how you're seeing customers move from pilots to broader deployments there? Is this still a pilot driven motion or are we seeing a lot more broader deployments?

And then when we think about the monetization of the product in terms of additional seats migration to higher-value SKUs like E7 and the consumption, what do you see as the main monetization or the main driver of monetization from here, please?

No, thank you, Adam, for that question. So let me start, and then Amy can add. I think, yes, it starts with, again, that product shape. As you can see, even within the quarter, the product shape has changed pretty dramatically, right? So we now have chat, co-work, autopilot, code all coming to essentially what is going to become this flagship super app that various roles can use it. And if you think about even the usage side, that's a place where, again, lots of interesting data there, right, which is time to usage has drastically come up. What used to be months is days, right? So when -- when a license is bought to usage. The usage intensity itself has gone up significantly. I mean we are talking about usage intensity that's sort of at the same level of what is an everyday communication tool like Outlook or Teams.

The second thing I would say is the overall enterprise wiring, right, so it's not like a tool that's isolated somewhere, but it's wired in, whether it is you brought up E7, right? So it's wired into the governance thesis with Agent 365, so then you have your IT ops, SecOps, SpinOps, all wired in as well as it's all the business processes. So for example, your CRM system, your ERP system, all of them are just skills and plug-ins that go into co-work. So you're able to take that enterprise-wide work workflow and wire it into the super app, right? That increases usage. So it all compounds.

And then the other one is the business model. So we now have this per seat business model, and so we also now have the usage business model, so it's seat plus usage. So we're already seeing the ARPU growth that comes from things like E7. But really, as we deliver more value to customer and customer outcomes at the enterprise level, in fact, if I think about historically Office compared to what Copilot is, is much more narrow. This is the first time where you really have an enterprise-wide tool which has a both per seat and usage-based pricing. So the TAM is much more expansive. We're going to be very, very focused on driving customer value and then expanding with it.

Yes, Adam, and I think I talked a little bit about it in my prepared remarks. But I do think what we've been seeing is, over the course of this year, some of the growth in ARPU was from E5 plus the Copilot license that Satya is talking about. We'll see a little bit more from E7 really has a lot of interesting value in the Agent 365 component, in particular, where Satya is talking about I mean having SecOps and FinOps think about, in general, everyone is going to need both observability of token spend and the manageability of token spend for all business processes. And that is what E7 brings. And so I think that is it was only in market for a part of the quarter, and I think we were quite encouraged by the value customers saw in that SKU.

I think we'll continue to focus on that through the year. And then finally, what Satya is talking about is this building TAM that grows through the year. And as I think about that expensive spending TAM, that's really where we're talking about this usage and consumption growth. And so as more of those experiences get wired in and as IT gets more involved in that process, it will be quite, I think, changing in terms of what people think of the M365 capabilities.

And It would be fun for you, Adam. I think 1 of your colleagues put out an ROIC document. So I took that document to copilot, which is a PDF. And I said, "Build me a new sort of PowerBI dashboard essentially." But here is the thing. It built a rich semantic model that went into my fabric with 1 lake that brought all the data in from the external sources, in fact, was current with all the SEC filings of all the MAG7.And then on top of that, the repo itself is in GitHub. But the artifact is sitting in my copilot as a site. That, to me, is a classic example of an enterprise-wide workflow [indiscernible] a knowledge worker could go create a dashboard, the data engineer can go to fabric and find the artifact. The professional developer can go to the repo and find it in GitHub. And by the way, it's all registered with Agent 365, right? That's a little bit of what Amy is describing as they're coming together of a new way to work even while at the same time, bringing IT, security and manageability of it.

The next question comes from the line of Brad Zelnick with Deutsche Bank.

Satya, appreciating cybersecurity is so core to everything Microsoft does. The playing field shifted recently with the latest Frontier model releases. And this week, you introduced Project Perception. Can you expand on what this moment means for your cyber business explicitly? And also what it means for trust in Microsoft more broadly?

Yes. Thank you for that question. I think you're right about saying that the entire, I would say, the overall physics of how both what is needed in terms of the fiber product and even the cyber operations because at the end of the day, you kind of have to sort of transform yourself on both the products, but also how you operate as a company to protect yourself have changed pretty dramatically. So what we are focused on is, first, again, take the same approach we have taken for knowledge work or coding, which is you've got to start with an intelligence-first, model-forward approach. And so what we launched with Perception is essentially saying, let's sort of really make sure that you have the red team agents that know how to find -- constantly are red teaming and finding the vulnerabilities. Then you have the blue team agent that is constantly going and making sure that you're triaging, and the green team that fix this. So you kind of create your own agentic system that's continuously operating to create the cyber defense unique.

It definitely feeds off of all the signals, right, whether it's the defenders, it's identity and trust signal, the Defender signal, the network signal, the app security signal, all that's sort of helping really do the context so that you can then truly create the protection. The other thing they also said is, especially in cyber, it becomes critical to have that multi-model approach to sort of the first question that was asked. Not just for cost. In fact, we proved with the MDASH data in cyber gym that essentially, you can have [indiscernible] level performance with 50% less cost because of this MAI cyber Flash One. And the reason is because 90% of the tasks are done by the Cyber Flash One model, and 10% of the tax, you still go to the Frontier, right? So this is sort of that mixing of the right model for the right task in what is essentially a pipeline job is a super important characteristic.

And so to us, I think this is an important piece. And the other thing I'd say is from a resilience perspective, right, for whatever reason, if a given model goes away, then you kind of can't be left high and dry. You need to be able to still continue your cyber operations. And that's the other piece. So it's cost and resilience is both an important criteria. And that's what we are trying to build in, whether it's in code, whether it's in cyber, whether it is in knowledge work. And we're very excited about Perception and what it means, quite frankly, for our security business going forward.

And the last question will come from the line of Gabriela Borges with Golman Sachs.

Amy, I'd like to ask you about, and do you give us color on the CapEx side of the equation, you've given us color on the monetization side of the equation. Maybe put those 2 pieces together for us. When you look at and track ROI on the CapEx decisions you're making today, how does that compare to a year ago? And what are some of the levers that you can still pull perhaps on the internal silicon sight, for example, as a driver of incremental monetization going forward?

Thanks, Gabriela. I don't know that -- quite frankly, my math has changed in terms of how I do it over the past year. I would say the way to think about it for me is more the confidence in the TAM expansion, the margin levers that we have in terms of both product improvements than the infrastructure improvements. And we talked about some already on the call today in terms of the levers we have to continue to get efficiencies across both the application part of the stack and then the infra part of the stack. But you're right, we didn't touch on all of the pieces. I think Satya actually commented on a number of them. We still have opportunities, obviously, as we continue to look for the best price performance on silicon, including our investments in first party.

The work, frankly, on model diversification also is a margin improvement opportunity being able to serve the best possible outcome with a more efficient or both efficient in terms of token usage and efficient in terms of cost structure are also margin levers. All of these things contribute, obviously, to your point, of increased confidence in ROIC, frankly, of the dollars that we're investing and continue to invest going forward. And as we think about the mix of the portfolio, being able to have a pretty broad pool across knowledge work, coding, security, then basically the agent layer, I'll call that Agent 365, it's kind of a cheap, but all of that also is an opportunity. And then, of course, what we talked about on the Azure side between model efficiency, silicon and component efficiency, including our investments in 1P solutions there and just the overall efficiency of running it at a hyperscale.

So we have quite a few levers to continue to see improvement that we're focused on. But as Satya mentioned, this is the grind work. This is like every day. You just get a little better, get a little better. And we actually are quite good at that grind and making sure that we can deliver that for customers.

Thanks, Gabriela. That wraps up the Q&A portion of today's earnings call. Thank you for joining us today, and we look forward to speaking with all of you soon.

Thank you very much.

Thank you.

Ladies and gentlemen, this does conclude today's conference. You may disconnect your lines at this time, and enjoy the rest of your day.

Greetings, and welcome to the Microsoft Fiscal Year 2026 Third Quarter Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded. It is now my pleasure to introduce Jonathan Neilson, Vice President of Investor Relations. Please go ahead.

Good afternoon, and thank you for joining us today. On the call with me are Satya Nadella, Chairman and Chief Executive Officer; Amy Hood, Chief Financial Officer; Alice Jolla, Chief Accounting Officer; and Brian DeFoe, Deputy General Counsel and Corporate Secretary.

On the Microsoft Investor Relations website, you can find our earnings press release and financial summary slide deck, which is intended to supplement our prepared remarks during today's call and provides the reconciliation of differences between GAAP and non-GAAP financial measures. More detailed outlook slide will be available on the Microsoft Investor Relations website where we provide outlook commentary on today's call.

On this call, we will discuss certain non-GAAP items. The non-GAAP financial measures provided should not be considered as a substitute for or superior to the measures of financial performance prepared in accordance with GAAP. They are included as additional clarifying items to aid investors in further understanding the company's third quarter performance in addition to the impact these items and events have on the financial results.

All growth comparisons we make on the call today relate to the corresponding period of last year, unless otherwise noted. We will also provide growth rates in constant currency when available as a framework for assessing how our underlying businesses performed, excluding the effect of foreign currency rate fluctuations. Where growth rates are the same in constant currency, we will refer to the growth rate only.

We will post our prepared remarks to our website immediately following the call until the complete transcript is available. Today's call is being webcast live and recorded. If you ask a question, it will be included in our live transmission, in the transcript and in any future use of the recording. You can replay the call and view the transcript on the Microsoft Investor Relations website.

During this call, we will be making forward-looking statements, which are predictions, projections or other statements about future events. These statements are based on current expectations and assumptions that are subject to risks and uncertainties. Actual results could materially differ because of factors discussed in today's earnings press release, in the comments made during this conference call and in the Risk Factors section of our Form 10-K, Forms 10-Q and other reports and filings with the Securities and Exchange Commission. We do not undertake any duty to update any forward-looking statement.

And with that, I'll turn the call over to Satya.

Thank you very much, Jonathan. It was a record third quarter powered by the continued strength of Microsoft Cloud, which exceeded $54 billion in revenue, up 29% year-over-year. Our AI business surpassed $37 billion ARR, up 123%.

We are at the beginning of one of the most consequential platform [ shifts ] that will change the entire tech stack as agents proliferate and become the dominant workload. This will drive TAM expansion and change the value creation equation across the entire economy. To capture this opportunity, we are executing against 2 priorities. First, we are building the world's leading cloud and AI infrastructure for a genetic computing era. Second, we are building high-value agentic systems across core domains such as productivity, coding and security. These 2 layers reinforce each other, and we are focused on driving competitive value and differentiation for customers across each so that they can evalmax their outcomes.

Today, I'll focus my remarks on both priorities, starting with infrastructure. We're optimizing every layer of the tech stack from DC design to silicon to system software, the model architecture as well as its optimization. This is translating into operational gains. We have reduced, [ docked the ] life times for new GPUs in our biggest regions by nearly 20% since the beginning of the year. Our Fairwater data center in Wisconsin came online earlier this month, 6 weeks ahead of schedule, allowing us to recognize revenue earlier. And we delivered a 40% improvement in inference throughput for our most used models across Copilot driven by our software and hardware optimization work.

All up, we added another gigawatt of capacity this quarter and remain on track to double our overall footprint in just 2 years. We are moving aggressively to add capacity aligned to our demand signals we see and we have announced new data center investments across 4 continents. We also continue to modernize our fleet with our first-party innovation alongside the latest from NVIDIA and AMD. Across our fleet, millions of servers are powered by our custom networking security and virtualization silicon, including Azure Boost as well as our first-party CPUs and accelerators. Our Maia 200 AI accelerator, which offers over 30% improved tokens per dollar compared to the latest silicon in our fleet is now live in our Iowa and Arizona data centers. Our Cobalt server CPU is deployed in nearly half of our DC regions running workloads at scale for customers like Databricks, Siemens and Snowflake.

As our largest customers scale their AI deployments, they're increasingly leveraging other services across our platform and choosing to run those workloads on Cobalt, and we are expanding Cobalt supply significantly to meet this demand.

The next [ way ] up from infrastructure is the agent app platform. It starts with model choice. We offer the broadest selection of models of any hyperscaler, so customers can choose the right model for the right workload across OpenAI, Anthropic, open source and more. Over 10,000 customers have used more than one model on Foundry, 5,000 used open source models, and the number who have used Anthropic and OpenAI models increased 2x quarter-over-quarter.

For example, Bayer is using multiple models in Foundry to create its own in-house agent platform with more than 20,000 active monthly users. All up, over 300 customers are on track to process over 1 trillion tokens on Foundry this year, accelerating 30% quarter-over-quarter.

We also remain focused on our first-party model work to differentiate our high-value copilots and agents and reduce COGS. We introduced MAI-Transcribe-1, a state-of-the-art speech-to-texdt model and MAI-Image-2, one of the top image generation models in the world. These models are already powering first-party scenarios like image generation in Bing and PowerPoint, and we are working towards having Transcribe-1-powered transcription in Copilot and Teams. Early signals show 67% increase in GPU efficiency with Transcribe-1 and up to 260% increase in Image-2.

We also brought MAI models to commercial customers like Shutterstock and WPP for the first time through Foundry. And we are innovating on OpenAI IP to drive product evals and lower COGS. Two recent examples of what we have done with multistep retrieval with Work IQ [ and ] Copilot and how reasoning adapts to intent complexity and researcher with much reduced latency and increased accuracy.

The next layer up is all about enterprise data in context across Fabric, Foundry, Microsoft 365 and our [ security graph ], we are building a unified IQ layer for organizational intelligence, thousands of enterprises already are accessing context across these IQ layers. And as AI usage grows, so does the context layer creating a flywheel that continuously improves the grounding, relevance and effectiveness of every agent they use and build making our IQ layers an unmatched context engine for organizational intelligence.

More broadly, our database business accelerated quarter-over-quarter. Cosmos DB alone saw 50% year-over-year revenue growth driven by AI app workloads. We now have 35,000 paid Fabric customers, up 60% year-over-year. And all of the amount of data in Fabric, OneLake data lake increased nearly 4x year-over-year. Over 15,000 customers now use both Foundry and Fabric, up 60% year-over-year as enterprises connect agents to real-time operational analytical and unstructured data that Fabric brings together. And we are very excited about the continued progress with Foundry Agent Service and how customers can now build durable stateful agents that run across time boundaries, orchestrate tools and models and close the loop with evals and improvement over long-running workflow.

Beyond Fabric and Foundry, we are also helping knowledge workers build agents with tools like Copilot Studio. Nearly 90% of the Fortune 500 now have active agents built with our low-code, no-code tools. And we are seeing fast growth of our co-pilot credit consumptive offer, up nearly 2x quarter-over-quarter as customers increasingly extend copilot with custom agents tailered to their workflows.

Finally, with Agent 365, we offer a control plane that extends company's existing governance, identity, security and management frameworks to agents. Tens of thousands of companies are already managing tens of millions of agents in Agent 365, and we expect this momentum to grow significantly as agents will increasingly need tools for identity governance, security and more.

Now let me turn to the high-value agentic systems we ourselves are building on this platform. We are evolving our family of copilots from synchronous assistance to async coworkers that can execute long-running tasks across key domains.

In knowledge work, it was another record quarter for Microsoft 365 Copilot seat ads which increased 250% year-over-year, representing our fastest growth since launch. Quarter-over-quarter, we continue to see acceleration and now have over 20 million Microsoft 365 Copilot paid seats. The number of customers with over 50,000 seats quadrupled year-over-year and Accenture now has over 740,000 seats, our largest Copilot win to date. And Bayer, Johnson & Johnson, Mercedes and Roche all committed to 90,000 or more seats. Copilot is uniquely valuable at work where nearly every task depends on organizational context, work IQ grounds, copilot responses in the full context of an organization, including people, roles, documents and communications, all within the company's security boundary.

The system of work behind Work IQ alone now spans more than 17 exabytes of data growing 35% year-over-year. The liquidity and freshness of that data matters with billions of e-mails, documents, chats, hundreds of millions of Teams meetings and millions of SharePoint sites added each day. And that context is getting even richer as copilot adoption grows, copilot and agent conversations and artifacts that create feedback into Work IQ, making it even more context rich.

We continue to increase the pace of future innovation across Microsoft 365 Copilot introducing over 625 updates over the past year, up 50%. And in Microsoft 365 Copilot, you now have access in chat to multiple models by default with intelligent auto routing in agents with critique and counsel. You can use multiple models together to generate optimal responses. As of last week, Agent Mode is now default experience across Copilot in Word, Excel and PowerPoint. And with Cowork, you now have a new way to delegate and complete work using Copilot. All this innovation is driving record usage intensity across Copilot. We have seen a surge in usage of our first-party agents with monthly active usage up 6x year-to-date. Copilot queries per user were up nearly 20% quarter-over-quarter. To put this momentum in perspective, weekly engagement is now at the same level as Outlook, as more and more users make Copilot a habit.

When it comes to biz apps, we are seeing a new pattern emerge as customers shift from traditional seat model to seats plus consumption. The customer service categories at the forefront of this transformation is nearly 60% of our service customers are already purchasing usage-based credits. For example, HSBC uses prebuilt agents with Dynamics 365 to manage customer inquiries across products, markets, regulatory requirements, reducing issue resolution time by over 30%. And our agentic products in LinkedIn Talent Solutions, which help hirers automate time-consuming tasks like sourcing, screening and drafting messages have already surpassed a $450 million annualized revenue run rate.

When it comes to developers, GitHub itself is seeing unprecedented growth driven by proliferation of agentic coding and we are hard at work to scale and meet this demand. We see this even with GitHub Copilot. Nearly 140,000 organizations now use Github Copilot and enterprise subscribers have nearly tripled year-over-year. The majority of users leverage multiple models. We're also seeing rapid adoption of GitHub Pilot CLI with usage nearly doubling month-over-month. And earlier this week, we announced our move to usage-based pricing model for GitHub Copilot as we align pricing to actual usage and cost.

When it comes to security, the physics of cybersecurity has changed as AI compresses the window between vulnerability and exploitation. To help mitigate risk immediately, we simship Defender protections when updates for AI discovered vulnerabilities are released. And we are on course to productize new multimodal AI-driven scanning harness as well. Already the number of Security Copilot customers increased 2x year-over-year, our data security triage agents alone handled over 2 million unique alerts this quarter, and we are helping customers secure their AI deployments as well. 35 billion Copilot interactions have been audited by Purview to date, up 7x year-over-year.

Finally, when it comes to our consumer business, we are doing the foundational work required to win back fans and strengthen engagement across Windows, Xbox, Bing and Edge. In the near term, we are focused on fundamentals, prioritizing quality and serving our core users better. You see this in the work underway across our consumer products. With Windows, we recently announced performance improvements for lower memory devices, streamline the Windows Update experience and brought back focus to core features and fundamentals that matter most to our customers. And you also see this in Xbox where the team is recommitting to our core fans and players and shaping the future of play. Last week's Game Pass changes are one example of how we are staying responsive to customer feedback. Monthly active Windows devices surpassed 1.6 billion, and over time, Windows value will extend to deliver [ unmeeted ] intelligence at the [ edge ]. Our Edge browser has taken share for 20 consecutive quarters and Bing monthly active users reached $1 billion for the first time. LinkedIn has 1.3 billion members, and we are seeing increased depth of conversation and it's the leading B2B sales and advertising channel for large and small businesses. We set new records for monthly Xbox active users in the quarter as well as game streaming hours. And in Microsoft 365 consumer, we now have nearly 95 million subscribers and early signals show increasing satisfaction as we make Agent Mode the default.

Across everything I've talked about. We're also hard at work changing the way we work. Our North Star remains the same, delivering customer value with highest quality and top-class innovation. And this is what gives me confidence in our ability to shape the next phase of growth for our company and our customers.

With that, let me turn it over to Amy to walk through our financial results and outlook.

Thank you, Satya, and good afternoon, everyone. We delivered results that exceeded expectations across revenue, operating income and earnings per share, driven by strong demand and execution.

As Satya shared, our AI business annual revenue run rate surpassed $37 billion this quarter, growing 123% year-over-year. And we're accelerating our pace of innovation as we execute against the expansive opportunity ahead. This quarter, revenue was $82.9 billion, up 18% and 15% in constant currency. Gross margin dollars increased 16% and 13% in constant currency, while operating income increased 20% and 16% in constant currency. Earnings per share was $4.27, an increase of 21% and 18% in constant currency when adjusted for the impact from our investment in OpenAI. And FX was roughly in line with guidance at the total company level.

Company gross margin percentage was 68% down year-over-year, driven by continued investment in AI infrastructure and growing AI product usage. The impact from these investments was partially offset by ongoing efficiency gains, particularly in Azure and M365 Commercial Cloud. Operating expenses increased 9% and 8% in constant currency, driven by continued investment in AI including R&D compute capacity, talent and data to support product development across the portfolio.

This quarter, growth was impacted by a low prior year comparable, particularly in sales and marketing and G&A expenses. Operating margins increased slightly year-over-year to 46%. Total company headcount declined year-over-year as we focus on building high-performing teams that operate with pace and agility. When adjusted for the impact from our investments in OpenAI, other income and expense was $961 million. Favorability was driven by gains on investments that were partially offset by losses on foreign currency remeasurement.

Capital expenditures were $31.9 billion, down sequentially due to the normal variability from structure buildouts and the timing of delivery of finance leases. And this quarter, roughly 2/3 of our CapEx was for short-lived assets, primarily GPUs and CPUs. The remaining spend was for long-lived assets that will support monetization over the next 15 years and beyond.

This quarter, total finance leases were $4.7 billion and were primarily for large data center sites. And cash paid for PP&E was $30.9 billion, roughly in line with capital expenditures as the impact from finance leases was partially offset by differences between the receipt of goods and payment.

Cash flow from operations was $46.7 billion, up 26%, driven by strong cloud billings and collections, partially offset by an increase in operating lease payments. And free cash flow was $15.8 billion, reflecting higher capital expenditures. And finally, we returned $10.2 billion to shareholders through dividends and share repurchases.

Now to our commercial results. Commercial bookings grew 7% when excluding the impact from OpenAI, driven by consistent execution in our core annuity sales motions. Bookings decreased 4% and 6% in constant currency when including Azure commitments from OpenAI. Commercial remaining performance obligation grew 26%, in line with historic seasonality when excluding OpenAI. RPO increased to $627 billion and was up 99% year-over-year with a weighted average duration of approximately 2.5 years when including OpenAI. Roughly 25% will be recognized in revenue in the next 12 months, up 39% year-over-year. The remaining portion recognized beyond the next 12 months increased 138%.

Microsoft Cloud revenue was $54.5 billion and grew 29% and 25% in constant currency, reflecting strong demand across the Azure platform and our first-party AI applications and services. Microsoft Cloud gross margin percentage was slightly better than expected at 66% and down year-over-year due to continued investments in AI, partially offset by the ongoing efficiency gains noted earlier.

Now to our segment results. Revenue from Productivity and Business Processes was $35 billion and grew 17% and 13% in constant currency. M365 Commercial Cloud revenue increased 19% and 15% in constant currency, ahead of expectations, strong execution and improving product quality drove accelerating M365 Copilot seat adds this quarter, with paid seats now over 20 million. ARPU growth was again led by both E5 and M365 Copilot. And paid M365 Commercial seats grew 6% year-over-year with installed base expansion across all customer segments, though primarily in our small and medium business and frontline worker offerings.

M365 Commercial products revenue increased 1% and decreased 3% in constant currency, down sequentially as Office 2024 transactional purchasing trends continue to normalize as expected. M365 Consumer Cloud revenue increased 33% and 29% in constant currency, again driven by ARPU growth. M365 consumer subscriptions grew 7%. LinkedIn revenue increased 12% and 9% in constant currency, with growth across all lines of business. Dynamics 365 revenue increased 22% and 17% in constant currency with continued share gains and growth across all workloads.

Bookings growth was impacted by weaker renewals as customers balance spend between the traditional per seat and the emerging seats consumption model. Segment gross margin dollars increased 18% and 13% in constant currency, and gross margin percentage increased slightly, again, driven by efficiency gains in M365 Commercial Cloud that were partially offset by continued investments in AI, including the impact of growing adoption and usage of Copilot.

[ I guess ] the low prior year comparable, operating expenses increased 11% and 9% in constant currency, driven by the shared R&D AI investments mentioned earlier as well as higher co-pilot advertising spend. Operating income increased 21% and 14% in constant currency, and operating margins increased year-over-year to 60%.

Next, the Intelligent Cloud segment. Revenue was $34.7 billion and grew 30% and 28% in constant currency. In Azure and other Cloud Services, revenue grew 40% and 39% in constant currency against a prior year that included accelerating growth. Results were ahead of expectations as we delivered capacity earlier in the quarter, enabling increased consumption across both AI and non-AI services. Strong customer demand across workloads, customer segments and geographic regions continues to exceed available capacity.

In our on-premises server business, revenue increased slightly and decreased 3% in constant currency with ongoing customer shift to cloud offerings. Segment gross margin dollars increased 19% and 18% in constant currency. Gross margin percentage decreased year-over-year driven by continued AI investment and increased GitHub Copilot usage, partially offset by ongoing efficiency gains in Azure. Operating expenses increased 9% and 7% in constant currency, driven by the shared R&D AI investment noted earlier. Operating income grew 24% and 23% in constant currency, and operating margins were 40%.

Now to More Personal Computing. Revenue was $13.2 billion and declined 1% and 3% in constant currency. Windows OEM and Devices revenue decreased 2% and 3% in constant currency. Windows OEM increased slightly and was ahead of expectations as OEM and channel partners continued to build inventory given increasing memory prices. Search advertising revenue ex TAC increased 12% and 9% in constant currency, was both driven by higher volume and revenue per search across edge and. And in gaming, revenue decreased 7% and 9% in constant currency. Xbox content and services revenue decreased 5% and 7% in constant currency against the prior year comparable that benefited from strong first-party content performance.

Segment gross margin dollars increased 6% and 4% in constant currency and gross margin percentage increased year-over-year driven by sales mix shift to higher-margin businesses. [ I guess ] the low prior year comparable, operating expenses increased 7% and 6% in constant currency driven by impairment and other related expenses in our gaming business as well as continued investments in shared R&D mentioned earlier that benefits the entire portfolio. Operating income increased 4% and 1% in constant currency, and operating margins increased year-over-year to 28%.

Now moving to our Q4 outlook, which unless specifically noted otherwise, is on a U.S. dollar basis. Based on current rates, we expect FX to increase revenue growth by roughly 1 point in Productivity and Business Processes and More Personal Computing with no meaningful impact to Intelligent Cloud. Overall impact to total revenue is expected to be less than 1 point. And FX should increase COGS growth by roughly 1 point with no impact to operating expense growth.

Starting with our commercial business. In commercial bookings, when adjusted for the impact from OpenAI, we expect healthy growth on a growing expiry base with consistent execution in our core annuity sales motions against a significant prior year comparable. Microsoft Cloud gross margin percentage should be roughly 64%, down year-over-year, driven by continued investments in AI and increased GitHub Copilot usage. Just this week, we announced a business model transition in GitHub Copilot that will align pricing with usage and value that takes effect on June 1 of this year.

Now to segment guidance. In Productivity and Business Processes, we expect revenue of USD 37 billion to USD 37.3 billion, a growth of 12% to 13%. In M365 Commercial Cloud, on an adjusted basis, we expect revenue growth to be between 15% and 16% in constant currency. When normalized for the prior year comparable that benefited from 2 points of in-period revenue recognition. And on a reported basis, we expect revenue growth to be between 13% and 14% in constant currency.

Building on the Copilot momentum we saw in Q3, we expect net paid seat adds to increase sequentially, which will drive continued ARPU growth. M365 Commercial products revenue should grow in the mid-single digits against the prior year that benefited from higher-than-expected Office 2024 transactional purchasing. As a reminder, M365 Commercial products includes components that can be variable due to in-period revenue recognition dynamics. M365 Consumer Cloud revenue growth should be in the low 20% range, down sequentially as we start to lap the benefit from last year's price increase. Growth will again be driven by ARPU and an increase in subscription volume. For LinkedIn, we expect revenue growth of approximately 10%. And in Dynamics 365, we expect revenue growth to be in the low double digits, down sequentially with impact from a strong prior year comparable and the bookings trends noted earlier.

For Intelligent Cloud, we expect revenue of USD 37.95 billion to USD 38.25 billion or growth of 27% to 28%. In Azure, we continue to focus on accelerating the delivery of capacity and increasing fleet efficiencies, and therefore, we expect Q4 revenue growth to be between 39% and 40% in constant currency against a strong prior year comparable that included accelerating growth. Broad and growing customer demand continues to exceed supply, and we continue to balance the incoming supply we can allocate here against our other high ROI priorities, first-party applications, R&D and end-of-life server replacement.

As a reminder, year-over-year Azure growth rates can vary quarter-to-quarter based on capacity, timing and contract mix.

In our on-premise server business, we expect revenue to decline in the mid-single digits with ongoing customer shift to cloud offerings.

In More Personal Computing, we are lapping strong prior year comparables, navigating complex PC market dynamics impacted by memory prices and refocusing on delivering quality and value to consumers. Therefore, we expect revenue to be USD 11.75 billion to USD 12.25 billion. Windows OEM revenue should decline in the high teens with roughly 6 points of impact from a prior year comparable that benefited from Windows 10 end of support, 6 points from inventory levels that we expect to come down for the quarter and 6 points from a lower PC market as prices increase due to memory cost. The range of potential outcomes remains wider than normal. Therefore, Windows OEM and Devices revenue should decline in the mid- to high teens. Search advertising revenue ex TAC growth should be in the high single digits, driven by revenue per search and volume with continued share gains across Bing and Edge. And in Xbox content and services, we expect revenue to decline in the low teens, reflecting a prior year comparable that benefited from strong first-party content as well as the recent price changes for Xbox Game Pass as we focus on delivering more value to gamers. Hardware revenue should decline year-over-year. Therefore, at the total company level, revenue should be between USD 86.7 billion and USD 87.8 billion or growth of 13% to 15%, with accelerating commercial growth, partially offset by our consumer business.

Our Q4 outlook for COGS and operating expenses includes roughly $900 million in onetime costs for the recently announced voluntary retirement program. Therefore, we expect COGS of USD 29.4 billion to USD 29.6 billion or growth of 22% to 23%, including roughly $350 million from the retirement program. And operating expense of USD 19.3 billion to USD 19.4 billion or growth of approximately 7%, including roughly $550 million from the retirement program.

Even as we invested through the year, [ and ] additional capacity to serve the growing AI platform, apps and services demand and inclusive of these onetime costs, we expect full year FY '26 operating margins to be up about 1 point year-over-year. Excluding any impact from our investments in OpenAI, other income and expense is expected to be roughly negative $100 million as interest income will be more than offset by interest expense, which includes the interest payments related to data center, finance leases. And we expect our adjusted Q4 effective tax rate to be approximately 19%.

Next, capital expenditures. We expect CapEx spend to increase to over $40 billion as we continue to bring more capacity online. The sequential increase includes roughly $5 billion from higher component pricing as well as the impact from finance leases, which add variability given the full value is recorded in the period of lease commencement. And we expect the mix of short-lived assets to remain similar to Q3.

For calendar year 2026, we expect to invest roughly $190 billion in capital expenditures, which includes approximately $25 billion from the impact of higher component pricing. We remain confident in the return on these investments, given higher demand signals and increasing product usage as well as the efficiencies we're already driving across the platform. Even with these additional investments and continued efforts to bring GPU, CPU and storage capacity online faster, we expect to remain constrained at least through 2026. Despite these constraints, and the continued need to balance incoming supply. We expect Azure growth to show modest acceleration in the second half of the calendar year compared with the first half.

Now I'd like to share some closing thoughts as we look to next fiscal year. First, we continue to evolve how we operate to increase our pace and agility. And therefore, we expect headcount will decrease year-over-year. Operating expense growth will be in the mid- to high single digits, reflecting ongoing investments in R&D, inclusive of AI investment in compute, data and talent to accelerate product innovation.

Next, as a reminder, we will lap strong prior year comparables impacted by Windows 10 end of support, elevated OEM inventory levels as well as increased office and server transactional purchasing. And finally, we remain focused on delivering a platform that enables customers to build and run AI solutions and on driving innovation in our first-party AI applications and services. And therefore, we expect another year of double-digit revenue and operating income growth in FY '27.

In closing, we are committed to delivering innovation that helps customers create new business value as we enter the final quarter of our fiscal year.

With that, let's go to Q&A, Jonathan.

Thanks, Amy. We'll now move over to Q&A. [Operator Instructions] Operator, can you please repeat your instructions?

[Operator Instructions] And our first question comes from the line of Keith Weiss with Morgan Stanley.

Congratulations on another really solid quarter. The Microsoft 365 Copilot numbers, super impressive, I think, way ahead of most people's expectations. I wanted to ask a broader question on demand. We've strong demand for a while. We see it in our CIO surveys and you guys definitely express it in what you're seeing in your business.

Maybe in the short term, you could talk to us about how that demand translates into commercial bookings and how that might be changing. You mentioned different contracting cycles between seats and consumption that may impact that. And then we also have to think about renewal basis. And then longer term, and maybe this opens it up to Satya, what is supporting this demand over time? Or is that another way, who is paying for all of this? Because while we see excitement for Microsoft in our CIO survey, like our overall IT spending expectations aren't increasing and GDP growth isn't really increasing. So at some point, like how does this get paid for? And you start to see the indications of where those dollars are going to come from?

You want to start...

Why don't I start with the first half of your question, Keith, around how these models impact bookings. And I think it's really important. You're right. You have the normal cyclical things that happen with bookings, it's the exploration [ phase ] or maybe large multiyear Azure commitments to get signed. And that stuff has always had some volatility to it. But I think if you take a step back, which is I think the broader question you're asking, and obviously, I'll let Satya talk to it, too, you're really thinking through as we go through using a model that's historically thought of as a per seat business. And suddenly, if you think about getting worked on and being more productive, it's thinking about being a seat or a worker plus an agent.

And when I think about that model, I start to think about it as a license business plus a consumption business and really applying far more broadly than I think people have thought about that. And so it starts to mean that over time, bookings will actually also look a little different. It will still have that per seat license logic, but it will also have a meter, just like you see in Azure. And it may not all flow through bookings in the same way. You'll just bill for usage. And if that usage has great value to customers, and I'll let Satya talk a little bit about this, then you'll keep spinning and keep using those agents if they're adding direct value or growth to your business. And so I think it's probably healthy to sort of start to think about that transition in a broader way. Well, you may not see it in the short term in bookings. I think if I were to frame how to think about the opportunity, I would probably think about it more in that light.

Yes, I think Amy captured it. I think the basic transformation of call it, say, any per user business of ours, whether it's productivity, coding, security will become a per user and usage business. That's the best way to think about. It's obviously already happening with coding. That's where you see it already perhaps at scale. Some of the business model changes even we made this quarter speak to that.

But it also speaks, I think, to the intensity of usage, right? Because where is these dollars going to come from. At the end of the day, it is going to come from some eval and outcome that a business has, where these agents that are working on behalf of users or with users is creating value. And so that's sort of where it starts, whether it's customer service, whether it's individual productivity, team productivity, a business process. Some costs [ per ] is either decreasing because of the use of agents or some revenue is increasing because of agents because it was able to compress these workflows.

And that's what you broadly start to seeing, right? Even when people talk about Copilot, it's obviously, they use chat, chat with reasoning, they use Cowork, they use Agent Mode inside a Word, Excel, PowerPoint, but it's all been done in the context of some task trajectory. And so when they start seeing that the trash trajectory is compressing the workflow, improving revenue, decreasing costs, that is what's driving usage. So it may not be, by the way, pure seat coverage type of motions like in the past. This is more about getting intense users and intense usage. And that's what we're focused on.

And Keith, maybe just to take a quick second. Just a big thank you to you. It's been a real privilege to work with you over many, many quarters. And just to say, we really appreciated your coverage over this time. And congratulations, [ making use your ] last earnings call.

And thank you so much, Keith. And [indiscernible] you.

The next question comes from the line of Karl Keirstead with UBS.

Okay. Great. Amy, could you elaborate a little bit on the CapEx guidance you just provided. Obviously, it requires a fairly material pickup in CapEx in the second half of the calendar year, maybe to the tune of $120 billion. I'm just curious your confidence in working through the physical component constraints to hit that number does it involve the greater use of partners? And how are you thinking about allocating that increased capacity between third-party and first-party, do you have a general framework you'd advise us to keep in mind?

Sure. Thanks, Karl. No, I actually -- I feel quite good about our ability to work through the physical sort of limitations. I think of the industrial logic of the supply chain, to be able to put that both -- some of that, as we've talked about, is getting capacity online, but a lot of that is far more short term in nature, being able to get CPUs, GPUs, storage put in place to be able to start to support better the demand signals we've been seeing. Trying to give some help on part of that being priced. I think that just helps give you a sense on volumes. And obviously, it leads more to short-term assets when you see that type of impact of price on the number.

I would also say, in terms of a sense of allocation, you should assume -- we talked a little bit about what you were seeing in Azure looking for 39% to 40% in constant currency in Q4 means that we're able to use some efficiencies to make sure we're able to meet demand as we can best do that in a balanced way across Azure, our Copilot usage, which I think you've seen in Q3 has really been on a different trajectory than we saw it up to this point that applies across coding. It applies across productivity, and I have some confidence it's also going to apply across security.

Then if you think about talking about some acceleration into what I would call the first half of calendar -- the first half of FY '27, the second half of the calendar year, it means we're getting some insights to our abilities to increasingly put pressure on efficiencies being able to speed up the deliveries into our data centers and make that what I would call revenue ready as quickly as we can. So I would expect the pressure between first-party usage and being able to meet Azure demand will persist as I said, but we're doing our best to be able to get things in as quickly as we can and hence, the CapEx number that we see in the second half of the year.

The next question comes from the line of Brent Thill with Jefferies.

One of the big pushbacks we all get is that AI is going to be really expensive. Yet you, Google and Amazon are showing higher margins tonight as you report. What are investors missing? And why is AI a potential better margin for the industry over time?

Thanks, Brent. I think -- we've been talking about sort of where this AI business of ours has been in the cycle compared to even the cycle we saw with the cloud, which now seems very long ago. And how margins were actually better and they remained better in our AI business versus where we saw in the cloud transition, looking back. And so I do feel like what we've been really focused on is making sure that the business models reflect how these applications are both getting built and the value that they're bringing. And so when you think about that type of value. It tends to be captured more and consumption and usage-based pricing models. And I think that's something that's probably been a little underappreciated as we look in terms of margins going forward.

I also think it's been important to us to make sure we leverage the IP we have the IP we get from our partnerships is obviously free to us for a long time. So we're able to take that and apply it and to benefit our margins in a healthy way. You've also seen us work hard on the first-party hardware stack being able to make sure we can take margins out of the infra stack as well. And then, of course, just the efficiency work. We've really been in an accelerated phase, as you know, of trying to get as much capacity as we can get into production. But when you go through that, you also can start to focus on the efficiency work. That's efficiency work on the hardware side as well as efficiency work on the software side to be able to deliver these types of margins.

I do believe that one of the real focuses that we've got to all have is that -- and this really dates back to the question that Keith asked in the beginning, which is that when you move to usage-based models, you have to make sure you're delivering incredibly high value to customers. So what we need to do is make sure the focus starts with customer usage that creates value. If that creates value and positive output than the TAM expansion here and the ROI will be very good.

The next question comes from the line of Mark Moerdler with Bernstein Research.

Congratulations on the quarter you delivered and the rate of growth and some of the commentary you've made on guidance, et cetera. What I'd like to drill in a little bit is on this whole question of of the CapEx and the spending that you're making. Obviously, the Commercial Cloud is growing fast. Azure is growing fast. AI is growing even faster within your overall business, but there's a bit of a disconnect that makes investors a bit nervous between how fast they're seeing CapEx growing and how fast they're seeing revenue growing. So can you give some color about how the timing works out, or how much needs to be spent on replacement of equipment or first party in order to build that confidence that as we look towards the strong spending on CapEx, that the core business will continue to be very, very healthy and that the margins will be good.

Thanks, Mark. Let me maybe start with Azure, which given its size and its growth rates where we talked about acceleration from where we are, which is the guide of 39% to 40% into a bigger number in the second half of the calendar year.

When you start to see that type of growth rate on the size of the business we have, the amount of spend being done on short-term assets, which is really the thing that correlates with revenue as opposed to these -- the 1/3 of that number-ish, that's going into 15-year assets or some lumpy timing from lease contracts that can kind of get confusing. I think in so many ways, this just reminds us of the last cycle. And when the TAM is so expensive and when shortages are generally, I think, growing seems to be the sentiment between supply and demand. It just gives you a lot of confidence in the ROI on certainly and starting with the platform side.

I think what you're really asking is whether as we see these usage plus consumption models emerge at the app and services layer, are we starting to see the benefits of that. And I think if you look past the last quarter, I think we saw some acceleration, which I felt good about in the M365 Commercial Cloud number this quarter. We're guiding for that to be better again in Q4. I think that's where you're starting to see, right? I think the thing that investors have been asking and Mark are asking about is we'll start to see that show up in revenue growth. And I think that's the first place you point to. We can also point to it, and I think you'll start to see it in in GitHub, right, where you see revenue growth rates and usage consumption models result in acceleration in the top line. And then in general, I think we continue to see that right?

And so when you think about spending that amount of capital, putting it into production, seeing some delay before it turns into revenue ready, having the book of business, I think it's over $600 billion of revenue that we still need to deliver, and that's before we're starting to see the acceleration in seats that we're seeing on Copilot, I do feel very good about, frankly, that number. And our real focus will be how much of that we could pull in as fast as we can. I just want to be transparent but when you have revenue that's sitting there that can be growing faster or efficiencies, the focus needs to be on doing that and landing this CapEx as quickly as we can and converting it to revenue as quickly as we can.

I mean I'm going to just add one point here, which is I think at some level, one of the things that we have learned even in the last, whatever, 2 years or so in AI and also build more conviction and confidence on is where is the TAM and the category economics of the TAM. And so this, I mean, it's fascinating that here we are in 2026 and the most exciting things are plug-ins in Word or Excel or CLIs in coding or -- and so when you see that, that means we have a structural position in knowledge work, coding, security, which are the big caps. And then you couple that with the right business model, which is what Amy was referencing multiple times, which is user plus usage. And then you take even the book of business we have, that sort of is a true line -- and if anything, we want to make sure we are getting the CapEx to get the capacity in time for those increases in usage, which I think is going to be very, very key.

And you got to remember the model capabilities are exponential. So we [ should ] think about even Agent Mode in Excel, it sort of kind of didn't work until it started working. And that's just because the model showed up. And so you have to be ready for those opportunities.

That's extremely helpful. I really do appreciate it. And again, congratulations on the quarter.

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

Satya, I would love to hear some of your reflections on the Copilot given the technical and commercial milestones that Microsoft just hit just in the last 3 months, so maybe share with us a little on your learnings from Copilot adoption to date. What do you think is working? What's not working? How is that now informing your E7 strategy and your Copilot Cowork strategy?

Thank you for the question. So I think the way to perhaps think about even Copilot, is the Microsoft 365 Copilot and knowledge work in some sense this pattern. We've learned a lot, as I said, even from coding. But if you sort of focus it on M365 Copilot, the first thing is to think about even the form factor and the shape of the product and how it's evolved, right? So there's chat, chat now with reasoning [ over ] Work IQ. So that's sort of one form factor. Then there is all the agents like researcher and analysts that you use within chat, or even custom agents that our customers are building. And then on top of that, you now have this [ edit more ], right? So if you think about it typical trajectory or a session in Copilot as it starts with chat, you do -- you ask some questions, you get some insights, you ask it to even generate an artifact, you open that artifact in Word, Excel, PowerPoint, you further refine it. So in other words, you continue the conversation.

And then, of course, we now have even a complete new form factor where you essentially delegate the task, right? So you're not even interactively working but you're delegating the task with Cowork. So these are all the various form factors. And one of the most interesting things to keep in mind is the usage of this is at the same level as Outlook. So this is not -- even to the previous question, are people using it, finding it useful. I mean this is like a daily habit of intense usage.

The other thing that is important when you think about what makes these form factors useful is intelligence and the intelligence is a function of 2 things. It's an intelligence of having multiple models coupled to context, that's the meetings, the documents, the e-mails, the teams, I mean all of that SharePoint data, all of that rich by the way, constantly updated, right? So this is not some static database. It's the most important database in any company that is constantly changing every second. So that's the context, the models brought together with the harness that's multi-model. This is essentially the same thing we do in GitHub, whether we do it in M365, we do this in security, which is our core goal is to decouple the harness from the models and then have the context richness show through because customers are going to use multiple models.

In fact, if you look at critique or console, that's a great example, or Rubber Duck in GitHub Copilot. These are good examples of why you want or even in Excel, I generate using Opus and I check with Codex. That's the type of things that you want users to have access to. And so that's really the -- and then couple that with the business model of user usage plus user pricing, I feel like that's what's happening, and we are seeing that all play out.

The next question comes from the line of Kirk Materne with Evercore ISI.

Amy, I was wondering, can you just talk a little bit about the change in the OpenAI agreement, if there's anything we should be aware of from a modeling perspective or from a financial perspective, that would would change today versus where we were maybe a couple of weeks ago? And then I guess, Satya, for you, it seems like an opportunity for you guys to continue to diversify from a model perspective. Any other takeaways we should be thinking in terms of where you guys landed with OpenAI and new framework.

Yes. Maybe I'll start. I mean, overall, we feel good about partnership with OpenAI, I'm always [ fairway ] focused on any partnership and ensuring that there's a win-win construct at all times. I mean that's how you can remain with partners. In this case, it starts with, quite frankly, IP, Amy referenced this. We have a frontier model. royalty-free with all the IP rights that we will have access to all the way to 32 and we fully plan to exploit it. And there are examples I talked about [ in you ] in my remarks earlier, and that's we are thankful for that, and that's sort of one part of the agreement.

The second part, of course, is the them as a customer of ours, they're a large customer of ours, not just on the AI accelerator side, but also on all the other compute side, and so we want to serve them well. And then, of course, we have our equity. And so overall, I think the construct as they have grown and we have grown and our customers also have different expectations in terms of their model diversity. So therefore, we've all revolved the partnership, but I feel very good about where we are.

Yes. I think the only maybe 2 things to keep in mind. I would say is having the revenue share exists through 2030 and the predictability of that is a real positive for us. And then as you put out, that Satya pointed out, the thinking about that as royalty free with the elimination of our rev share to them.

Thanks, Kirk. Operator, we have time for one last question.

The last question will come from the line of Rishi Jaluria with RBC Capital Markets.

I wanted to go back the discussion we've been having today on seat-based models and consumption. And maybe kind of the philosophy for how this changes over time. We'll get a complete agreement with what you're seeing out there and totally makes sense. Maybe I want to drill into -- you announced E7, which will come out. That is predominantly seat-based with some consumption components. You're doubling down on that seat element. And it seems increasingly customers still want the predictability of seat-based models, as we see with all the kind of usage issues that companies are running to as AI has kind of gone out of control. Can you maybe understand how to bring all these pieces together how to maintain predictability within the customer base while increasingly growing consumption, and maybe if we were to fast forward 3, 5 years in the future, what we -- how should we be thinking about that? What that mix of consumption versus traditional seat base looks like?

Yes. I mean at a high level, maybe we should add to it. But I think you said it, which is, customers want predictability, especially for budgets and procurement. And the seat-based pricing is just entitlements to some consumption right. And so that's, I think, the way to think about it, which is there is some base usage rights that get bundled in or packaged in to see it's a convenient way for people to buy some essentially consumption packs that happen to be assigned to see agents.

And then beyond a certain level, there's overages go into pure consumption. And even there, if you have commitments, long-term commitments to consumption, you get discounting that is appropriate with it. So I feel like that's the direction of travel. And then -- the other thing you mentioned is how is this going to -- from a customer perspective, they're more evaluated by evals. Where are they seeing the value of tokens, as simple as that. So where they see the outcome, the eval and the token, whether it's improving revenue, improving efficiency, and that's what will define.

Like when we talk about IT budgets, IT budgets are going to have to be reshaped by a combination of business outcomes, making their way into into IT budgets and maybe reallocation from other line items on the income statement like OpEx.

Thanks, Rishi. That wraps up the Q&A portion of today's earnings call. Thank you for joining us today, and we look forward to speaking with all of you soon.

Thank you.

Thank you.

This concludes today's conference. You may disconnect your lines at this time, and enjoy the rest of your day.

Excellent. Thank you, everyone, for joining. I'm Keith Weiss, I run the software equity research franchise in the U.S. here at Morgan Stanley. And really thrilled to have with us from Microsoft, Satya Nadella, Chairman and CEO. Satya, thanks so much for joining us.

Thank you so much. It's thrilled to be here.

Excellent. So it's really an amazing time to be talking to you about what's going on in software, partly because of how much innovation is going on in software right now, partly because of how much uncertainty there is in this room amongst investors, and we're seeing it in the stock market. And I'd say over the past several weeks, we've seen even renewed pressure, even more uncertainty about real fundamental questions on how software is going to be built, how it's going to be delivered, how it's going to be priced on a go-forward basis. So maybe we could start there and get your perspective. What's your view on how software is fundamentally going to change over the next 3 years, over the next 5 years?

Yes. I mean it's fascinating, right? Here we are in 2026. And you sit around and you say, God, there are start-ups that are getting founded and funded in massive amounts. If you're an Excel plug-in, you are a VS Code fork, people are in love with file systems again. I mean yesterday, there was a tweet which I loved, which is somebody said, oh, instead of just using grep, I can even do an index now. And I mean like this is mind-blowing in the sense that something is happening where we are rediscovering some of the oldest things we had, like CLIs and IDEs and Excel plug-ins. So it's a good question. Like what has changed? I think to me, I go back to what is this AI thing all about, right? It's a pretty powerful capability we now have to essentially predict trajectories, right?

That's what AI does. AI sort of takes what is, call it, knowledge work or a consumer sort of task, right? It could be a shopping task or a research task or I'm trying to create a sophisticated Excel model or I'm trying to do a forecast. Those are just basically knowledge work, their trajectories and AI is just great at predicting them. So a couple of things are happening as you put these models as they get better and better into our work and into our workflow. The first thing that's happening is the velocity is going up. If you think about -- there's all the stats, right, which is Dario, I think, was here last night, and he talked about sort of what's happening in coding. And you see that even what he's seeing in -- with Claude, we see it in GitHub in the sense the repos are sort of skyrocketing, right, where the public repos have what, 4%, 5% are all basically generated by these coding agents. And the same equivalent is happening in office. I mean it's crazy. I mean our SharePoint, OneDrive, I mean, the artifact creation has become the thing that everybody does.

And because it's cheaper, it's so much easier to create sophisticated. And it's not just bad artifacts. These are sophisticated artifacts that are getting created as part of a trajectory, as part of a task, as part of a workflow. And so in fact, last night, I was playing with this where I wanted to track all start-ups and said this is funding in start-ups. So I put up -- we have tasks now in Copilot that's in research preview. Think of it as our coworker. And it created this fantastic spreadsheet, right? It went off, did a bunch of work.

And the spreadsheet -- the problem -- so what it did is it generated an unbelievable spreadsheet at the end of a long-running sort of horizon task. Then it was great because I handed my cognitive load to it, but then it transferred it all the way back to me by creating an artifact and said, now go, here is the output, go make sense of it. So this back and forth between human and AI is going to be the next thing, which is AI is going to be great at doing cognitive work because it offloads cognitive work. But then it also is great, like a GitHub repo that was sort of completely white coded, what the heck are you going to do with it?

I mean you got to then go again, apply AI to understand it. So the same thing is happening even in office and in knowledge work. So what I did, in fact, was go into Excel agent mode and use it essentially as a new linter, new reasoner, a new profiler because that's kind of what Excel does, right, which is it allows you to make sense of -- sort of my tool to reason over sophisticated outputs. In the past, only humans created it. Now humans and AI create it. And so that is sort of what's happening. And the one other thing, Keith, to your point, which is also fascinating, the network effects of intelligence is how I think of it.

Think about -- I'm sitting in front of my CLI and GitHub and coding away and I say, God, we have all these design meetings that are going on. I updated -- I know one of my colleagues updated sort of a spec. It's in SharePoint. I want to make sure that the code that is in my repo is in sync with all of that, right? Literally, meetings and some document. I can go to Work IQ, which is the database underneath M365, right?

In fact, I always said that's the most important database in the company that is least used as a database because they use it as just a transactional store. But now you can. I can load it up as an MCP server and GitHub. And I can say, please go check all my meeting notes, transcripts and SharePoint documents and make sure that the repo is in sync. And literally, it goes off and creates a plan and then starts executing a plan.

That unlock of intelligence or the network effects of intelligence. In fact, I'd like to say that the next office, which is bigger than all of the office that came before, may be headless. And that is sort of going to be the TAM expansive part for us.

Right. Let's dig into that TAM expansive part and maybe focusing on Microsoft 365 and that franchise. It's always been a core franchise for Microsoft. You're talking about the increase in volume in Artifacts and GitHub. One of the other things we've seen is an increase in subscriptions. I think over 2025, we're over a 20% increase in subscriptions. So the users are increasing as well. And Microsoft, particularly Microsoft 365 has always been about enabling those users. So when you're thinking about this Agentic future, is it more users? Or is it just more agents? And how does Microsoft make sure that they stay in front of that curve?

Yes. I mean I think -- remember, subscriptions are just entitlements because it's easier to budget, right? Like why doesn't everybody just do consumption? The reason why everybody doesn't do consumption is sometimes it's easier to forecast. In fact, we are running into this now, right? With coding agents, we're running into it, which is because you kind of have this effect of sort of -- there are people who are using lots of tokens. There are people who are using what I'll call normal sort of volume of tokens.

And so that by modality or the distribution is not yet stable. And so I think what's going to happen there, first, even for just users because there's -- if you think about there are 3 types of modes, right? There's chat, let's call it, what we will call tasks or cowork, right? It's delegated access. So I give something else, which is a synchronous, my credentials and say, please work on my behalf. So that's the second thing.

And the third would be a full digital worker, right? So which is a -- it will have its own identity. It will have its own tools, it will have its own desktop, whatever, right? That sort of is the third thing. And so if you take all these 3, I think the combination of subscriptions with some limits of usage plus a meter is where I think we will end up. And it doesn't matter if it's a person or an agent. And I think the first place where this is happening already with the business models is in coding. And I think information work also will happen. So from a TAM expansive perspective, for us, I look at all agents as users and with maybe more flexibility on how people license that.

Got it. Got it. And if we think about sort of the conversation of evolutionary versus revolutionary, when I look at a lot of the dynamics of Microsoft 365 Copilot, but also what we're seeing more broadly in the environment and what the AIs are trying to do with their generic computing vision, it feels to me like an extension of what we've been talking about for a decade with Power Automate and Power Flow and Power BI of just bringing more visibility in terms of your systems, your data, who you're communicating with into one place. So do you see Copilot really emerging as a new AI-powered user interface for all that, for the systems, for the data, for your teammates for your communication?

Yes. I mean the interesting thing on this design side as to what happens to UI in general or what have you is -- the dream was always the case that can you have some form of a universal interface to all of computing. I mean that's kind of how operating systems evolved. The browser came next and so on. And so this -- the power of natural language and the access to all of the tools with an agentic loop, I think, makes it very powerful.

The question is, if you look at even a model company, right, it sort of -- they started with chat and now there's multiple apps. So everybody has to -- this is where the power of UI, right? I mean what did OpenAI do with Codex? They have a Codex app. and makes sense, right? Because when you have something that is so strong in the sense of usage, then having a heads down usage app for that makes sense. But the workflow between this is what you have to manage. Like for example, even in the thing that I posted this morning was I kind of generate -- I assigned a task, it generated the spreadsheet.

I actually use the spreadsheet right on Canvas inside of "task interface, right, inside of Copilot. I even have fantastic edit capability. By the way, this has been going on forever, right, which is how does Google Docs work versus Office? I mean it has an IR, an intermediate format. So basically, you kind of take office docs and you have an intermediate format in which you manipulate. So all these things will have some intermediate format in which you will work.

And then ultimately, you can export out back to the full UI if you want to, so that you can go deep. And the other big thing is sharing, right? One thing, a lot of AI still is not yet fully cracked. In fact, you see this in coding. We've not fully cracked multiuser. So the next phase is inside of Teams for us, how do you crack agent to agent to human inside a channel. And that's where UI will matter because even if AIs are all doing their work, I as a human need to reason about it. I need the UI that I use. And that, I think, is where it will all come down to.

Oh, by the way, the one other thing I should mention is we are looking at -- like I was looking at why are we spending so many tokens on all this, right? Sometimes it's kind of the most token inefficient things, right? The smart thing that I think everybody is going to start doing in the next year is tools use. That's sort of all software. And that is the way you make AI token efficient. So don't think of this as like tools. Think of it as even if you're all AI pilled, go and say, oh my God, I need a lot more software to make my AI efficient. That's another way this new world will be born.

Got it. Can we talk about sort of with Microsoft 365 Copilot and the -- getting this into your customer base. You have 450 million information workers that are using the suite. Investors got concerned by some demos that we saw from Claude with Claude Cowork. And for the first time in really the last, I would say, definitely 5 years, maybe a decade, there's more concern about the positioning that Microsoft has with Copilot.

But then on the other side of the equation, when I look at my CIO surveys, 80% of the CIOs say that they're either using Copilot today or expect to use in the next 12 months. So CIOs are seeing something in Copilot that I think investors are missing. Can you help us bridge that gap? Like what are the CIOs looking for from Copilot that investors aren't seeing?

Yes. I mean I think, first of all, it's great to see, again, this is where the model capability and also these form factors, right? The -- and this happened even in coding, right, which is get up Copilot, pave the way. And then now we have, quite frankly, a lot of good competition. But the market is kind of massive, right? It's no longer the tools business that I grew up with for the last, whatever, 40 years, and this is a very different market, which is 100x what we were competing for.

And so the form factor innovation that's happening, right, whether it's Chat or Now Cowork or Worker, that's sort of the direction. So for us, we want to do what we have always done, right, which is what we did even in the previous era, which is we have our permission in especially commercial customers through IT. And so we want to love our IT customers because they are the ones who care about lots of things. They care about security, they care about compliance. They care about all of the observability, right?

I mean the reality is come on. I mean I can't launch open claw as Microsoft. I mean, it just wouldn't work. I don't have permission to do that because that would be considered Microsoft launching a virus. I mean that's just not a thing. And so what we have to do -- but at the same time, it's a fantastic innovation, right? That's where -- I mean, the fact is think about like what the last 15-year journey in enterprise was. They didn't want local file systems. And this is about exposing a full-on local file system.

I mean, Morgan Stanley, if I go to [ 10 ] and say, hey, I have news for you, like please I launch something that where I want all the state and the local file system. I mean you'll keep me out of the room. So the thing that we have to be mindful of is really know where our permission comes from and also know what the expectations of our customers are and do a fantastic job of putting the form factors that people love. That's kind of one -- that's the path we are on.

And if you -- to your point, what's happened is the unlock of Work IQ, which is, again, the database underneath M365. So you were showing me the one note. I mean, think about it, right? I prepped for a meeting. I just go in and say, hey, I'm meeting Keith, can you prep my meeting brief, right? It will literally know everything from 2014 to now, every meeting I've had with you, everything you said and which our folks noted and reason over it, not like a search, right? It's a state full agent.

In fact, you can think of WorkIQ as our frontier model, right? It's basically the data plus the model embedded together. That's what the CIOs see the value, right? It's sort of compounding. When I say network effects of intelligence, that's a massive value for them, which is they've invested for 15 years on all of this, and now you can even compound the value.

Got it. It's a good segue into talking about the model layer. And there's definitely a lot of debate in the industry about the AI model layer and whether it's going to concentrate to a small number of frontier models that are just going to get bigger and better over time or whether we should be thinking about a broader set of models, more SaaS-specific models, to open source models. So how do you think about that trade-off between is this market going to consolidate? Or should we be thinking about an ecosystem of models?

I think there are a couple of things. One is, I think at this point, it's fairly clear industry structure-wise when it comes to the narrow way we talk about frontier models, like the American ones are closed, a lot of the Chinese are open. That, I think, is going to be multi-model, right? So there is not going to be a company or a product company other than the model companies, of course, are going to be homogenous, right?

Any one of them -- but the reality is everyone sort of cotton on to look, if you're building any product, whether it's for coding agents or for knowledge work or wealth management or whatever, you want to access multiple models, right? That's going to be the case. And that means there's lots of very careful design that you want to do, which is you want to have the harness not get coupled with the model layer. So people are getting pretty sophisticated in making sure that the harness layer is decoupled.

The other is the context layer also should be decoupled, right? You don't want effectively the one model to, in fact, vertically integrate into these 2. And that will be the game that will be played. But I'm pretty clear industry structure-wise, where we will go. Then the other question in the -- by the way, the other thing is people also will use a frontier model, like, let's say, there's something -- you have an eval, like everything is all about evals, right? So you have an eval, you want to max your eval, you use all these models.

Let's say, one particular model was doing well on that eval. The next thing you do is you take all your traces and then you train the model that's the cheapest to be better and hill climb on that eval. And so that's what everybody does, right? So it's no one wants to sort of be inefficient in their token spend, whether it's a company app or whether it is an ISV. And so the game would always be about continuous optimization of these models. And then the next thing that also -- you asked the question, how many models should be there, right? I think there should be as many models as companies because if you think about it, the real sovereignty question in the world, it's not about even country sovereignty.

It's about -- look, if really knowledge work, tacit knowledge of any enterprise of a firm is in the human capital they have, then it also needs to be in some model weights that they control. Otherwise, there's just going to be enterprise value transfer. Forget the software companies. It's like the rest of the economy. I mean that's where that [ Citrini ] thing sort of goes off on, which is that is what -- like -- and everybody is going to wake up.

I mean I just don't think we live in a real world and a real political economy. And so therefore, people are going to say, oh, no, that's not going to happen in my house. I'm going to have my model that is sovereign and independent of any frontier model. And that's the other sort of direction of travel.

Got it. Can we talk about the model that Microsoft is building. Of course, you have a great relationship with OpenAI. You have research IP, right? So you get access to their models. But Mustafa is also working on models within Microsoft. Is there a differentiation between what Mustafa is working on versus what OpenAI is working on?

Yes. So we have 2 sort of -- we have the OpenAI models and the OpenAI IP. And so we have -- essentially, if you look inside of Copilot, whether it's GitHub or M365, we have lots of models. It's not a single model. The one model would be a midtrained OpenAI lineage or an MAI model. And so we basically are continuing to optimize these models. And what are we optimizing for our evals, right? So all of us now at this point, the game is about private evals because we know the benchmarks are all saturated.

So the question is, how do you really make sure that any flops we assign to any research compute, right, whether it's post training of an OpenAI model or even our own model is all about maxing the evals that are private to -- so that means some product impact. That's the job #1. The second piece inside of that is all about COGS reduction, right, which is the auto mode, right? One of the things that I think this week or last week, there was a nice paper by the vLLM guys, the semantic router, right? If you remember when GPT5 came out, they had a router. And now even the sophistication of the routers are getting good, right? That means the models that you choose, the modes of the models you choose, the tools that you choose with it.

That's kind of -- that's -- think of it as a meta model, right? That's the most important thing. So our auto mode inside a Copilot, M365 Copilot, GitHub Copilot, is essentially doing that optimization of maxing the evals, reducing the COGS on a continuous basis. And so our frontier work, our super intelligence work will all be about really getting good and hill climbing on it. So the great news for us is we have access for whatever, 5 more years or what have you until '32. So therefore, we want to max the value of what the OpenAI R&D is while also making sure we have all the capability.

And it's no different than, by the way, what we are doing in silicon. I think you have Jensen here next, and we work deeply with him there, which he's doing lots of the innovation. And then we are also working on Maia and so a similar approach.

Got it. Got it. If we talk about the OpenAI relationship? It started 2018, very early before we were talking about any large language models or generative AI, but it's evolved a lot since then. What's the current status of that relationship? And how should we expect that to evolve on a go-forward basis? It's a very important relationship for Microsoft.

Yes. I mean I think -- yes, I mean, going back to 2018, I don't think any of us thought of -- we'll be here in 2026 talking about what is it, like close to $1 trillion company, right? I mean, that OpenAI is. And so we are thrilled to have invested in them, partnered with them. And although none of those things were sort of in my head at that time, who would have thought that I'll sort of earn the credit of being a good investor, but here I am. That's a great investment.

But the thing that was foundational to it, it's fascinating. I always think back at it. The paper I read, which, by the way, was authored, one of the authors was Dario, when he was at OpenAI on scaling laws. And that was the bet, which is when they all sort of sort of have the conviction, Ilya, Dario, Sam, all of these chaps were really into, hey, look, the scaling stuff works and we should go at it.

They showed me those charts, and I kind of understood it, but then I sort of said, now, these chaps are really good at this and they -- and I was always obsessed about NLP. So therefore, I wanted to take that bet. And so it's an important partnership. It's worked very well for them. It's worked very well for us. We also know that they're now a big company that they need some flexibility, and we've given them the flexibility. But they're also going to be a customer.

We're an investor. We are going to be a partner. And they're also -- even the latest partnerships they have are all fine by us because at some level, our exclusivities, our rev share, our equity partnership is all sort of what's going to be healthier because of all this.

For me, one of the things that has marked your tenure as CEO at Microsoft is really embracing openness and open ecosystems, right? I think one of the foundational moments for me is when you came on stage and said, we're going to treat Linux as a first-class citizen on Azure. It almost feels like OpenAI and OpenAI relationship is similar, right, that it behooves both parties to have more flexibility, more openness to let them grow, you get the IP, you get the revenue share, and you both benefit from a more open relationship.

Yes. I mean I think since this is an investor conference and not a feel good conference, right, the thing that I go back to is there are very few zero-sum battles. And I think we overstate that a lot. And so when the big strategic mistakes are made, when you're not clear about, wow, what could be an expansive way for me to think about my opportunity versus getting narrowly stuck on everything being a zero-sum battle.

In fact, our biggest mistakes, I would say, strategically would have been historically made when we didn't view -- like somebody else's success doesn't need to be a failure if you can write it. And it's sort of a thing that needs to be talked about more. And in fact, it is, like without Intel, I don't know if Windows would have happened, right? Without -- in fact, without Mac, I wonder whether Office would have happened, right? I mean that's sort of the world I come from, and I'm always looking for, first, what's the non-zero sum where we can add value to our customers. And then, of course, there are zero-sum battles and we'll compete.

Excellent. So if we talk about vertical integration. You talked about it a little bit in terms of needing to maintain flexibility with the harnesses and the context layers within the model. But there's also some really great innovation going on from Google, where they're integrating the TPU layer with the model layer. Is that something that Microsoft has to pursue more aggressively?

Because I know you guys do have your own silicon initiative. How important it is to have that strong integration between sort of the silicon layer and the evolving software layers on top of it?

It's -- I think the way I think about all of this is if you take the -- like the broad perhaps investor question on this, which is what's the best way to think about ROIC on capital spend? I would say there are a couple of different dimensions to it, right? One is you want TCO to be great. You can't be upside down on cost. Now my belief is, since this is a multi sort of generation thing, it's not about any one time, right? So you really want to make sure you are always going to have the best TCO.

And so that's why my fundamental bet is you want to have system software over a heterogeneous infrastructure. That way, then the latest innovation, right? I'm sure Jensen will come talk about sort of some very cool stuff he's working on next and let's call it, on some low latency inference. God, I want that, right? Why would I not -- like -- and maybe I'm also working on it, but it may be 2 years from now or 1 year.

But no, I want the best always in my fleet. And then I want my system software laid out on top of right? So that's kind of where I like our thing about, hey, we have a Maia 300, 200 today that does competitively, right? It's, in fact, marked on TCO per dollar -- performance per dollar, it's leading, in fact. But that's leading today that is not a guarantee that it's leading all the time. So therefore, I shouldn't get too high on my own supply here just because I have this means that's the end of the game.

No, I'm on that. And then let me then sort of make sure I'm also getting everybody else's innovation in, use software to then schedule across all of this and manage TCO by generation. Then the next layer, of course, is utilization, right? That's again a software game, really max the utility on it. Then make sure you have a diverse demand, right? We have 1P and 3P, even 3P. We have OpenAI book, we have Anthropic book, but we want to also have the long tail of enterprise IT. And so all 3 of these, I think, are more important than just like, oh, I have a model and sort of an accelerator, like TCO and cost economics is -- you got to have diverse customers, great utilization, multi-generation TCO curve. So it's a lot more sophisticated than just one simple thing.

Got it. You're talking about unit economics and the ROI of the CapEx spend. I think it's equally as surprised as we would have been in 2018 to be talking about how big OpenAI is. I think we'd be equally surprised to see how much we're spending on CapEx today, how much infrastructure we're building out in terms of data center. And it has caused concern amongst investors, does this fundamentally change the economics for Microsoft? Does this fundamentally make it a more capital-intensive business that's going to bring down gross margins? How do you see that equation playing out in the near term and then longer term?

Yes. I mean I think the -- I mean that capital intensity has been true even in the Claude era. And that, by the way, is the other thing that is probably worth pulling the thread on, right? We talk a lot about AI accelerators, which are obviously super important. But these agents are really driving up, guess what, regular compute uses, right? We like the innovation on the micro VMs and the containers and the security boundaries on this and storage requirements.

So basically, it's a full-on system upgrade of everything, the network, the compute, the storage, this AI accelerator. And that's all obviously all exploding. And so therefore, the capital intensity increases. So I fundamentally think that this is a capital-intense business, but it's a knowledge-intense business as well, right? So that's a software business. So that's sort of the way I think that this industry is probably unique in the sense that the -- you can't think of it as just a pure industrial. This is where interestingly enough, even on the schedule of how profit comes, right, you think about it, it's not like we have retired any GPUs yet.

Even though we have all the power constraints and so on, just because we have gotten our software so good and some of the workloads optimized, so good on that, that we are able to take the life of these things and really max them out. And so overall, I would say, yes, we have to manage a capital-intensive business, but using all the levers software gives us in managing TCO, managing utilization, optimizing the kernels by workload, ensuring that there's a diverse class of customers, using the diurnal curve even of utilization to create new SKUs to clear markets. I mean those are all things that I think will generate great ROIC.

And this is probably unique. I know the way we have set it up is you track the Azure KPI. The thing that I keep complaining to Amy always is like, hey, I love the Azure KPI, but I don't allocate my capital to Azure KPI, right? I allocate my capital to the Microsoft KPI because I have these -- all these other businesses that are using the same capital. So my 1P, and it has a different gross margin profile. So that's kind of at least how we will, between Amy and me, will take a look at the long-term ROIC and margin profile of the business.

And the margin dollars, not even just the margin percentage because even if it's a capital-intensive business, and we have the ability to use software to generate margin dollars, this will be a much bigger business than we ever had. I mean that was the thing. If you remember, Keith, it's funny that you've tracked us all through, right? I remember when I became CEO, everybody said, oh my God, isn't it too late man, like why bother to even build a public cloud because Amazon is so far ahead. And we knew it was going to be multiplayer. We knew that there is going to be margin, and we kept building. And so that is, I think, what -- and the reason why that was the case is it's software.

Got it. You touched on the allocations and capacity constraints, definitely a hot button topic for investors, the need to allocate towards R&D and first party versus just Azure. I was having dinner with [ Johnson Nielsen ] last night and said, the real solve to all of this is more capacity. It's kind of the first problem -- first-class problem to have of you have too much demand versus your capacity. How should we think about the time frame for that settling out for the capacity coming online that you could feed all of the mouths within Microsoft?

Yes. I think -- look, I think the challenge is what we are seeing is not just linear growth in some of the demand. And so therefore, I think the capacity challenge -- and it's also not any one thing, right? It's not just power, it's sort of silicon, it's wafers, it's substrates, it's memory, it's storage, it's everything. It's compute. And so I feel like it's going to take a lot more upstream capacity to get built, quite frankly, before we have some relief.

So this is where -- this is there to stay. Then the question is what you said, which is how do we make sure we allocate things here that are strategic. And that's where I think the tension is, right, which is -- and I get it, which is you would like -- and not all of you uniformly, but some of you would like that allocation to work in ways that just maximize the short term. That makes no strategic sense for Microsoft, right, or Microsoft investors who are long term because you want to be careful, right?

You don't want to sacrifice your own 1P, which is going to be higher margin by just not feeding that product. I mean that's not a sensible thing. Even like when we say, well, we're going to allocate to our R&D, as I said, we are not allocating to our R&D in ways to just sort of win some contest. No, I want to allocate my R&D to create a model that's going to COGS reduce, which is, again, sort of margins for investors long term, right?

So therefore, that is the asymmetry we will have, right, between what decisions I'll make versus what you may want me to make in the short term, given the supply constraint. And so that's where we will just have to be very disciplined and by the way, one other thing that I'll mention is we also want to be careful about pricing, which is -- look, we see this even in our supply side, which is people are going to take different strategies. Some are going to be very long-term oriented.

Some will be short-term oriented. And we'll remember that, by the way. And the thing that we want to be is long-term oriented even with our customers because these cycles will go -- come and go, and we don't want -- we want to earn trust in periods like this with customers who depend on us. And so these are all fairly important sort of criteria for us.

Outstanding. Unfortunately, that takes us to the end of our time, but really exciting times in software, really exciting times for Microsoft. feels like you guys are really positioned well for this opportunity. So thank you for coming.

And thank you, Keith, for your covering of us over all these years throughout my entire tenure, I deeply appreciate it. I know you're going off to do newer and better things. And so we wish you all the best on that one.

Thank you so much.

Greetings, and welcome to the Microsoft Fiscal Year 2026 Second Quarter Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded. It is now my pleasure to introduce Jonathan Neilson, Vice President of Investor Relations. Please go ahead.

Good afternoon, and thank you for joining us today. On the call with me are Satya Nadella, Chairman and Chief Executive Officer; Amy Hood, Chief Financial Officer; Alice Jolla, Chief Accounting Officer; and Keith Dolliver, Corporate Secretary and Deputy General Counsel.

On the Microsoft Investor Relations website, you can find our earnings press release and financial summary slide deck, which is intended to supplement our prepared remarks during today's call and provides the reconciliation of differences between GAAP and non-GAAP financial measures. More detailed outlook slides will be available on the Microsoft Investor Relations website where we provide outlook commentary on today's call.

On this call, we will discuss certain non-GAAP items. The non-GAAP financial measures provided should not be considered as a substitute for or superior to the measures of financial performance prepared in accordance with GAAP. They are included as additional clarifying items to aid investors in further understanding the company's second quarter performance in addition to the impact these items and events have on the financial results.

All growth comparisons we make on the call today relates to the corresponding period of last year, unless otherwise noted. We will also provide growth rates in constant currency when available as a framework for assessing how our underlying businesses performed, excluding the effect of foreign currency rate fluctuations. Where growth rates are the same in constant currency, we will refer to the growth rate only.

We will post our prepared remarks to our website immediately following the call until the complete transcript is available.

Today's call is being webcast live and recorded. If you ask a question, it will be included in our live transmission, in the transcript, and in any future use of the recording. You can replay the call and view the transcript on the Microsoft Investor Relations website.

During this call, we will be making forward-looking statements, which are predictions, projections or other statements about future events. These statements are based on current expectations and assumptions that are subject to risks and uncertainties. Actual results could materially differ because of factors discussed in today's earnings press release, in the comments made during this conference call, and in the Risk Factors section of our Form 10-K, Forms 10-Q and other reports and filings with the Securities and Exchange Commission. We do not undertake any duty to update any forward-looking statement.

And with that, I'll turn the call over to Satya.

Thank you very much, Jonathan. This quarter, the Microsoft Cloud surpassed $50 billion in revenue for the first time, up 26% year-over-year, reflecting the strength of our platform and accelerating demand. We are in the beginning phases of AI diffusion and its broad GDP impact. Our TAM will grow substantially across every layer of the tech stack as this diffusion accelerates and spreads. In fact, even in this early innings, we have built an AI business that is larger than some of our biggest franchises that took decades to build. Today, I'll focus my remarks across the 3 layers of our stack, cloud and token factory, agent platform and high-value agentic experiences.

When it comes to our cloud and token factory, the key to long-term competitiveness is shaping our infrastructure to support new high-scale workloads. We are building this infrastructure out for the heterogeneous and distributed nature of these workloads, ensuring the right fit with the geographic and segment specific needs for all customers, including the long tail. The key metric we're optimizing for is tokens per watt per dollar, which comes down to increasing utilization and decreasing TCO using silicon systems and software. A good example of this is the 50% increase in throughput we were able to achieve in one of our highest volume workloads, OpenAI inferencing, powering our Copilots.

And another example was the unlocking of new capabilities and efficiencies for our Fairwater data centers. In this instance, we connected both Atlanta and Wisconsin site through an AI WAN to build a first of its kind AI super factory. Fairwater's 2-storey design and liquid cooling allow us to run higher GPU densities and thereby improve both performance and latencies for high-scale training. All up, we added nearly 1 gigawatt of total capacity this quarter alone.

At the silicon layer, we have NVIDIA and AMD and our own Maia chips, delivering the best all up fleet performance, cost and supply across multiple generations of hardware. Earlier this week, we brought online our Maia 200 accelerator. Maia 200 delivers 10-plus petaFLOPS at FP4 precision with over 30% improved TCO compared to the latest generation hardware in our fleet. We will be scaling this starting with inferencing and synthetic data gen for our Superintelligence Team as well as doing inferencing for Copilot and Foundry.

And given AI workloads are not just about AI accelerators, but also consume large amounts of compute, we are pleased with the progress we are making on the CPU side as well. Cobalt 200 is another big leap forward, delivering over 50% higher performance compared to our first custom build processor for cloud-native workloads. Sovereignty is increasingly top of mind for customers, and we are expanding our solutions and global footprint to match. We announced DC investments in 7 countries this quarter alone, supporting local data residency needs. And we offer the most comprehensive set of sovereignty solutions across public, private and national partner cloud, so customers can choose the right approach for each workload with the local control they require.

Next, I want to talk about the agent platform. Like in every platform shift, all software is being rewritten. A new app platform is being born. You can think of agents as the new apps and to build, deploy and manage agents, customers will need a model catalog, tuning services, harness for orchestration, services for context engineering, AI safety, management, observability and security. It starts with having broad model choice. Our customers expect to use multiple models as part of any workload that they can fine tune and optimize based on cost, latency and performance requirements. And we offer the broadest selection of models of any hyperscaler. This quarter, we added support for GPT-5.2 as well as Claude 4.5. Already over 1,500 customers have used both Anthropic and OpenAI models on Foundry. We are seeing increasing demand for region-specific models, including Mistral and Cohere as more customers look for sovereign AI choices, and we continue to invest in our first-party models, which are optimized to address the highest value customer scenarios such as productivity, coding and security.

As part of Foundry, we also give customers the ability to customize and fine-tune models. Increasingly, customers want to be able to capture the tacit knowledge they possess inside of model weights as their core IP. This is probably the most important sovereign consideration for firms as AI diffuses more broadly across our GDP and every firm needs to protect their enterprise value. For agents to be effective, they need to be grounded in enterprise data and knowledge, that means connecting their agents to systems of record and operational data, analytical data as well as semi-structured and unstructured productivity and communications data. And this is what we are doing with our unified IQ layer, spanning Fabric, Foundry and data powering Microsoft 365.

In the world of context engineering, Foundry knowledge and Fabric are gaining momentum. Foundry knowledge delivers better context with automated source routing an advanced agentic retrieval while respecting user permissions. And Fabric brings together end-to-end operational real-time and analytical data. 2 years since it became broadly available, Fabric's annual revenue run rate is now over $2 billion with over 31,000 customers, and it continues to be the fastest-growing analytics platform on the market with revenue up 60% year-over-year. All of the number of customers spending $1 million plus per quarter on Foundry grew nearly 80%, driven by strong growth in every industry. And over 250 customers are on track to process over 1 trillion tokens on Foundry this year. There are many great examples of customers using all of this capability on Foundry to build their own agentic systems. Alaska Airlines is creating natural language flight search. BMW is speeding up design cycles, Land O'Lakes is enabling precision farming for co-op members, and SymphonyAI is addressing bottlenecks in the CPG industry. And of course, Foundry remains a powerful on-ramp for the entire cloud. The vast majority of Foundry customers use additional Azure solutions like developer services, app services, databases as they scale.

Beyond Fabric and Foundry, we are also addressing agent building by knowledge workers with Copilot Studio and Agent Builder. Over 80% of the Fortune 500 have active agents built using these low-code/no-code tools. As agents proliferate, every customer will need new ways to deploy, manage and protect them. We believe this creates a major new category and significant growth opportunity for us. This quarter, we introduced Agent 365, which makes it easy for organizations to extend their existing governance, identity, security and management to agents. That means the same controls they already use across Microsoft 365 and Azure, now extend to agents they build and deploy on our cloud or any other cloud. And partners like Adobe, Databricks, Genspark, Glean, NVIDIA, SAP, ServiceNow and Workday are already integrating Agent 365. We are the first provider to offer this type of agent control plane across clouds.

Now let's turn to the high-value agentic experiences we are building. AI experiences are intent-driven and are beginning to work at task scope. We are entering an age of macro delegation and micro steering across domains. Intelligence using multiple models is built into multiple form factors. You see this in chat, in new agent inbox, apps, coworkers, scaffoldings, agent workflows embedded in applications and IDs that are used every day or even in our command line with file system access and skills. That's the approach we are taking with our first-party family of copilot spanning key domains.

In consumer, for example, Copilot experiences span chat, news, feed, search, creation, browsing, shopping and integrations into the operating system, and it's gaining momentum. Daily users of our Copilot app increased nearly 3x year-over-year. And with Copilot checkout, we have partnered with PayPal, Shopify and Stripe, so customers can make purchases directly within the app. With Microsoft 365 Copilot, we are focused on organization-wide productivity.

Work IQ takes the data underneath Microsoft 365 and creates the most valuable stateful agent for every organization. It delivers powerful reasoning capabilities over people, their roles, their artifacts, their communications and their history and memory all within an organization security boundary. Microsoft 365 Copilot's accuracy and latency powered by Work IQ is unmatched, delivering faster and more accurate work grounded results than competition, and we have seen our biggest quarter-over-quarter improvement in response quality to date. This has driven record usage intensity with average number of conversations per user doubling year-over-year. Microsoft 365 Copilot also is becoming true daily habit with daily active users increasing 10x year-over-year.

We're also seeing strong momentum with researcher agent, which supports both OpenAI and Claude, as well as agent mode in Excel, PowerPoint and Word. All up, it was a record quarter for Microsoft 365 Copilot seat adds, up over 160% year-over-year. We saw accelerating seat growth quarter-over-quarter and now have 15 million paid Microsoft 365 Copilot seats and multiples more enterprise chat users.

And we are seeing larger commercial deployments. The number of customers with over 35,000 seats tripled year-over-year. Fiserv, ING, NASA, University of Kentucky, University of Manchester, U.S. Department of Interior and Westpac, all purchased over 35,000 seats. Publicis alone purchased over 95,000 seats for nearly all its employees. We are also taking share in Dynamics 365 with built-in agents across the entire suite. A great example of this is how Visa is turning customer conversations data into knowledge articles with our customer knowledge management agent and dynamics. And how Sandvik is using our sales qualification agent to automate lead qualification across tens and thousands of potential customers.

In coding, we are seeing strong growth across all paid GitHub Copilot. Copilot Pro Plus subs for individual devs increased 77% quarter-over-quarter, and all up now, we have 4.7 million paid Copilot subscribers, up 75% year-over-year. Siemens, for example, is going all in on GitHub adopting the full platform to increase developer productivity after a successful Copilot rollout to 30,000-plus developers. GitHub Agent HQ is the organizing layer for all coding agents like Anthropic, OpenAI, Google, Cognition and xAI in the context of customers GitHub repos. With Copilot CLI and VS Code, we offer developers the full spectrum of form factors and models they need for AI-first coding workflows.

And when you add Work IQ as a skill or an MCP to our developer workflow, it's a game changer, surfacing more context like e-mails, meetings, docs, projects, messages and more. You can simply ask the agent to plan and execute changes to your code base based on an update to a spec-in Sharepoint or using the transcript of your last engineering and design meeting in Teams.

And we're going beyond that with GitHub Copilot STK. Developers can now embed the same run time behind Copilot CLI, multi-model, multistep planning tools, MCP integration, Ops streaming directly into their applications. In security, we added a dozen new and updated security Copilot agents across Defender, Entra, Intune, and Purview. For example, Icertis, the SOC team used Security Copilot agent to reduce manual triage time by 75%, which is a real game changer in an industry facing a severe talent shortage.

To make it easier for security teams to onboard, we are rolling out security copilot to all our E5 customers and our security solutions are also becoming essential to manage organization's AI deployments. 24 billion Copilot interactions were audited by Purview this quarter, up 9x year-over-year.

Finally, I want to talk about 2 additional high-impact agenetic experiences. First, in health care, Dragon Copilot is the leader in its category, helping over 100,000 medical providers automate their workflows. Mount Sinai Health is now moving to a system-wide Dragon Copilot deployment for providers after a successful trial with its primary care physicians. All up, we helped document 21 million patient encounters this quarter, up 3x year-over-year.

And second, when it comes to science and engineering, companies like Unilever in consumer goods and Synopsys in EDA are using Microsoft Discovery to orchestrate specialized agents for R&D end-to-end. They're able to reason over scientific literature and internal knowledge, formulate hypotheses, spin up simulations and continuously iterate to drive new discoveries.

Beyond AI, we continue to invest in all our core franchises and meet the needs of our customers and partners, and we are seeing strong progress. For example, when it comes to cloud migrations, our new SQL server has over 2x the IaaS adoption of the previous version. In Security, we now have 1.6 million security customers, including over 1 million who use 4 or more of our workloads. Windows reached a big milestone, 1 billion Windows 11 users up over 45% year-over-year. And we had share gains this quarter across Windows, Edge and Bing, double-digit member growth in LinkedIn with 30% growth in paid video ads.

And in gaming, we are committed to delivering great games across Xbox, PC, Cloud and every other device, and we saw record PC players in paid streaming hours on Xbox.

In closing, we feel very good about how we are delivering for customers today and building the full stack to capture the opportunity ahead.

With that, let me turn it over to Amy to walk through our financial results and outlook, and I look forward to rejoining for your questions.

Thank you, Satya, and good afternoon, everyone. With growing demand for our offerings and focused execution by our sales teams, we again exceeded expectations across revenue, operating income and earnings per share, while investing to fuel long-term growth. This quarter, revenue was $81.3 billion, up 17% and 15% in constant currency. Gross margin dollars increased 16% and 14% in constant currency, while operating income increased 21% and 19% in constant currency.

Earnings per share was $4.14, an increase of 24% and 21% in constant currency when adjusted for the impact from our investment in OpenAI. And FX increased reported results slightly less than expected, particularly in Intelligent Cloud revenue. Company gross margin percentage was 68%, down slightly year-over-year, primarily driven by continued investments in AI infrastructure and growing AI product usage that was partially offset by ongoing efficiency gains, particularly in Azure and M365 Commercial Cloud as well as sales mix shift to higher-margin businesses.

Operating expenses increased 5% and 4% in constant currency, driven by R&D investments in compute capacity and AI talent as well as impairment charges in our gaming business. Operating margins increased year-over-year to 47%, ahead of expectations. As a reminder, we still account for investment in OpenAI under the equity method. And as a result of OpenAI's recapitalization, we now record gains or losses based on our share of the change in their net assets on their balance sheet as opposed to our share of their operating profit or losses from their income statement. Therefore, we recorded a gain which drove other income and expense to $10 billion in our GAAP results.

When adjusted for the OpenAI impact, other income and expense was slightly negative and lower than expected driven by net losses on investments. Capital expenditures were $37.5 billion, and this quarter, roughly 2/3 of our CapEx was on short-lived assets, primarily GPUs and CPUs. Our customer demand continues to exceed our supply. Therefore, we must balance the need to have our incoming supply better meet growing Azure demand with expanding first-party AI usage across services like M365 Copilot and GitHub Copilot, increasing allocations to R&D teams to accelerate product innovation and continued replacement of end-of-life server and networking equipment.

The remaining spend was for long-lived assets that will support monetization for the next 15 years and beyond. This quarter, total finance leases were $6.7 billion, and were primarily for large data center sites. And cash paid for PP&E was $29.9 billion. Cash flow from operations was $35.8 billion, up 60%, driven by strong Cloud billings and collections. And free cash flow was $5.9 billion and decreased sequentially, reflecting the higher cash capital expenditures from a lower mix of finance leases. And finally, we returned $12.7 billion to shareholders through dividend and share repurchases, an increase of 32% year-over-year.

Now to our commercial results. Commercial bookings increased 230% and 228% in constant currency, driven by the previously announced large Azure commitment from OpenAI that reflects multiyear demand needs as well as the previously announced Anthropic commitment from November and healthy growth across our core annuity sales motions.

Commercial remaining performance obligation, which continues to be reported net of reserves increased to $625 billion, and was up 110% year-over-year with a weighted average duration of approximately 2.5 years. Roughly 25% will be recognized in revenue in the next 12 months, up 39% year-over-year. The remaining portion recognized beyond the next 12 months increased 156%. Approximately 45% of our commercial RPO balance is from OpenAI. The significant remaining balance grew 28% and reflects ongoing broad customer demand across the portfolio.

Microsoft Cloud revenue was $51.5 billion and grew 26% and 24% in constant currency. Microsoft Cloud gross margin percentage was slightly better than expected at 67%, and down year-over-year due to continued investments in AI that were partially offset by ongoing efficiency gains noted earlier.

Now to our segment results. Revenue from Productivity and Business Processes was $34.1 billion and grew 16% and 14% in constant currency. M365 Commercial cloud revenue increased 17% and 14% in constant currency with consistent execution in the core business and increasing contribution from strong copilot results. ARPU growth was again led by E5 and M365 Copilot, and paid M365 commercial seats grew 6% year-over-year to over 450 million with installed base expansion across all customer segments, though primarily in our small and medium business and frontline worker offerings.

M365 Commercial products revenue increased 13% and 10% in constant currency, ahead of expectations due to higher-than-expected Office 2024 transactional purchasing. M365 consumer cloud revenue increased 29% and 27% in constant currency, again driven by ARPU growth. M365 consumer subscriptions grew 6%. LinkedIn revenue increased 11% and 10% in constant currency driven by Marketing Solutions.

Dynamics 365 revenue increased 19% and 17% in constant currency with continued growth across all workloads. Segment gross margin dollars increased 17% and 15% in constant currency, and gross margin percentage increased, again driven by efficiency gains at M365 Commercial Cloud that were partially offset by continued investments in AI, including the impact of growing copilot usage.

Operating expenses increased 6% and 5% in constant currency, and operating income increased 22% and 19% in constant currency. Operating margins increased year-over-year to 60%, driven by improved operating leverage as well as the higher gross margins noted earlier.

Next, the Intelligent Cloud segment. Revenue was $32.9 billion and grew 29% and 28% in constant currency. In Azure and Other Cloud services, revenue grew 39% and 38% in constant currency, slightly ahead of expectations with ongoing efficiency gains across our fungible fleet, enabling us to reallocate some capacity to Azure that was monetized in the quarter.

As mentioned earlier, we continue to see strong demand across workloads, customer segments and geographic regions, and demand continues to exceed available supply. In our on-premises server business, revenue increased 2% and 1% in constant currency, ahead of expectations, driven by demand for our hybrid solutions, including a benefit from the launch of SQL Server 2025, as well as higher transactional purchasing ahead of memory price increases. Segment gross margin dollars increased 20% and 19% in constant currency. Gross margin percentage decreased year-over-year, driven by continued investments in AI and sales mix shift to Azure, partially offset by efficiency gains in Azure.

Operating expenses increased 3% and 2% in constant currency, and operating income grew 28% and 27% in constant currency. Operating margins were 42%, down slightly year-over-year as increased investments in AI were mostly offset by improved operating leverage.

Now to More Personal Computing. Revenue was $14.3 billion and declined 3%. Windows OEM and devices revenue increased 1%, and was relatively unchanged in constant currency. Windows OEM grew 5% with strong execution as well as a continued benefit from Windows 10 end of support. Results were ahead of expectations as inventory levels remained elevated with increased purchasing ahead of memory price increases.

Search and news advertising revenue ex TAC increased 10% and 9% in constant currency, slightly below expectations, driven by some execution challenges. As expected, the sequential growth rate moderated as the benefit from third-party partnerships normalized.

And in Gaming. Revenue decreased 9% and 10% in constant currency. Xbox content and services revenue decreased 5% and 6% in constant currency, and was below expectations driven by first-party content with impact across the platform. Segment gross margin dollars increased 2% and 1% in constant currency, and gross margin percentage increased year-over-year, driven by sales mix shift to higher-margin businesses.

Operating expenses increased 6% and 5% in constant currency, driven by the impairment charges in our gaming business noted earlier, as well as R&D investments in compute capacity and AI talent. Operating income decreased 3% and 4% in constant currency, and operating margins were relatively unchanged year-over-year at 27% as higher operating expenses were mostly offset by higher gross margins.

Now moving to our Q3 outlook, which unless specifically noted otherwise, is on a U.S. dollar basis. Based on current rates, we expect FX to increase total revenue growth by 3 points. Within the segments, we expect FX to increase revenue growth by 4 points in Productivity and Business Processes, and 2 points in Intelligent Cloud and More Personal Computing. We expect FX to increase COGS and operating expense growth by 2 points. As a reminder, this impact is due to the exchange rates a year ago.

Starting with the total company. We expect revenue of USD 80.65 billion to USD 81.75 billion or growth of 15% to 17%, with continued strong growth across our commercial businesses, partially offset by our consumer businesses. We expect COGS of USD 26.65 billion to USD 26.85 billion or growth of 22% to 23%, and operating expense of USD 17.8 billion to USD 17.9 billion or growth of 10% to 11%, driven by continued investment in R&D, AI compute capacity and talent against a low prior year comparable. Operating margins should be down slightly year-over-year.

Excluding any impact from our investments in OpenAI, other income and expense is expected to be roughly $700 million, driven by a fair market gain in our equity portfolio and interest income, partially offset by interest expense, which includes the interest payments related to data center finance leases. And we expect our adjusted Q3 effective tax rate to be approximately 19%.

Next, we expect capital expenditures to decrease on a sequential basis due to the normal variability from cloud infrastructure build-outs and the timing of delivery of finance leases. As we work to close the gap between demand and supply, we expect the mix of short-lived assets to remain similar to Q2.

Now our commercial business. In commercial bookings, we expect healthy growth in the core business on a growing expiry base when adjusted for the OpenAI contracts in the prior year. As a reminder, the significant OpenAI contracts signed in Q2 represents multiyear demand needs from them, which will result in some quarterly volatility in both bookings and RPO growth rates going forward. Microsoft Cloud gross margin percentage to be roughly 65%, down year-over-year, driven by continued investments in AI.

Now to segment guidance. In Productivity and Business Processes, we expect revenue of USD 34.25 billion to USD 34.55 billion or growth of 14% to 15%. In M365 Commercial Cloud, we expect revenue growth to be between 13% and 14% in constant currency with continued stability and year-over-year growth rates on a large and expanding base. Accelerating Copilot momentum and ongoing E5 adoption will again drive ARPU growth. M365 commercial products revenue should decline in the low single digits, down sequentially, assuming office 2024 transactional purchasing trends normalize.

As a reminder, M365 commercial products include components that can be variable due to in-period revenue recognition dynamics. M365 consumer cloud revenue growth should be in the mid- to high 20% range, driven by growth in ARPU as well as continued subscription volume. For LinkedIn, we expect revenue growth to be in the low double digits. And in Dynamics 365, we expect revenue growth to be in the high teens with continued growth across all workloads.

For Intelligent Cloud, we expect revenue of USD 34.1 billion to USD 34.4 billion or growth of 27% to 29%. In Azure, we expect Q3 revenue growth to be between 37% and 38% in constant currency against a prior year comparable that included significantly accelerating growth rates in both Q3 and Q4. As mentioned earlier, demand continues to exceed supply, and we will need to continue to balance the incoming supply we can allocate here against other priorities.

As a reminder, there can be quarterly variability in year-on-year growth rates depending on the timing of capacity delivery and when it comes online as well as from in-period revenue recognition depending on the mix of contracts. In our on-premises server business, we expect revenue to decline in the low single digits as growth rates normalize following the launch of SQL Server 2025, though increased memory pricing could create additional volatility in transactional purchasing.

In More Personal Computing, we expect revenue to be USD 12.3 billion to USD 12.8 billion. Windows OEM and devices revenue should decline in the low teens. Growth rates will be impacted as the benefit from Windows 10 end of support normalizes and as elevated inventory levels come down through the quarter. Therefore, Windows OEM revenue should decline roughly 10%.

The range of potential outcomes remains wider than normal, in part due to the potential impact on the PC market from increased memory pricing. Search and News advertising ex TAC revenue growth should be in the high single digits. Even as we work to improve execution, we expect continued share gains across Bing and Edge with growth driven by volume, and we expect sequential growth moderation as the contribution from third-party partnerships continues to normalize.

And in Xbox content and services, we expect revenue to decline in the mid-single digits against a prior year comparable that benefited from strong content performance, partially offset by growth in Xbox Game Pass. And hardware revenue should decline year-over-year.

Now some additional thoughts on the rest of the fiscal year and beyond. First, FX. Based on current rates, we expect FX to increase Q4 total revenue and COGS growth by less than 1 point with no impact to operating expense growth. Within the segments, we expect FX to increase revenue growth by roughly 1 point in Productivity and Business Processes and More Personal Computing, and less than 1 point in Intelligent Cloud.

With the strong work delivered in H1 to prioritize investment in key growth areas and the favorable impact from a higher mix of revenue in our Windows OEM and commercial on-prem businesses, we now expect FY '26 operating margins to be up slightly. We mentioned the potential impact on Windows OEM and on-premises server markets from increased memory pricing earlier. In addition, rising memory prices would impact capital expenditures, though the impact on Microsoft Cloud gross margins will build more gradually as these assets depreciate over 6 years.

In closing, we delivered strong top line growth in H1, and are investing across every layer of the stack to continue to deliver high-value solutions and tools to our customers. With that, let's go to Q&A, Jonathan.

Thanks, Amy. We'll now move over to Q&A. Out of respects for others on the call, we request that participants please only ask 1 question. Operator, can you please repeat your instructions?

[Operator Instructions] And our first question comes from the line of Keith Weiss with Morgan Stanley.

I'm looking at Microsoft print where earnings is growing 24% year-on-year, which is a spectacular result. Great execution on your part, top line growing well, margins expanding. But I'm looking at after-hours trading and the stock is still down. And I think one of the core issues that is weighing on investors is CapEx is growing faster than we expected and maybe Azure is growing a little bit slower than we expected. And I think that fundamentally comes down to a concern on the ROI on this CapEx spend over time. So I was hoping you guys could help us fill in some of the blanks a little bit in terms of how should we think about capacity expansion and what that can yield in terms of Azure growth going forward. More to the point, how should we think about the ROI on this investment as it comes to fruition?

Thanks, Keith. And let me start and Satya can add some broader comments, I'm sure. I think the first thing, I think you really asked a very direct correlation that I do think many investors are doing, which is between the CapEx spend and seeing an Azure revenue number. And we tried last quarter, and I think, again, this quarter to talk more specifically about all the places that the CapEx spend, especially the short-lived CapEx spend across CPU and GPU and where that will show up.

Sometimes, I think it's probably better to think about the Azure guidance that we give as an allocated capacity guide about what we can deliver in Azure revenue. Because as we spend the capital and put GPUs specifically, it applies to CPUs, the GPUs more specifically, we're really making long-term decisions. And the first thing we're doing is solving for the increased usage in sales and the accelerating pace of M365 Copilot as well as GitHub Copilot, our first-party apps. Then we make sure we're investing in the long-term nature of R&D and product innovation. And much of the acceleration that I think you've seen from us and products over the past a bit is coming because we are allocating GPUs and capacity to many of the talented AI people we've been hiring over the past years.

Then, when you end up, is that, you end up with the remainder going towards serving the Azure capacity that continues to grow in terms of demand. And a way to think about it, because I think, I get asked this question sometimes, is if I had taken the GPUs that just came online in Q1 and Q2 in terms of GPUs and allocated them all to Azure, the KPI would have been over 40. And I think the most important thing to realize is that this is about investing in all the layers of the stack that benefit customers. And I think that's hopefully helpful in terms of thinking about capital growth, it shows in every piece, it shows in revenue growth across the business and shows as OpEx growth as we invest in our people.

Yes, I think you -- Amy covered it. But basically, as an investor, I think when you think about our capital and you think about the GM profile of our portfolio, you should obviously think about Azure. But you should think about M365 Copilot and you should think about GitHub pilot, you should think about Dragon Copilot, Security Copilot. All of those have a GM profile and lifetime value. I mean if you think about it, acquiring an Azure customer is super important to us, but so is acquiring an M365 or a GitHub or a Dragon Copilot, which are all by the way incremental businesses and TAMs for us. And so we don't want to maximize just 1 business of ours, we want to be able to allocate capacity while we're sort of supply constrained in a way that allow us to essentially build the best LTV portfolio. That's on one side. And the other one that Amy mentioned is also R&D. I mean you got to think about compute is also R&D, and that's sort of the second element of it. And so we are using all of that, obviously, to optimize for the long term.

The next question comes from the line of Mark Moerdler with Bernstein Research.

Congrats on the quarter. One of the other questions we believe investors want to understand is how to think about your line of sight from hardware CapEx investment to revenue and margins. You capitalized servers over 6 years, but the average duration of your RPO is 2.5 years, up from 2 years last quarter. How do investors get comfortable that since this is a lot of this CapEx is AI-centric that you'll be able to capture sufficient revenue over the 6-year useful life of the hardware to deliver solid revenue and gross profit dollars growth, hopefully, one similar to the CPU revenue.

Thanks, Mark. Let me start with at a high level and Satya can add as well. I think when you think about average duration, I think what you're getting to is -- and we need to remember, is it, average duration is a combination of a broad set of contract arrangements that we have. A lot of them around things like M365 or our BizApps portfolio, are shorter dated, right, 3-year contracts. And so they have, quite frankly, a short duration. The majority then that's remaining are Azure contracts are longer duration. And you saw that this quarter when we saw the extension of that duration from around 2 years to 2.5 years. And the way to think about that is the majority of the capital that we're spending today, and a lot of the GPUs that we're buying are already contracted for most of their useful life. And so a way to think about that is much of that risk that I think you're pointing to isn't there, because they're already sold for the entirety of their useful life. And so part of it exists because you have this shorter-dated RPO because of some of the M365 stuff. If you look at the Azure only, RPO is a little bit more extended. A lot of that is CPU basis. It's not just GPU. And on the GPU contracts that we've talked about, including for some of our largest customers, those are sold for the entire useful life of the GPU. And so there's not the risk to which I think you may be referring. Hopefully, that's helpful.

Yes. And just to -- one other thing I would add to it is, in addition to sort of what Amy mentioned, which is it's already contracted for the useful life is we do use software to continuously around even the latest models on the fleet that is aging, if you will. So that's sort of what gives us that duration. And so at the end of the day, we want to have -- that's why we even think about aging the fleet constantly, right? So it's not about buying a whole lot of gear 1 year. It's about each year, you write the Moore's Law, you add, you use software, and then you optimize across all of it.

And Mark, maybe to state this in case it's not obvious, is that as you go through the useful life, actually, you get more and more and more efficient at delivery. So where you've sold the entirety of its life, the margins actually improved with time. And so I think that may be a good reminder to people as we see that, obviously, in the CPU fleet all the time.

The next question comes from the line of Brent Thill with Jefferies.

Amy, on 45% of the backlog being related to OpenAI. I'm just curious if you can comment, there's obviously concern about the durability. And I know maybe there's not much you can say on this, but I think everyone is concerned about the exposure. And if you could maybe talk through your perspective and what both you and Satya are seeing.

I think maybe I would have thought about the question quite differently, Brent. The first thing to focus on is the reason we talked about that number is because 55% or roughly $350 billion is related to the breadth of our portfolio, a breadth of customers across solutions, across Azure, across industries, across geographies. That is a significant RPO balance, larger than most peers, more diversified than most peers. And frankly, I think we have super high confidence in it.

And when you think about that portion alone growing 28%, it's really impressive work on the breadth as well as the adoption curve that we're seeing, which is I think what I get asked most frequently, it's grown by customer segment, by industry and by geo. And so it's very consistent. And so then if you're asking about how do I feel about OpenAI and the contract and the health, listen, it's a great partnership. We continue to be their provider of scale. We're excited to do that. We sit under one of the most successful businesses built, and we continue to feel quite good about that. It's allowed us to remain a leader in terms of what we're building and being on the cutting edge of app innovation.

The next question comes from the line of Karl Keirstead with UBS.

Amy, regardless of how you allocate the capacity between first party and third party, can you comment qualitatively on the amount of capacity that's coming on. I think the 1 gigawatt added in the December quarter was extraordinary and hence that the capacity adds are accelerating. But I think a lot of investors have their eyes on Fairwater Atlanta, Fairwater Wisconsin, and would love some comments about the magnitude of the capacity adds regardless of how they're allocated in the coming quarters.

Yes, Karl, I think we've said a couple of things. We're working as hard as we can to add capacity as quickly as we can. You've mentioned specific sites like Atlanta or Wisconsin, those are multiyear deliveries. So I wouldn't focus necessarily on specific locations. The real thing we've got to do, and we're working incredibly hard doing it, is adding capacity globally. A lot of that will be added in the United States, the 2 locations you've mentioned, but it also needs to be added across the globe to meet the customer demand that we're seeing and the increased usage. We'll continue to add both long-lived infrastructure. The way to think about that is we need to make sure we've got power and land and facilities available and we'll continue to put GPUs and CPUs in them when they're done as quickly as we can.

And then finally, we'll try to make sure we can get as efficient as we possibly can on the pace at which we do that and how we operate them so that they can have the highest possible utility. And so I think it's not really about 2 places, Karl, I would definitely abstract away from that. Those are multiyear delivery time lines. But really, we just need to get it done every location where we're currently in a build or starting to do that. We're working as quickly as we can.

The next question comes from the line of Mark Murphy with JPMorgan.

Satya, the performance achievements of the Maia 200 accelerator for inference, look quite remarkable, especially in comparison to TPUs and Trinium and Blackwell, which have just been around a lot longer. Could you put that accomplishment in perspective in terms of how much of a core competency you think silicon might become for Microsoft. And Amy, are there any ramifications worth mentioning there in terms of supporting your gross margin profile for inference costs going forward?

Yes, thanks for the question. So a couple of things. One is we've been at this in a variety of different forms for a long, long time in terms of building our own silicon. And so we're very, very thrilled about the progress with Maia 200, and -- especially when we think about running a GPT-5.2 and the performance we were able to get in the GEMS at FB4, just proof point that when you have a new workload, a new shape of a workload, you can start innovating end-to-end between the model and the silicon and the entire system. It's just not even about just the silicon, the way the networking works at rack scale that's optimized with memory for this particular workload.

And the other thing is we're obviously round-tripping and working very closely with own super intelligence team with all of our models, as you can imagine, whatever we build will be all optimized for Maia. So we feel great about it. And I think the way to think about all up is we're in such early innings. I mean, even just look at the amount of silicon innovation and systems innovation. Even since December, I think the new thing is everybody is talking about low latency inference, right? And so one of the things we want to make sure is we are not locked into any one thing. If anything, we have great partnership with NVIDIA, with AMD, they are innovating, we're innovating. We want a fleet at any given point in time to have access to the best TCO. And it's not a one-generation game. I think a lot of folks just talk about who's ahead. It's just remember, you have to be ahead for all time to come. And that means you really want to think about having a lot of innovation that happens out there to be in your fleet, so that your fleet is fundamentally advantaged at the TCO level. So that's kind of how I look at it, which is we are excited about Maia. We're excited about Cobalt. We're excited about our DPU, our NIC. So we have a lot of systems capability. That means we can vertically integrate. And because we can vertically integrate doesn't mean we just only vertically integrate. And so we want to be able to have the flexibility here, and that's what you see us do.

The next question comes from the line of Brad Zelnick with Deutsche Bank.

Satya, we heard a lot about Frontier transformations from Judson at Ignite, and we've seen customers realize breakthrough benefits when they adopt the Microsoft AI stack. Can you help frame for us the momentum in enterprises embarking on these journeys? And any expectation for how much their spend with Microsoft can expand in becoming frontier firms?

Yes. Thank you for that. So I think one of the things that we are seeing is the adoption across the 3 major suites of ours, right? So if you take M365, you take what's happening with security and you take GitHub. In fact, it's fascinating. I mean these 3 things had effectively compounding effects for our customers in the past, like something like Entra as an identity system, or Defender as the protection system, across all 3 was sort of super helpful.

But so what now you're seeing is something like Work IQ, right? So I mean just to give you a favor for it, the most important database underneath for any company that uses Microsoft today is the data underneath Microsoft 365. And the reason is because it has all those tacit information, right, who are your people, what are their relationships, what are their projects they're working on, what are their artifacts, their communications. So that's a super important asset for any business process, business workflow context.

In fact, the scenario I even had in my transcript around you can now take Work IQ as an MCP server and GitHub Repo and say, "Hey, please look at my design meetings for the last month in Teams and tell me if my repo reflects it." I mean that's a pretty high-level way to think about how, what is happening previously perhaps with our tools business and our GitHub business are suddenly now being transformative, right? That agent black plane is really transforming companies in some sense, right? That's, I think, the most magical thing, which is you deploy these things. And suddenly, the agents are helping you coordinate, bring more leverage to your enterprise.

Then on top of it, of course, there is the transformation, which is what businesses are doing. How should we think about customer service. How should we think about marketing. How should we think about finance. How should we think about that and build our own agents. That's where all the services in Fabric and Foundry. And of course, GitHub tooling is helping them or even the low-code/no-code tools. I had some stats on how much that's being used. But one of the more exciting things for me is the new agents systems, M365 Copilot, GitHub Copilot, Security Copilot, all coming together to compound the benefits of all the data and all the deployment, I think, is probably the most transformative effect right now.

Thanks, Brad. Operator, we have time for 1 last question.

And the last question will come from the line of Raimo Lenschow with Barclays.

The last few quarters we talked -- besides the GPU side, you talked about CPU as well on the Azure side and you had some operational changes at the beginning of January last year. Can you speak what you saw there? And maybe put it more on a bigger picture in terms of clients realizing that their move to the cloud is important if you want to deliver proper AI. So what are we seeing in terms of cloud transition?

I didn't quite...

Sorry, Raimo, you were asking about the SMC CPU side? Or can you just repeat the question, please?

Yes. Sorry. So I was wondering about the CPU side of Azure because we had some operational changes there. And we also hear from the [ field a lot ] that people are realizing they need to be in the Cloud if you want to do proper AI and if that's kind of driving momentum.

Yes. I think I get it. So first of all, I had mentioned in my remarks that when you think about AI workloads, you should think of AI workloads as just AI accelerator compute, right? Because in some sense, you take any agent, the agent will then spawn through tools used maybe a container, which runs obviously on compute. In fact, we have -- whenever we think about even building out of the fleet, we think of in ratios or even for a training job, by the way. An AI training job requires a bunch of compute and a bunch of storage very close to compute. So therefore -- and same thing in inferencing as well.

So in inferencing with agent mode would require you to essentially provision a computer or computing resources to the agent. So not -- they don't need GPUs. They're running on GPUs, but they need computers, which are compute and storage. So that's what's happening even in the new world.

The other thing you mentioned is the Cloud migrations are still going on. In fact, 1 of the stats I had was SQL -- latest SQL server growing as an IaaS service in Azure. And so -- that's one of the reasons why we have to think about our commercial cloud and keep it balanced with the rest of our AI Cloud because when clients bring their workloads and build new workloads, they need all of these infrastructure elements in the region in which they are deploying.

That wraps up the Q&A portion of today's earnings call. Thank you for joining us today, and we look forward to speaking with you all soon.

Thank you all.

Thank you.

Thank you. This concludes today's conference. You may disconnect your lines at this time, and we thank you for your participation. Have a great night.

Good morning.

Good morning, everyone.

Good morning for day 2. Actually, I'll leave out the day with like 2 congratulations. Judson and Craig, both of you have expanded roles in the organization. So first of all, congratulations.

Thank you.

Thank you.

The Judson, so at the moment, it's really exciting times if you think about the technology world. Can you -- one of the things that was really interesting when I was here at Incyte a couple of weeks ago was this notion of the frontier transformation. Can you talk a little bit -- what does it mean to you? What's the message here?

Yes. So we're really working with customers to have sort of an evolution in the wave of AI transformation because if you look at the first couple of years of progress, largely focused on efficiency and productivity and largely tech-driven. And if you look at the corpus of AI projects out there in the market, I think Craig would agree with this, you see that there is an extraordinarily high failure rate of AI projects, north of 80%, depending on the research that you study.

And if you pick apart the reasons why, some of it is classic tech to business, misalignment with business goals. Others are tied back to sort of the disorganization of data in the enterprise and lacking a real data estate or foundation to build quality and capabilities.

And then if you poke further, it's down into not having the right kind of AI development tools that allow researchers to become productive in an environment. And so if we look at by contrast, the places where we have seen strong success across our customer base, across multiple industries and around the world, it's largely tied to this notion of business-led transformation. What are you doing to enrich your employee experience and tie that back to your own KPIs? How are you're looking at it in the frame of customer engagement, driving top line revenue, not just bottom line savings?

How are you reshaping business processes, not just throwing tech at existing ones, but actually stopping and saying, how do I create an AI-first business process that's grounded in human ambition and empowered by assistance in an agent ecosystem?

And then four, how are you putting AI to work for innovation? If you have this success framework where you're tying it back to business-led transformation and then applying the technology portfolio to it, the success rate goes up materially. So we're coining this notion of frontier transformation as being a business-led AI evolution that really allows for a fundamental reinvention of the business empowered by AI rather than the other way around. So it may sound like a subtlety, but there is a massive difference in how we're tracking progress with customers on these business-led transformations. And so much of what we've announced in terms of product portfolio is meant to help and stimulate that kind of growth.

Is that -- and how do you compare and contrast that to like previous tech transformation, we had like the move to the cloud. We had the Internet, et cetera, like this feels slightly different.

I mean I think it's fair to say almost every wave of technology has been largely grounded in efficiency, right? How do I throw tech out of process, make it faster, drive down cost? And AI certainly has the potential to do that, right?

I mean much of what you read about is the impact on white-collar work and how is it going to change? But if you pull the thread on it, the bottom line is AI can do a lot more for humanity and needs to do a lot more for humanity than simply drive efficiencies. And so if you look at AI and drug discovery, for example, we are shaving an order of magnitude off the cycles that are required to get new drugs to market and new drugs to market safely without having to risk human trials, only to have them fail in the last mile. And there's objective evidence of making real progress there. The same can be said with material science, advanced -- it's in quantum and even in our own Quantum labs at Microsoft, I think, is a big reason why we're ahead in getting quantum chipsets to market is because of the use of AI and material science.

And so if you pivot away from, gosh, how can I just save money and eliminate jobs with AI versus like, hey, actually, how can I get more done? How can I unlock creativity and innovation with AI? It's a remarkable change in terms of the business outcomes that you can drive.

And then that's actually a good segue into my first question for Craig, like our industry, like financial services, we -- sorry, if I say that, like historically, you're not like first adopters, but like I see a big movement around AI. How do you feel about AI for us?

The trends. So look, I think we'd all agree the opportunity is quite significant. And virtually everything about AI is changing at pace. When I sort of reflect on some of the key trends and probably reflecting on the more recent influence of sort of generative AI, sort of 5 things sort of come to mind.

Firstly, I think you're starting to see the shift from experimentation into operationalization, slower than, in fact, what we might have anticipated a little bit to Judson's point. But AI is starting to change roles and workflows in a really fundamental way. Take your customer service agent as an example, who now has real-time expert support, guiding them through a call, summarizing the results of that call, assessing their development needs throughout that call and then in a position to create personalized development content, that's not augmentation. It's a shift in the way work is happening and it's a shift in how capability is being developed.

And you're starting to see that pattern across operations, across technology, across trust risk management. I think the second thing that you see happening, Raimo, is this shift from AI use in back office to the point of customer interface from human in the loop to sort of more autonomous workflows.

And the significance of that shift is it elevates the importance of governance. And by that, I don't mean just guardrails. I mean understanding the role of each agent in the organization, be it human or AI, understanding their commissions, understanding the decisions they make, where you need to have oversight. And I think the organizations that get that right will establish autonomy with accountability.

And within financial services, that's going to be key. I think the third thing that I see, and Judson touched on this, is whilst we often talk about AI, the real battleground is, in fact, data. Yes, of course, data fuels AI. But too often that data is not accessible, not of the right quality, not structured in the way to support real-time decisions.

And I think most organizations are discovering the same fundamental truth. And that is you cannot scale AI while your data is trapped in silos, while you can't link activity across the full customer life cycle and while you don't have the lineage and controls in your data to have trust in that data.

So while AI often gets the headlines, actually, it's in data that I think real advantage will be created. Then I think you're seeing infrastructure being recreated. I think we all will sort of understand that to use AI at scale, yes, I need access to powerful models across multiple clouds. Yes, I need increased compute density. Yes, I need much greater power capacity.

But I think the really important shift is in resilience that is AI, in particular, agents and autonomous agents become more integral to critical workflows in organizations like financial institutions, the expectations on resilience will actually raise quite sharply. And I think that's going to have a profound impact across the entire stack from networks to multi-cloud infrastructure to data platforms to APIs.

And I think you're going to see much greater focus come on to this question of resilient AI provisioning within organizations. And then finally, I think AI is emerging as much as a threat as it is an opportunity. You think about those traditional cyber exploits, identifying vulnerabilities, polling of systems, creating malware, all of that can be executed now at a scale and pace that just wasn't attainable before.

But it's probably the new exploits that are emerging or the new opportunities for exploits that are emerging that are most concerning. Think of a mass disinformation campaign executed across multiple social platforms, a vast target audience using deep fakes executed almost instantly. That's a very different type of threat that we now need to begin to mitigate and arguably probably one of the more important of those sort of -- those trends.

Yes. And then Judson, Craig mentioned a couple of interesting points, data as a really core kind of point here as well. How do you think about it at Microsoft in itself? Like yesterday, I was talking with Gina from ServiceNow and asking her about is she eating her own dog food. She's had champagne actually. like what's the -- how are you going about it? And how do you think about that data aspect as well?

Yes. So the data piece of it is super critical for us. I mean we just had our Ignite conference a couple of weeks ago and had over 70 product announcements, but really in sort of 2 major areas, one around intelligence and the other on trust, both sort of really grounded in data.

And what we did on the intelligence front was a pretty massive lift to help our customers get their arms around their data in a much more usable way. Let me explain this in a little bit of detail here just so you all can understand because there's a fundamental difference on asking AI engineers, even the best and brightest of AI developers to go after data sources using connectors and APIs versus using a semantic layer through what we call MCP servers, allowing agents to speak to agents. If you take a model provider's approach towards trying to synthesize a business process, a model provider will take data through 1,000 small straws, hoover it up and try to inference over that data, and we use inference as if it's a super cool smart technical word, it's fancy guessing. So you're hovering up tons of data through lots of small straws and guessing on an outcome.

What we've done by contrast is to serve up these intelligent layers. The first one being what we call Work IQ. Work IQ is basically the brain inside of Microsoft 365 Copilot. It knows how you work, with whom you work, the content over which you collaborate, and it knows it precisely, not guessing. And those are actual workflows. When you delegate an important task, it knows to whom you delegate it and has the history of how you've done so for years.

We've served up Work IQ now, not just as the brain inside of Microsoft 365 Copilot, but basically as an Azure meter. So you can build agents on top of this intelligence layer that provide a higher degree of accuracy, speed and trust. You can reason over confidential and encrypted documents, if that's what your rights and privileges entitle you to do. And then it's a material leap forward versus what a model provider can do or any other company that is trying to somehow sit data through 1,000 straws or frankly, an enterprise customer trying to have their AI engineers do that themselves. It's super hard. We've done this at every layer of the stack. We've also done it on Fabric.

We released Fabric IQ. Fabric is our data services product that allows you to reason over multiple different data sources. So BigQuery running on Google Cloud platform or Amazon S3 data stores and of course, all of the Azure data services and even data services running in your environment. With Fabric IQ, what we've done is we've taken the semantic layer inside of Power BI. So the sort of the one binding link across all of those data services across multiple clouds is the fact that most of them use Power BI to understand the semantic layers within their business.

We've also served that up now as a singular API, so you can reason over data far more accurately in the way in which your business understands it through Fabric IQ regardless of what cloud you want to run on. And then, of course, then the final tier, the foundry layer where AI applications are actually built. We have unlocked all of the knowledge bases that sit into agent-to-agent communication, all of the Azure search foundational elements into Foundry IQ. So you can develop agents far more effectively and efficiently than ever before with this intelligent layer. And everything I've just described to you is model diverse.

So a new model comes out tomorrow, great, snap it in. We support over 11,000 models. It's open and heterogeneous at every layer of the stack. And so we think -- back to this like how do we have to drive frontier transformation versus throwing tech at existing business processes and hoping things get better. You have to serve up this layer of intelligence to allow business to get more work done. And so it's been a big investment for us, and I think a pretty strategic advantage for customers that rely on and trust Microsoft.

Yes. Perfect. And then, Craig, like if you think about it on the Barclays side, like there's so many potential things that you -- that AI could touch. Like how do you think about our journey? Like what's the -- how do you prioritize about that?

Yes. So look, I think we were quite early into generative AI, largely off the back of, I think, the investments we've made across infrastructure, data security, those foundations. And if I sort of wind forward to where we are today, it's emerging out of what you might want to think of as the creative chaos phase, lots of experimentation, rapid learning, uneven value. But I think through that, a level of insight around where the opportunities are and how we need to go about executing against those opportunities, sort of 3 areas that I think are priorities.

The first, and Judson touched on, this is process transformation. And I tend to think of generative AI, in particular, as almost the third generation of automation, the first rule-based, the second, third base, now language-based. And it's when you combine the potential of AI, digital and data that I think you create an enormous opportunity to rethink, redesign process to unlock true -- real value there.

And I think this is Judson's point around frontier transformation. It's about working front to back and left to right across the organization, and that's exactly what we're doing. Two, colleague enablement, just getting the technology into the hands of colleagues, allowing them to innovate, simple augmentation, more complex autonomous workflows.

For me, the challenge there isn't a technical challenge. I'm not even sure it's a skills challenge. Yes, skills is important. It's absolutely a mindset challenge. It's about how you actually inspire people to sort of embrace the art of the possible, set aside traditional ways of working, invest the time and energy in discovering what you can do with this technology. And I would say that's a war not yet won, but certainly one that we're really focused on. And then finally, tech modernization. And that goes beyond just software development. It's about how we're using AI to shift from legacy architectures, legacy technologies, legacy practices to more contemporary ones. It's an area where we see sort of crews, so cross-functional teams of agents and engineers having enormous potential. It's allowing us to refactor legacy technology in a fraction of the time that we previously did. So the opportunity to accelerate tech modernization far greater than what we might have seen even 2 years ago.

Now I would say this, Raimo, if you'll ask me this question again in 12 or 24 months' time, I may very well have a different answer, right? And I think it's the nature of this technology. We're still learning. The technology is evolving really quickly. So what's key is, yes, be curious, be adaptable, keep really focused on customer value is the only way I think you can approach this.

And Judson, from your customer conversations, like I'm sure you're getting that as well, like how do I think about the return like -- of all these kind of adopting this innovation, et cetera? Like what's the conversation like?

So that was the second big part of our work at Ignite. I mentioned intelligence and trust. On the trust side -- trust means a lot of things, of course, there's a security element to trust, which I'll come to. But there's just good old-fashioned trust in the business and trust in the partnership is this journey that we're about to embark upon of embracing AI to reinvent business processes and become frontier is that something that we can trust.

Can we trust that the ROI at the end is actually going to be something that makes the juice worth to squeeze, right? So we announced a lot of new capability around this observability aspect that Craig mentioned, we announced a new product called Agent 365. What Agent 365 allows you to do is basically visualize all of your AI artifacts across the entire enterprise, whether they're built on the Microsoft platform or any other third-party platform.

And allows you to register these agents, provide identity and access privileges to them. But far more importantly, it actually allows you to visualize how they come together in workflows and how they come together with human interaction across those workflows. So we turned on the product because part of the new role is I've inherited IT at Microsoft. We turned on Agent 365 before we launched and announced the product. And we have 138,000 agents being used by 88,000 employees on a weekly basis, which I would offer you all up to turn on Agent 365, it's free in your environment because I would be willing to bet you have more AI happening inside your organization than you know about.

But the beauty of this is you can instantly understand how these agents come together in a process, in a workflow and then also the usage intensity so that you can then go back and optimize. And say within a supply chain flow, for example, you can say, well, we have an out-of-stock agent and an inventory agent and availability to promise agent, wow, it looks like the availability to promise agent is getting hit by a lot of people.

The usage intensity is really high, and it's costing us a lot of money and actually impacting ROI on the overall process. So you can actually zero in on that, go in and understand what the model is underneath that agent, fine-tune the model, optimize the process and streamline it. And you can actually do this as a part of the dev process even before you decide to say, I'm going to take these agents -- production in the wild.

So we could meet with a large insurance company who wants to reinvent claims processing with AI. And before they have to commit to some massive body of work, we can actually run water through, here's what an AI claim cost to process per claim.

And if you're going to do 100 million claims in a year, great. This is the bill. Here's the ROI. You actually see these things upfront, manage the security aspects of the same. It all binds to the same identity platform that you use for your employees and your contingent staff, you can register your agents that way. And you can use all of our data classification tools to make sure that the data over which those agents are reasoning is what you, in fact want. And so this idea of providing observability at every layer of the stack with transparency around the ROI for AI is also something we think is going to be a huge lift for the acceleration of real AI adoption because if you can build with intelligence and then trust the process and understand the ROI before you invest, we think it's a pretty big unlock.

Yes. And talk about trust, like Craig, when we made the decision to roll out Copilot, to me, that was probably the single most kind of project I've seen in Barclays since I've been here, and that's kind of many years...

Aided by my friends on the left here.

Yes, yes. Can you a little bit -- can you talk to that like it seems like there was an urgency that I haven't seen before. What was the thinking here? And what are you seeing in terms of outcomes so far?

Sure. So look, I think when we started with Copilot, we started, as you do with often with new technologies, which is sort of a small pilot to understand impact. I think what was really surprising, Raimo, was the groundswell of interest in being involved more so than I think any other technology that I've had the opportunity to bring into an organization.

What we learned from those early pilots is that when we put the technology in the hands of colleagues, they did innovate. They did find new ways of working for themselves and their teams. And that gave us confidence to scale and make the commitment to enabling 100,000 colleagues across the organization. Now with that commitment then came also a challenge, which is how do you create across the broader organization what we actually saw in those early adopters and innovators?

How do you inspire the same enthusiasm, the same curiosity, the same desire to discover the potential and capability of this technology on a much, much larger scale? So we lent into that really hard. We ran 3 global hackathons, 6,000 colleagues involved in those, another 8,000 that we didn't have capacity for. And I say that just as a measure of the scale of interest across the organization.

That wasn't just tech, that was organization-wide. We ran escape rooms, prompt the fonds, hundreds of demos and fostered quite a sort of a vibrant community of interest across, firstly, teams and then Viva Engage. So a lot of energy in creating sort of not just skill and understanding, but also the mindset shift and holding up role models within the organization.

Where are we today according to Viva Insights, we've just crossed 1 million hours of productivity through that process. But actually, the other measure, I think, is just the demand that continues to exist. And that demand has meant that the enablement that was to run to the middle of next year will now be concluded at the end of this year. Where to next? Look, it is about continuing to inspire, continuing to educate, hold up the role models, putting new capability, Copilot agents into the hands of those early adopters and innovators and seeing what they can do with autonomous workflows in support of themselves and their teams.

Judson, is it -- are we like a typical customer there? And then the other thing I had to -- that one is like it does feel like the Copilot initially was seen as like, okay, I ask a question and do something more, but it -- now like it feels much, much broader. And I'm only now realizing, okay, this is actually the gateway to everything.

No. Look, we've had a very strong relationship for a long time. And Craig and I have done a lot of work together across our teams. The partnership has been strong because of the feedback loop between the organizations.

I think Craig has been very strong at helping us even taste make the product itself. So Barclays on that side of things is a strong early adopter. I think the other thing just to even kind of come back to the Frontier concept, it's no accident that Craig is now in the role that he's in, right?

Because you move from just leading technology to leading the business and applying AI for the outcomes is frankly where we're all headed. So if you take that back to the comparison of the Copilot journey itself, you were to make an abstraction, Copilot is for AI like the iPhone is for personal computing or like Windows was for the PC. It's a platform.

It's designed to drive a lot of great personal productivity, and it is achieving that. It's the fastest-growing product we've ever launched with the highest utilization. Agents are basically like the apps on your iPhone. They provide that accelerant into the process that you're trying to achieve, the ambition you're trying to unlock, the creativity or the innovation that you're trying to go and pursue.

And so this next step that Craig talks about connecting agents back into the flow of how humans get things done, how they innovate. That is where we are right now, sort of the apex of where we are with the bulk of our customers. We're super excited about the adoption of Copilot. We think we're just getting started in terms of the real AI unlock because if you couple human ambition with Copilot and an agent ecosystem, that's really the formula for really driving true frontier transformation across every facet of the business across all industries.

Yes. And then last question for me, and then I have to let you go. Time fly -- flew by here quickly. if you look out the AI potential that you see out there, how do you think about kind of adoption curves, but also kind of the next steps that are coming out of that? And it's more like we saw a Copilot, but like what's the stuff that we are not thinking about at the moment?

I think it actually comes down to this reinvention of the business that has to be pursued, right? And so there is so much we can achieve with the technology that's less of the question and more of the applied use of the same. I'm very excited about where we're headed with these Agentic business processes with empowering our customers with this intelligence layer that allows people to build those more effectively, more efficiently and with greater confidence.

And then at the same time, the counterbalance on observability so that there are some predictions on the ROI and the outcomes, I think that's a huge confidence booster for business and the adoption cycles. And I do think you'll see a lot more definitive cases of real growth and abundance versus simply efficiency and productivity.

Yes. Craig, for you, same question?

Yes. So look, I think it's redefining what digital transformation means for organizations like ours and increasing its urgency. It is a uniquely powerful capability that changes everything. It changes how we think about designing personalized intelligent digital experiences. It changes how we think about creating autonomous workflows. It changes how we think about modernizing our technology.

And I think it's both distinctive but also new. So there's a sort of a window of opportunity at the moment to differentiate. And I don't think anyone can afford to be left behind, which is why you see the greater urgency. But I think it requires some really important shifts. Judson called out the one that I think is absolutely right, which is this sort of frontier transformation. It's, hey, look, pick 3 or 4 key journeys, 3 or 4 end-to-end processes, go deep, not experimentation, real transformation. And in that, solve not just with AI, but the integrated capability of digital data and AI because the 3 in combination are far greater in capability.

Then I think this issue of just preparing the organization for adoption. The technology is ready. The question is, is the organization ready? And I think that's a different question. It comes down to how you get the technology in the hands of people, how you create the right climate, the right mindset, how you start to educate leaders to manage teams that are now hybrid teams of agents and individuals.

And then this sort of question of just getting the foundations right, like -- getting infrastructure right, getting data right, getting security right, none of that stuff is glamorous. But ultimately, it will set apart those that actually pilot AI and those that really scale the capability, and I think they're critical.

Yes. But it sounds like an exciting journey. And I'm glad I'm on as well. Thank you.

Thanks.

Thank you.

Okay. Let's get started. I'm Karl Keirstead covering Microsoft here, and we are always so honored to have Microsoft keynote our event. I was saying to Rajesh, it wouldn't be the same without Microsoft up on stage. So Rajesh, thank you so much for flying in for this.

Well, thank you so much for having me, Karl.

Yes. Rajesh, when I went online to look at all the parts of Microsoft that report up to you, it took me a little while because there are so many, but I summed up the revenue stream, and it was a very, very large number. So a very big part of Microsoft's portfolio sums up to this gentleman. Can you talk a little bit about what some of the common threads are between the different parts of the portfolio that report into you?

Yes. So I lead the experiences and devices at Microsoft. It's got Office, Teams, so M365. It has our business applications, Dynamics, Power platform, Windows, Surface and then, of course, M365 Copilot. And the common theme, Karl, for experiences and devices, it's what the name suggests, the experiences and devices for information workers on the globe to allow them to be productive, to be collaborative and ultimately drive business outcomes, economic opportunity. So it's a per user. Think of it as we are focused on the users in these organizations, small or large.

Got it. That makes sense. So I think the -- there are several areas we're going to talk about, but clearly, one of the largest pieces is the M365, Office 365 portfolio, where it's still amazing to me that, that business as large as it is and as ubiquitous a product it is, is still growing by 15%. And Amy gave some good stats around seat growth and ARPU growth. But maybe you could just describe a little bit about the underlying growth drivers to get to that 15%. What's really behind the numbers, Rajesh, and in a way that might help us get confidence in the durability of, let's say, a teens growth rate?

Well, M365, as we've said, is we are upwards of 400 million subscribers, paid subscribers. And the seat growth, we continue to see a healthy seat growth, even though we are mature in the enterprise space, small businesses, medium-sized businesses, first-line workers continue to drive seat growth for us. But the biggest drivers of the business now are really on the ARPU growth side. And the ARPU growth are primarily driven out of customers, 3 things that drive the ARPU growth. The first is customers choosing all of our suite, which is they may have in different flavors, different offers and they go to M365, the entire suite. And then the premium offering of M365, which is our E5 offering is probably the largest driver of the ARPU growth. And then increasingly, and in fact, the second largest driver of the ARPU growth is Copilot. So we feel good about continued seat growth. We feel good about the ARPU growth. And yes, it's a large PxQ business.

Let's -- we'll talk about a few of those things, but maybe let's start with the seat side. So I think in the most recent quarter, you guys announced 6% seat growth. Again, still amazing given the scale of that business. Obviously, investors for all SaaS businesses are worried about headcount cuts. Now the pace at which that seat growth is declining to me, speaks more to just scale and maturity. I don't really see any strong evidence that there's sort of an AI-induced seat pressure in that number yet. Do you mind opining a little bit on that dynamic?

No, we don't see that either. I'm not seeing AI as driving down seats in M365. If anything, I think AI is going to be an opportunity for us to drive seat growth. If you think about an organization in the future, it's going to have more agents than people. Many of these agents are going to be embodied. They're going to have an identity. They're going to be in the address book. They're going to have a mailbox. They're going to need a computer to do its computation in a secure way. You want to mention them. It's going to want to chat with you. It's going to want to join a meeting. To me, all of those embodied agents are seat opportunities.

Yes. And then on the ARPU lift coming from customers embracing M365 and in particular, E5, how far into that opportunity, Rajesh, are we? And if you could remind the audience, is that still a security-led decision given the number of security features you're putting in E5?

The upside on E5 varies vastly by segment. First-line workers, less so, but in the enterprise, in the midsized markets, continue to see opportunity for E5 penetration. Then in terms of is security the biggest driver? I would say, yes. And increasingly, what we call compliance in a world of AI, data classification, data governance has become incredibly important. So the purview aspect of E5 in addition to the defender aspect of E5 is also starting to be a meaningful driver and a pull.

As customers want to go more broadly with AI and agents, they want to get the data governance of their data estate in a way that they can set the right policies for what's accessible to AI and what are the handling policies for AI and which so -- it's security end.

Got it. And Rajesh, you mentioned that now M365 Copilot might rank behind the E5 adoption as key driver of ARPU growth. So given the importance of that, let's talk a little bit about that. Microsoft obviously hasn't yet, at least maybe someday you will, give a metric on Copilot such as seat growth, et cetera. So what's your advice to the audience in terms of what we on the outside can monitor to get some sense of the traction that Copilot is having? Is it more anecdotal comments sort of ranking it as a key driver of ARPU? Is that it? Or is there anything else, Rajesh, you recommend we watch for?

We continue to consider what else we would share and when. But there are a couple of things I can point to. As we recently disclosed, 2 quarters in a row, the daily active engagement with Copilot has more than doubled quarter-over-quarter. That, to me, is a real sign of product with increased intensity of usage and more users engaging with the Copilot experience.

Another one I would point to, maybe a couple of quarters ago, we said, what, 70% of the Fortune 500 had Copilot. Now that's up to 90%. And I do think the ARPU growth is a good one to think about. As we discussed, most of our seat growth today is in lower ARPU base, small businesses, first-line workers. And the fact that our ARPU growth has been stable, Copilot is a contributor to that. So I think the ARPU growth rate is it's a good one to watch. Our engagement rate is a good one to watch.

Okay. That usage improvement, I can tell you, I don't think you were in the audience, but maybe some Microsoft members were. I did a panel this morning with a couple of the lead IT executives at UBS. As you know better than anybody, we've become a very large Copilot customer. So I asked our executives this question. And I think the stuff they shared with me was that UBS employee Copilot usage is up something like 9x on a year-over-year basis. So that's a testimony to what you're describing in terms of usage improvement.

Yes, yes, we see that.

Okay. Is there anything Microsoft can do to help large Copilot customers like UBS drive usage? How is Microsoft helping us get there?

Yes, there's a ton we can do and a ton that we are doing and continuing to invest in, and we learn from our customers what else we have to go do. But let me just talk about a few of those things. The first, of course, is our customer success teams, all the way from deployment to change management, to scenario planning to business value workshop. And by the way, we've decided deliberately to have a bunch of that CSM or customer success management to be core out of the core product team. So the feedback loop between customers and the product team stays super high.

The second one, I would say, as more and more customers, including UBS, are starting to go beyond Copilot as an assistant to agentic workflows. We have FDEs, our forward deployed engineers, again sitting out, of my team, working with customers for the first 5 of the first 10 of these agents that we pick with the customer on either top line drivers or real cost reducers or efficiency or quality drivers. And then we are working with our SI partners so they can scale out and do some of the more -- some of the same.

A third thing that we're building into the product itself is called Copilot Analytics because we want customers to actually be able to measure the ROI and see the business value. So the Copilot Analytics, customers can now do cohort analysis. They can give Copilot to a bunch of users, have a control group and do an agent for one group, not do an agent for the other group. Join the metrics that we see in terms of collaboration and productivity and join that with their own KPIs and see what the ROI is and see what drives the usage at what time across which application. So we want to build more of that stuff in the product itself, and that's what the Copilot Analytics is about.

And finally, I think a big unlock increasingly is going to be agent governance and not just agent governance, broadly governance, even with UBS. Should they enable web grounding for Copilot or should they not enable web grounding for Copilot. So there are a set of controls they would like. When I issue a prompt, what part of the prompt actually leaves the compliance boundary? Do we have the right checks and balances? Because every deployment has to pass risk and governance and security. So we are now year 2 of this thing, and we are getting more and more mature in the governance. And likewise, on agent governance, it's a huge ask from customers.

So I've never seen an announcement in this realm generate the sort of excitement with the IT community that we had at Ignite, which is Agent 365. Because the big worry for customers is, hey, do we democratize the creation of agents? If so, do I have an agent sprawl where it's like the -- all Access database sprawl or VB Macro sprawl. With Agent 365, we give them one observability plane, one data classification plane, one identity plane, whether the agents are built by Microsoft, built by a third party. So we have a lot of partners in the state. So Agent 365 is going to be the other big product investment that we are making that is going to go and enable customers to go bigger with these things.

And Rajesh, in terms of other drivers that might -- the catalyst for unlock needs to come from your partners, not from Microsoft, is the model performance. Yes. I think a lot of us listened to Sam Altman a couple of weeks back and a pod he did. Where he was actually almost admitting that OpenAI needs to make the models even better as an unlock to get more people using AI applications like Copilot. So where do you think we are on the model evolution? And do we need to wait essentially for GPT-6 to get the next big step function improvement in Copilot that would then drive even more usage? How dependent do you think the usage trajectory is on underlying model performance?

I mean, clearly, if the model gets better, the experiences are going to get better. If architected correctly, you want your experiences to be modeled forward. The model gets better, the experiences get better. That's what modern experiences have to be crafted to go to. But I think there are 3 things that are already elevating the capabilities of Copilot and those -- I think the first is the reasoning model finally unlocks AI and graphical user interfaces to work much more effectively.

So I'm very excited in the time -- in the January, February time frame, you're going to see agent mode inside of Office, inside of Excel, inside of Word, inside of PowerPoint, in your calendar, in your meeting, inside your Teams channel. And that is going to be a big one. It's one thing to go and say there's AI off on the side, which I think is a very interesting user experience because you get to express your intent in a natural language. But then to be able to take that same capability and bring that to existing workflows that hundreds of millions of people are already on with a graphical user interface and have those 2 coexist, I think the agent mode is going to be a big unlock. So that's thing one.

The second thing is what we talked about, enabling more agentic workflows. And this is where Agent 365 as a governance and a compliance plane enables customers to go without the worry of a sprawl or a management issue on their hands. And finally, with Copilot, we brought multiple sources of intelligence into Copilot. Just like if I hire an employee with liberal arts background to work on a project and I hire an employee with a STEM background to work on a project, they're going to bring different strengths. And so with Copilot, it's not just about the model. It's about how the models come together with the graphical user interface, the innate capabilities of the model, multiple lineages of the model to actually deliver the outcome or the workflow that you're planning to go do, whether in our applications or agents that you will build on your own.

So let's press maybe a little bit on the enterprise Agentic adoption, which you're describing as some interesting things coming that could be an unlock. So Rajesh, we do talk to some people who are more at the skeptical end of the spectrum where they say that it's going to be a long haul to get companies really embracing agents, especially agents acting relatively autonomously. You sound a lot more optimistic. What's the gap here that gives you that confidence? And where do you think generally enterprises are in investing heavily in Agentic apps?

I mean for sure, I mean, I will agree with the fact that it's not a uniform thing. I mean any enterprise -- any new technology goes through its enterprise curve with the risk and management, ROI analysis, business value, all of that stuff. That being said, let me just say, I mean, the word agent is so overused. I'd like to think about agents as 3 -- think of a spectrum. On one end of the spectrum, you've got what used to traditionally be an end user going and creating a great template or an access database or an Excel, if I'm a lawyer doing IPE rebuttals, I have a 20-page contract as a template that my work group uses. Those things can be actually implemented very capably as an agent, built at the departmental level by a power user and that kind of a thing today has been really bottled up by a lot of customers because of the governance issue. And the Agent 365 really unlocks that because now you have governance, you have -- you can see what's been shared by who, how often is it used, what should get aged out.

So that's one end of the spectrum. And by the way, companies are going to have to enable that because no top-level IT or business leader of a large organization have visibility into the workflows that impact the daily lives of a lot of information workers. So anyway, that's one end of the spectrum. The middle end of the -- in the spectrum is classic high-value applications. And these are applications that might be really important from a cost perspective, revenue perspective, quality perspective. And then agents are a great implementation. Take the GenAI engine, get the right grounding, right enterprise policies and you go enable that.

Now let me give you an example of such a thing that we talked about at Ignite finally getting to general availability with huge excitement by our customers. It's called employee self-service. Think of it as a template for an agent. Every customer that we talk to, every large organization has help desk for IT. They have HR tickets coming in. They have people making requests for conference rooms, event management, all of those things. And so this is broadly applicable having an agentic workflow that can take all your business policies, quickly be able to tell you, hey, your kid is getting 26 that need to come off your medical plan. It is trivial for such an agent to get the right one click, let me go take care of that when you go to this agent.

But this is not a simple agent. And I was telling somebody like if you're an employee and you ask, hey, what's the paternity leave policy and you happen to be in Norway, HR is very particular for that particular thing. There needs to be no AI. It needs to be the authoritative HR policy that needs to be given back to the employee. If somebody files a ticket saying, "Hey, I'm depressed, I'm looking for resources." You don't want HR adjudicating. You want that employee to be connected to a human.

So the ESS is kind of a broadly applicable app pattern, agent pattern that we are starting to go build to enable and bootstrap this kind of an agent deployment. And then the other end of the spectrum of agents are digital workers. They're going to show up in your org chart. They're going to show up in a meeting roster. They're going to show up as a part of the channel member. You're going to mention somebody who's a new digital marketing employee in your marketing group to go do research and give you back a document that they post back into the Teams channel. That is probably the most immature end, but 2026, you're going to start to see those things happen as well.

Okay. That's exciting. Let me ask you a couple of more questions about Copilot, then we'll move on to other parts of your portfolio. One of the interesting announcements that Microsoft made recently was that, well, Copilot has to date been run largely on OpenAI models. You announced a partnership where customers can access essentially an Anthropic version of Copilot. Can you describe that decision? What's the thinking behind that?

Yes. Again, I mean, with Copilot, our goal here is to make sure you can get to the outcome that you were trying to get done as an information worker or in the context of a business process. And we noticed that the Anthropic model does a pretty good job with different perspective. If I ask OpenAI model to do research on a given topic, it tends to be well written, well crafted. If I ask the Anthropic model to do the same thing, I'm going to see many more charts, it's going to be much more concise, neither is better than the other.

And so I want to give a choice to the user as to what works best for them. And they could run both, pick the one that they want. And so it's no different than like I said, if you hire 2 employees in your work group and they come from 2 different backgrounds, 2 different schools, 2 different skill sets, they're going to do different -- they're going to do very competent work, but it's going to be different. And so for us, it's about the best possible outcomes.

And by the way, it's not just the Anthropic model, the OpenAI model. I think one of the more interesting things that we've seen customers go do is if you have a business process that is unique to you, that is your differentiated IP as a financial services and insurance company as a pharmaceutical company, you want the model to be tuned to your IP with the weights belonging to you, not the weights collapsing into the model.

And so that's what Copilot tuning is about. We'll let customers take a base model, tune it with no Microsoft eyes on it with their subject matter experts, create a fine-tuned model, measure the efficacy of the model against the base model and then deploy it back into the Copilot experiences that our employees are going to go use. So it's less about the model. It's making sure that the entire experience is optimized for the outcomes you're trying to drive.

Okay. Last one I want to ask you on Copilot is around competition. So I'm sure when you and the leadership team launched M365 Copilot, you were under no illusion that you would have that market to yourself. It's always competitive in every corner of the software space. But your -- one of your key partners, OpenAI now is trying to scale into their targets, partly on the back of ChatGPT for enterprise, which many organizations might use alongside Copilot, maybe in some cases, instead of. Google with Gemini now is pushing Gemini for enterprise. So getting a little bit more competitive. What are your thoughts, Rajesh, on the ability of Microsoft to stay ahead of those 2 rivals in particular?

It's a great question. So let me first just say, look, I mean, with -- let me talk about Copilot. What's differentiated, what's unique. The first thing we got to do is to make sure we meet users' expectations with what they expect generative AI to do, be a great writer, be a great analyzer, be able to tell you the weather, create an image and all of that stuff. So that's table stakes. We're going to go do that, and we're going to be competitive like any other Gen AI tool on that. Now let's talk about what's unique and what's differentiated. The first thing is what we call Work IQ, which is understanding your work context.

Most of the people powering economic activity around the globe are on M365. So Work IQ gives us a pretty good understanding of the projects that matter to you, who your work group is, what events are interesting, who are your customers, your e-mails, your meetings, your documents. We have a pretty good understanding of that stuff. Then you'll hear the other tools talk about connectors. Connector is trying to drink through a very thin straw and understand your work context. Now you can take many sips from a thin straw, it is still a thin straw. And so we have customers who've done side-by-side analysis of Copilot in the context of your work and connector-based things and the gap is very significant and growing.

So anyway, that's just thing one. Thing two, though, as people use Copilot both as a new endpoint as a peer to what Excel, PowerPoint, Outlook, Teams, as a pure endpoint. But they have billions of transactions every hour that go through these applications. We have an opportunity to serve the same Work IQ intelligence with agent mode inside of these existing applications that people use, we can uniquely do that well. That's thing two.

Thing three, human-to-human collaboration, communication, productivity, highly regulated. We have 15 years' experience on how to do eDiscovery, sensitivity label, all sorts of things, legal hold, data classification, data governance. All of that works with human to AI conversations in Copilot automatically for customers. Our 15 years of maturity on that stuff is enterprise-grade maturity and human-to-human productivity. Human and AI productivity gets all of that for free. The other providers are going to have to go build all of that capability. Thing 4, can I keep going or you can keep coming.

Yes. So thing 4, though, it's like can I go back to we are not beholden to a model. We are beholden to the best outcome. And so we will build experiences that are powered by whether it be our IP partner, OpenAI, whether it be Anthropic, whether it be copilot tuning, whether it be an open source model whose weights you want to go create as your own unique IP. So we are multi-model.

Finally, the more important thing that I go back to the comment I made earlier, if you think that the future workforce is going to have more agents than humans, and this is going to be true in every organization. How is the human supposed to get work? I thought that the beauty of AI was I don't have to worry about tools anymore. I have one natural language interface that abstracts away the tools, and I express my intent.

In a world where now I'm surrounded by hundreds of agents that are semiautonomous, some that I'm managing, some that I have to rendezvous with, how does a human navigate all of that stuff? Copilot is going to be the UI through which they orchestrate the agent. We are going to be the search engine for these agents. We're going to be the relevance engine for these agents. We're going to be the orchestration engine for these agents. IT is going to have a bunch of controls. Karl, you're in this department. The following agents are the most relevant to you. You work group created an agent that you tend to use a lot. Copilot will know to use that. So when you put all of these things together, I really like our ability to go serve M365 customers with M365 Copilot.

Yes. Rajesh, I had so many other questions on other parts of your portfolio like Teams and Windows that I don't even have time for. But I noticed we have 2 minutes left, and I do want to get to one last one that I know is of interest to this group. And that is the compute capacity. So Microsoft obviously needs to stand up an enormous amount of compute capacity for multiple constituents. You've obviously got partners like OpenAI that have demands on you. You've got enterprise customers like UBS that over time, if not today, need Azure-based compute capacity.

Absolutely.

But this conversation is about really Microsoft's first-party applications because you need to ensure that the Copilot and Agentic AI experience is robust. So you need to make sure that you've got enough compute capacity to serve your first-party apps. So I guess the question is, how big a constraint is that? How big of a problem is it given that Microsoft, as Amy described to all of us, is very compute constrained right now. So how do you ensure, Rajesh, that you knock on Amy's door and make sure that she's carving out sufficient compute for your business?

Yes. You're right, Karl. I mean this is a multidimensional thing that we spend a lot of time on. But holistically, it's what you would expect. We take a left to right view across our third-party commitments, our first-party commitments. Inside of the first-party commitments, clearly, M365 Copilot, GitHub Copilot, these are the huge priorities. But it isn't as simple as which is priority 1, which is priority 2. There are multiple dimensions to this. I mean, not all scenarios need all sorts of hardware. Not all load happens at the same time across the globe.

We have different geo ring-fencing constraints in different parts of our businesses. Saturday morning peak, Saturday, 3:00 a.m. is our lowest use time. What parts of the AI can be done asynchronously during that time? How much latency is affordable for which segment? How much optimization can we bring out, not just at the chip level, but in the global routing level to optimize for all of these things. And so we work this multidimensional thing even as we add more and more capacity, and we look left to right against all the prioritization. So suffice it to say, this is top of mind, and I feel good about the process we have in place to meet all the stakeholders and navigate that.

Got it. Okay. I think that's all the time we have. Rajesh, I learned a lot. Thank you. And I'm very proud to work for what's now become a pretty significant Microsoft Copilot and Azure customer in...

Keep the feedback coming, Karl.

Thank you. Thank you so much.

Thank you, Karl.

Greetings, and welcome to the Microsoft Fiscal Year 2026 First Quarter Earnings Conference Call. [Operator Instructions] As a reminder, this conference is being recorded.

It is now my pleasure to introduce Jonathan Neilson, Vice President of Investor Relations. Please go ahead.

Good afternoon, and thank you for joining us today. On the call with me are Satya Nadella, Chairman and Chief Executive Officer; Amy Hood, Chief Financial Officer; Alice Jolla, Chief Accounting Officer; and Keith Dolliver, Corporate Secretary and Deputy General Counsel.

On the Microsoft Investor Relations website, we will provide our earnings press release and financial summary slide deck, which is intended to supplement our prepared remarks and provides the reconciliation of differences between GAAP and non-GAAP financial measures. More detailed outlook slides will be available on the Microsoft Investor Relations website.

On this call, we will discuss certain non-GAAP items. The non-GAAP financial measures provided should not be considered as a substitute for, or superior, to the measures of financial performance prepared in accordance with GAAP. They are included as additional clarifying items to aid investors in further understanding the company's first quarter performance in addition to the impact these items and events have on the financial results.

All growth comparisons we make on the call today relate to the corresponding period of last year, unless otherwise noted. We will also provide growth rates in constant currency when available as a framework for assessing how our underlying business performed, excluding the effect of foreign currency rate fluctuations. Where growth rates are the same in constant currency, we will refer to the growth rate only.

We will post our prepared remarks to our website. Today's call is being recorded. If you ask a question, it will be included in our live transmission, in the transcript and in any future use of the recording. You can replay the call and view the transcript on the Microsoft Investor Relations website.

During this call, we will be making forward-looking statements, which are predictions, projections or other statements about future events. These statements are based on current expectations and assumptions that are subject to risks and uncertainties. Actual results could materially differ because of factors discussed in today's earnings press release, in the comments made during this conference call and in the Risk Factors section of our Form 10-K, Forms 10-Q and other reports and filings with the Securities and Exchange Commission. We do not undertake any duty to update any forward-looking statements.

And with that, I'll turn the call over to Satya.

Thank you, Jonathan. It was a very strong start to our fiscal year. Microsoft Cloud revenue surpassed $49 billion, up 26% year-over-year. And our commercial RPO increased over 50% to nearly $400 billion with a weighted average duration of only 2 years. We are seeing increasing demand and diffusion of our AI platform and family of Copilots, which is fueling our investments across both capital and talent.

When it comes to infrastructure, we're building a planet scale cloud and an AI factory, maximizing tokens per dollar per watt while supporting the sovereignty needs of customers and countries. We are innovating rapidly across the family of Copilots spanning the high-value domains of information work, coding, security, science, health and consumer.

And as you saw yesterday, we closed a new definitive agreement with OpenAI, marking the next chapter in what is one of the most successful partnerships and investments our industry has ever seen. This is a great milestone for both companies and we continue to benefit mutually from each other's growth across multiple dimensions. Already, we have roughly 10x-ed our investment. OpenAI has contracted an incremental $250 billion of Azure services, our rev share, exclusive IP rights and API exclusivity for Azure continue until AGI or through 2030. And we have extended the model and product IP rights through 2032.

And we are also energized to innovate and pursue AI advancements with both talent and compute investments that have real-world impact. With that, let's turn to our momentum across our AI platform and Copilots as well as with agents.

We have the most expansive data center fleet for the AI era, and we are adding capacity at an unprecedented scale. We will increase our total AI capacity by over 80% this year and roughly double our total data center footprint over the next 2 years, reflecting the demand signals we see. Just this quarter, we announced the world's most powerful AI data center, Fairwater in Wisconsin, which will go online next year and scale to 2 gigawatts alone.

And we have deployed the world's first large-scale cluster of NVIDIA GB300s. We are building a fungible fleet that's been continuously modernized and spans all stages of the AI life cycle from pretraining to post training to synthetic data generation and inference. And it also goes beyond GenAI workloads to recommendation engines, databases and streaming. We're optimizing this fleet across silicon systems and software to maximize performance and efficiency.

It's this combination of fungibility and continuous optimization that allows us to deliver the best ROI and TCO for us and our customers. For example, during the quarter, we increased the token throughput for GPT-4.1 and GPT-5, two of the most widely used models by over 30% per GPU. We also have the most comprehensive digital sovereignty platform. Azure customers in 33 countries are now developing their own cloud and AI capabilities within their borders to meet local data residency requirements.

In Germany, for example, OpenAI and SAP will rely on Azure to deliver new AI solutions to the public sector. On top of this infrastructure, we are building Azure AI Foundry to help customers build their own AI apps and agents. We have 80,000 customers, including 80% of the Fortune 500. We offer developers and enterprise access to over 11,000 models more than any other vendor, including as of this quarter, OpenAI's GPT-5 as well as xAI's Grok 4. For example, Ralph Lauren used Foundry to build conversational shopping experience in its app, enabling customers to describe what they're looking for and get personalized recommendations. And OpenEvidence used Foundry to create its AI-powered clinical assistant which surfaces relevant medical information to physicians and help streamline charting.

When it comes to our first-party models, we are excited by the performance of our new MAI models for text, voice and image generation, which debuted among the top in the industry leader boards. And we continue to make great progress with our Phi family of SLMs, which now have been downloaded over 60 million times, up 3x year-over-year.

Beyond models in Foundry, we are providing everything developers need to design, customize and manage AI applications and agents at scale. Our new Microsoft Agent Framework helps developers orchestrate multi-agent systems with compliance, observability and deep integration out of the box. For example, KPMG used the framework to modernize the audit process, connecting agents to internal data with enterprise-grade governance and observability. These kinds of real production scale AI deployments are driving Azure's overall growth. And once again, this quarter, Azure took share.

Now let's turn to applications and agents we ourselves are building on this platform. We now have 900 million monthly active users of our AI features across our products. And our first-party family of Copilots now has surpassed 150 million monthly active users across the information work, coding, security, science, health and consumer.

When it comes to information work, we continue to innovate with Microsoft 365 Copilot. Copilot is becoming the UI for the agentic AI experience. We have integrated chat and agentic workflows into everyday tools like Outlook, Word, Excel, PowerPoint and Teams. Just 9 months since release, tens of millions of users across Microsoft 365 customer base are already using chat. Adoption is accelerating rapidly, growing 50% quarter-over-quarter, and we continue to see usage intensity increased.

This quarter, we also introduced Agent Mode, which turns single prompts into export quality Word documents, Excel spreadsheets, PowerPoint presentation and then iterate to deliver the final product much like agent mode in coding tools today. We're thrilled by the early response, including third-party benchmarks that rank it best-in-class.

Beyond individual productivity, Copilot is multiplayer, with Teams mode announced this week, you can now invite colleagues into a Copilot conversation. And our collaborative agents like Facilitator and Project Manager, prep meeting agendas, take notes, capture decision, kick off group tasks.

We are seeing a growing Copilot agent ecosystem with top ISVs like Adobe, Asana, Jira, LexisNexis, SAP, ServiceNow, Snowflake and Workday, all building their own agents that connect to Copilot. And customers are also building agents for their mission-critical business processes and workflows using tools like Copilot Studio and integrating them into Copilot. The overall number of agent users doubled quarter-over-quarter. And just yesterday, we announced App Builder, a new Copilot agent that lets anyone create and deploy task-specific apps and agents in minutes grounded in Microsoft 365 context.

All this innovation is driving our momentum. Customers continue to adopt Microsoft 365 Copilot at a faster rate than any other new Microsoft 365 suite, all up more than 90% of the Fortune 500 now use Microsoft 365 Copilot, Accenture, Bristol-Myers Squibb, EY Global and the U.K.'s Tax and Payments and Customs Authority all purchased over 15,000 seats this quarter. Lloyds Banking Group has deployed 30,000 seats, saving each employee an average of 46 minutes daily.

And a large majority of our enterprise customers continue to come back to purchase more seats. Our partner, PwC, alone added 155,000 seats this quarter and now has over 200,000 deployed across its global operations. In just 6 months, PwC employees interacted with Microsoft 365 Copilot over 30 million times, and they credit this agentic transformation with saving millions of hours an employee productivity.

When it comes to coding, GitHub Copilot is the most popular AI pair programmer now with over 26 million users. For example, tens of thousands of developers at AMD use GitHub Copilot, accepting hundreds of thousands of lines of code suggestions each month and crediting it with saving months of development time. All up, GitHub is now home to over 180 million developers and the platform is growing at the fastest rate in its history, adding a developer every second. 80% of new developers on GitHub start with Copilot within the first week.

Overall, the rise of AI coding agents is driving record usage with over 500 million pull requests merged over the past year. And just yesterday, at GitHub Universe, we introduced Agent HQ. GitHub Copilot and Agent HQ is the organizing layer for all coding agent, extending GitHub privatives like PRs, issues, actions to coding agents from OpenAI, Anthropic, Google, Cognition, xAI as well as OSS and in-house models. GitHub now provides a single mission control to launch, manage and review these agents, each operating from its own branch with built-in controls, observability and governance.

We are building a similar system in security with over 3 dozen agents in Copilot integrated across Entra, Defender, Purview and Intune. For example, with our Phishing Triage Agent in Defender, studies show that analysts can be up to 6.5x more efficient in detecting malicious mails.

In health, Dragon Copilot helps providers automate critical workflows. This quarter alone, we helped document over 17 million patient encounters, up nearly 5x year-over-year. More than 650 health care organizations have purchased our ambient listening tech to date, including University of Michigan Health where over 1,000 physicians are actively using it.

Finally, when it comes to AI consumer experiences, we are excited about all the progress Copilot is making, starting with Windows. Every Windows 11 PC now is an AI PC. Just 2 weeks ago, we introduced new ways to speak naturally to your computer, including a Copilot wake word. With vision, Copilot sees what you see on your screen and you can have a real-time conversation about it. And with action, it takes real action on your behalf, interacting with both web and desktop apps.

In Edge, we are introducing first-of-its-kind AI features to automate multistep workflows within the browser and help you pick up right where you left off. Edge now has taken share for 18 consecutive quarters. In Bing, our overview pages now include embedded conversational capabilities, we took share again in search. And daily users of our Copilot consumer app increased nearly 50% quarter-over-quarter. Among many updates we made last week is groups, which turns Copilot for the first time into a shared experience.

We also are creating a great consumer subscription offer with Microsoft 365 premium. It brings together our Office applications and advanced Copilot features with high usage limits, giving individuals the flexibility to bring their own AI to work in a secure way.

Finally, in gaming, Copilot provides a voice-first immersive experience across PC, mobile and our new Xbox Ally.

Beyond our family of Copilots and AI platform, we are seeing strong momentum across the portfolio. Cloud migrations are accelerating. In data and analytics, Fabric revenue grew 60%, which is faster than any other data and analytics platform in the industry. We now have 28,000 paid Fabric customers. In databases, SQL, DB, hyperscale revenue was up nearly 75%, 50% in Cosmos DB. In Business Applications, Dynamics 365 gained share. In security, our end-to-end stack is now informed by 100 trillion daily signals. 1 billion monthly active users of Entra, 16 billion Copilot interactions audited by Purview, up 72% quarter-over-quarter. 40,000 Sentinel customers, and we took share across all categories we serve in security. In LinkedIn, nearly 1.3 billion members.

And finally, in gaming, we expanded our reach across every endpoint focused on our high-margin content and services. We launched critically acclaimed games like Keeper, Ninja Gaiden 4 and Outer Worlds 2, reaching 155 million monthly active users. Minecraft, an all-time high, and set new record for overall content and services revenue for the quarter. We also saw a great response to Xbox Ally launch 2 weeks ago and set new records for players on PC.

In closing, our planet-scale Cloud and AI factory together with Copilots across high-value domains is driving broad diffusion and real-world impact. And we continue to increase our investments in AI across both capital and talent to meet the massive opportunity ahead.

With that, let me turn it over to Amy to walk through our financial results and outlook, and I look forward to rejoining for your questions.

Thank you, Satya, and good afternoon, everyone. First, as you heard from Satya, we were pleased to announce the next phase of our partnership with OpenAI yesterday. They continue to choose Microsoft to power their workloads. And together, we remain committed to driving innovation that meets real-world needs. Our Q1 results were not impacted by the deal signed this week.

Now on to the quarter, we delivered a strong start to our fiscal year, exceeding expectations across revenue, operating income and earnings per share. We also saw continued share gains across many of our businesses, demonstrating our leadership position in key markets.

This quarter, revenue was $77.7 billion, up 18% and 17% in constant currency. Gross margin dollars increased 18% and 16% in constant currency, while operating income increased 24% and 22% in constant currency. And earnings per share was $4.13 and an increase of 23% and 21% in constant currency when adjusted for the impact of our investments in OpenAI. FX impact was roughly in line with guidance.

Company gross margin percentage was 69%, down slightly year-over-year, driven by investments in AI, including the impact of scaling our AI infrastructure and the growing usage of our AI product features. This was partially offset by ongoing efficiency gains, particularly in Azure and M365 Commercial cloud.

Operating expenses increased 5% and 4% in constant currency, driven by investments in Cloud and AI engineering, including compute capacity and AI talent to support product development across the portfolio. Operating margins increased year-over-year to 49% and were ahead of expectations with stronger-than-anticipated results and high-margin businesses this quarter.

When adjusted for the impact from our investments in OpenAI, other income and expense was $401 million as interest income more than offset interest expense, which includes the interest of payments related to data center finance leases.

Capital expenditures were $34.9 billion, driven by growing demand for our Cloud and AI offerings. This quarter, roughly half of our spend was on short-lived assets, primarily GPUs and CPUs, to support increasing Azure platform demand, growing first-party apps at AI solutions, accelerating R&D by our product teams as well as continued replacement for end-of-life server and networking equipment.

The remaining spend was for long-lived assets that will support monetization for the next 15 years and beyond, including $11.1 billion of finance leases that are primarily for large data center sites. And cash paid for PP&E was $19.4 billion. As a reminder, the difference between total CapEx and cash paid for PP&E is primarily due to finance leases as well as the normal timing of goods received, but not yet paid.

Cash flow from operations was $45.1 billion, up 32%, driven by strong Cloud billings and collections, partially offset by higher supplier payments. And free cash flow increased 33% to $25.7 billion with minimal impact from a sequential increase in CapEx, given the higher mix of finance leases.

And finally, we returned $10.7 billion to shareholders through dividends and share repurchases.

Now to our commercial results. Commercial bookings increased 112% and 111% in constant currency and were significantly ahead of expectations, driven by Azure commitments from OpenAI as well as continued growth in the number of $100 million-plus contracts for both Azure and M365. These results do not include any impact from the incremental $250 billion Azure commitments from OpenAI announced yesterday.

Commercial remaining performance obligation increased to $392 billion and was up 51% year-over-year. The balance has nearly doubled over the past 2 years. And even with this growth, our weighted average duration has been relatively stable at approximately 2 years.

Microsoft Cloud revenue was $49.1 billion, ahead of expectations and grew 26% and 25% in constant currency. Microsoft Cloud gross margin percentage was slightly better than expected at 68% and down year-over-year due to the investments in AI that were partially offset by ongoing efficiency gains as noted earlier.

Now to segment results. Revenue from Productivity and Business Processes was $33 billion and grew 17% and 14% in constant currency. M365 Commercial Cloud revenue increased 17% and 15% in constant currency with 1 point of benefit from in-period revenue recognition. Year-over-year growth was driven by both ARPU and seats with ARPU growth again led by E5 and M365 Copilot. Paid M365 Commercial seats grew 6% year-over-year with installed base expansion across all customer segments though primarily in our small and medium businesses and frontline worker offerings.

M365 Commercial Products revenue increased 17% and 14% in constant currency, ahead of expectations due to higher-than-expected Office 2024 transactional purchasing. M365 Consumer Cloud revenue increased 26% and 25% in constant currency, again driven by ARPU growth. M365 consumer subscriptions grew 7% to over 90 million.

LinkedIn revenue increased 10% and 9% in constant currency driven by Marketing Solutions. The Talent Solutions business was impacted by continued weakness in the hiring market. Dynamics 365 revenue increased 18% and 16% in constant currency with continued growth across all workloads.

Segment gross margin dollars increased 19% and 16% in constant currency, and gross margin percentage increased driven by efficiency gains in M365 Commercial Cloud that were partially offset by investments in AI including the impact of growing usage in M365 Copilot chat. Operating expenses increased 6% and 5% in constant currency, and operating income increased 24% and 20% in constant currency. Operating margins increased 3 points year-over-year to 62%, driven by the higher gross margin noted earlier as well as improved operating leverage.

Next, the Intelligent Cloud segment. Revenue was $30.9 billion and grew 28% and 27% in constant currency. In Azure and other Cloud services, where we continue to see accelerating demand, revenue grew 40% and 39% in constant currency. Results were ahead of expectations, driven by better-than-expected growth in our core infrastructure business, primarily from our largest customers.

Azure AI services revenue was generally in line with expectations, and this quarter, demand again exceeded supply across workloads, even as we brought more capacity online.

In our on-premise server business, revenue increased 1% and was relatively unchanged in constant currency. Results were ahead of expectations, driven by transactional purchasing of Windows Server 2025. Segment gross margin dollars increased 20% and 19% in constant currency, and gross margin percentage decreased year-over-year, driven by investments in AI that were partially offset by efficiency gains in Azure. Operating expenses increased 4% and operating income grew 27%. Operating margins were 43%, down only slightly year-over-year as increased investments in AI were mostly offset by improved operating leverage.

Now to More Personal Computing. Revenue was $13.8 billion and grew 4%. Windows OEM and Devices revenue increased 6% year-over-year, significantly ahead of expectations, driven by strong demand ahead of Windows 10 end of support as well as a benefit from inventory levels that remain elevated.

Search and news advertising revenue ex TAC increased 16% and 15% in constant currency, driven by growth in volume as well as the continued benefit from third-party partnerships that was better than expected. And in gaming, revenue decreased 2% and 3% in constant currency against a strong prior year comparable. Xbox content and services revenue increased 1% and was relatively unchanged in constant currency, driven by better-than-expected performance from third-party content.

Segment gross margin dollars increased 11% and 10% in constant currency, and gross margin percentage increased year-over-year driven by sales mix shift to higher-margin businesses. Operating expenses increased 4% and 3% in constant currency, and operating income increased 18% and 16% in constant currency. Operating margins increased 3 points year-over-year to 30% driven by the higher gross margin noted earlier.

Now moving to our Q2 outlook, which unless specifically noted otherwise, is on a U.S. dollar basis. Based on current rates, we expect FX to increase total revenue growth by 2 points. Within the segments, we expect FX to increase revenue growth by 2 points in Productivity and Business Processes and Intelligent Cloud and 1 point in More Personal Computing. We expect FX to increase COGS and operating expense growth by 1 point.

Starting with the total company, we expect revenue of USD 79.5 billion to USD 80.6 billion or growth of 14% to 16%. We expect COGS of USD 26.35 billion to USD 26.55 billion or growth of 21% to 22%. And operating expense of USD 17.3 billion to USD 17.4 billion, growth of 7% to 8%. Operating margins should be relatively flat year-over-year and down sequentially, aligned with historic seasonality.

Now other income and expense. The combination of OpenAI's conversion to a public benefit corp and the ongoing nature of our partnership will result in increased volatility. Therefore, going forward, we'll provide our outlook, excluding any impact from our investments in OpenAI. On that basis, in Q2, other income and expense is estimated to be roughly $100 million as interest income will more than offset interest expense. And we expect our Q2 effective tax rate to be approximately 19%.

Next, capital expenditures. With accelerating demand and a growing RPO balance, we're increasing our spend on GPUs and CPUs. Therefore, total spend will increase sequentially, and we now expect the FY '26 growth rate to be higher than FY '25. As a reminder, there can be quarterly spend variability from cloud infrastructure build-outs and the timing of delivery of finance leases.

Next, our commercial business. In commercial bookings, we expect healthy growth in the core business on a low expiry base when adjusted for the OpenAI contracts in the prior year. And we expect commercial bookings will be positively impacted by the significant OpenAI commitments announced yesterday. As a reminder, larger long-term Azure contracts, which are more unpredictable in their timing, drive increased quarterly volatility in our bookings growth rate.

Microsoft Cloud gross margin percentage should be roughly 66%, down year-over-year, driven by the continued investments in AI as well as the mix shift to Azure.

Now to segment guidance. In Productivity and Business Processes, we expect revenue of USD 33.3 billion to USD 33.6 billion or growth of 13% to 14%. In M365 Commercial Cloud, we expect revenue growth to be between 13% and 14% in constant currency, with business trends that remain relatively stable quarter-over-quarter. ARPU growth will again be driven by E5 and M365 Copilot. M365 Commercial products revenue growth should be in the low to mid-single digits. As a reminder, M365 Commercial products includes components that can be variable due to the in-period revenue recognition dynamics.

M365 Consumer cloud revenue growth should be in the mid-20s driven by growth in ARPU. For LinkedIn, we expect revenue growth of approximately 10%. And in Dynamics 365, we expect revenue growth to be in the mid- to high teens with continued growth across all workloads.

For Intelligent Cloud, we expect revenue of USD 32.25 billion to USD 32.55 billion or growth of 26% to 27%. In Azure, we expect Q2 revenue growth of approximately 37% in constant currency as demand remains significantly ahead of the capacity we have available. And while we're accelerating the amount of capacity we're bringing online, we will continue to balance Azure revenue growth with the growing needs across our first-party apps and AI solutions, our own R&D efforts and the end-of-life server replacements. Therefore, we now expect to be capacity constrained through at least the end of our fiscal year.

As a reminder, there can be quarterly variability in the year-on-year growth rates depending on the timing of capacity delivery and when it comes online as well as from in-period revenue recognition depending on the mix of contracts.

In our on-premises server business, we expect revenue to decline in the low to mid-single digits with ongoing customer shift to cloud offerings.

In More Personal Computing. We expect revenue to be in the USD 13.95 billion to USD 14.45 billion. Windows OEM and Devices revenue should decline in the mid-single digits. We expect continued momentum from Windows 10 end of support, although growth rates will be impacted by elevated inventory levels at the end of Q1 that we expect to come down through the quarter. Therefore, Windows OEM revenue should decline low to mid-single digits. The range of potential outcomes remains wider than normal. Devices revenue should decline year-over-year. Search and news advertising ex TAC revenue growth should be in the low double digits, down sequentially as growth rates normalize following the benefit from third-party partnerships noted earlier. Growth will continue to be driven by volume and revenue per search across Edge and Bing.

And in Xbox content and services, we expect revenue to decline in the low to mid-single digits against the prior year comparable that benefited from strong first-party performance, partially offset by growth in subscriptions. And hardware revenue should decline year-over-year.

And in closing, demand signals across bookings, RPO and product usage are accelerating faster than we expected. We're investing in infrastructure, AI talent and product innovation to capture that momentum and expand our leadership position. And we remain focused on delivering real value to our customers that results in durable revenue growth for the long term.

With that, let's go to Q&A, Jonathan.

Thanks, Amy. We'll now move over to Q&A. [Operator Instructions]. Operator, can you please repeat your instructions?

[Operator Instructions] Our first question comes from the line of Keith Weiss with Morgan Stanley.

Congratulations on another outstanding quarter. And if I'm looking at Microsoft, this is 2 quarters in a row, we're really seeing results that are well ahead of anybody's expectations when we were thinking about this company a year ago or 5 years ago, 111% in Commercial bookings growth was not on anybody's bingo card, if you will, yet the stock is underperforming in the broader market. And the question I have is kind of getting at the zeitgeist that I think is weighing on the stock. And is something about to change?

And I think AGI is kind of a nomenclature or a shorthand for that. And it's something that still included in your guys' OpenAI agreement. So Satya, when we think about AGI or we think about how application and computing architectures are changing, is there anything that you see on the horizon, whether it's AGI or something else, that could potentially change what appears to be a really strong positioning for Microsoft in the marketplace today where that strength will perhaps weaken on a go-forward basis. Is there anything that you're worrying about in that evolution and particularly the evolution of these generative AI models.

Thank you, Keith, for the question. So here's what I would say, I think there are 2 parts. We feel very, very good about even this, I'd say, the new agreement that we now have with OpenAI because I think even, it just creates more certainty to all of the IP relationship we have as it relates to even this definition of AGI. But beyond that, I think your question touches on something that's pretty important, which is how are these AI systems going to truly be deployed in the real world and make a real difference and make a return for both the customers who are deploying them and then obviously, the providers of these systems. And I think the best way to characterize the situation is that even as the intelligence capability increases, let's even say, exponentially like model version over model version, the problem is it's always going to still be jagged, right? I think the term people use is the jagged intelligence, even -- or spiky intelligence, right?

So you may even have a capability that's fantastic at a particular task, but it may not uniformly grow. So what is required is in fact, these systems, whether it is GitHub Agent HQ or the M365 Copilot system. Don't think of this as a product. Think of it as a system that in some sense smooths out those jagged edges, and really helps the capability.

I mean just to give you a flavor for it, right? So if I am in M365 Copilot, I can generate an Excel spreadsheet. The good news is now an Excel spreadsheet does understand Office JS, has the formulas in it. It feels like, wow, it is a great spreadsheet created by a good model. The more interesting thing is I can go into agent mode in Excel and iterate on that model. And yet, it will stay on rail. It won't go off rail, it will be able to do the iteration. Then I can even give it to the analyst agent, and then it will even make sense of it like a data analyst would of our Excel model.

The reason I say all of that is because that's the type of construction that will be needed even when the model is magical, all powerful. I think we will be in this jagged intelligence phase for a long time. So one of the fundamental things that these -- whether it's GitHub, whether it's security, whether it's M365, the 3 main domains we're in, we feel very, very good about building these as organizing layers for agents to help customers.

And by the way, that's the same thing that we want to put into Foundry for our third-party customers. So that's kind of how people will build these multi-agent systems. So I feel actually pretty good about both the progress in AI. I don't think AGI as defined at least by us in our contract is ever going to be achieved anytime soon. But I do believe we can drive a lot of value for customers with advances in AI models by building these systems. So it's kind of the real question that needs to be well understood. And I feel very, very confident about our ability to make progress.

The next question comes from the line of Brent Thill with Jefferies.

Amy, on the bookings blowout. I guess many are somewhat concerned about concentration risk. And I think you noted a number of $100 million contracts, not to go into a lot of detail, but can you just give us a sense of what you're seeing in that 51% RPO and 110-plus percent bookings growth that gives you confidence about what you're seeing in terms of the breadth and extent of some of these deals on a global basis.

Thanks, Brent. A couple of things to maybe take a step back on RPO. With a nearly $400 billion balance, we've been trying to help people understand sort of how to think about really the breadth of that. It covers numerous products. It covers customers of all sizes. It -- that's been a balance that we've been growing obviously at a good clip. But what people need to realize is it sits across multiple products because of the things Satya is talking about around creating systems and where we're investing.

And if you're going to have that type of balance and then more importantly, have the weighted average duration be 2 years, it means that most of that is being consumed in relatively short order. People are not consuming, and I say this broadly, unless there's value. And I think this is why we keep coming back to, are we creating real-world value in our AI platforms, in our AI solutions and apps and systems.

And so I think the sort of the way to think about RPO is it's been building across a number of customers. We're thrilled to have OpenAI be a piece of that. We're learning a ton and building leading systems because of it that are being used at scale that benefits every other customer. And so it's why we've tried to give a little bit more color to that RPO balance because I do understand that there have been a lot of concerns or questions about is it long dated, is it coming over a long period of time. And hopefully, this is helpful for people to realize that these are contracts being signed by customers who intend to use it in relatively short order. And at that type of scale, I think that's a pretty remarkable execution.

The next question comes from the line of Mark Moerdler with Bernstein Research.

Congratulations on the quarter. It's pretty amazing what you guys are doing. Satya and Amy, I'd like to ask you the #1 question I receive, whether from investors or at AI conferences I attend, how much confidence do you have that the software, even the consumer [indiscernible] business can monetize all the investments we're seeing globally? Or frankly, are we in a bubble? In fact, Amy, what would be the factors you'd be watching for to assure that you're not overbuilding for current demand and the demand will sustain.

Maybe I'll start, Satya and then you could add. Let me talk a little bit about maybe connecting a couple of the dots because with $400 billion of RPO, that's sort of short-dated as we talked about, our needs to continue to build out the infrastructure is very high. And that's for booked business today. That is not any new booked business we started trying to accomplish on October 1, right?

And so the way to think about that, and you saw it this quarter in particular, and as we talked about '26, the remainder, number one, we're pivoting toward -- increasingly, we talked about this short-lived assets, both GPUs and CPUs, Again, we talk about all these workloads are burning both in terms of app building. Now when that happens, short-lived assets generally are done to match sort of the duration of the contracts or the duration of your expectation of those contracts. And so I sometimes think when people think about risk, they're not realizing that most of the lifetimes of these and the lifetime of the contracts are very similar.

And so when you think about having revenue and the bookings and coming on the balance sheet, the depreciation of short-lived assets, they're actually quite matched, Mark. And as you know, we've spent the past few years not actually being short GPUs and CPUs per se, we were short the space or the power is the language we used to put them in. So we spent a lot of time building out that infrastructure. Now we're continuing to do that also using leases. Those are very long-lived assets, as we've talked about 15 to 20 years. And over that period of time, do I have confidence that we'll need to use all of that, it is very high.

And so when I think about sort of balancing those things, seeing the pivot to GPU, CPU short-lived, seeing the pivot in terms of how those are being utilized, we are -- and I said this now, we've been short now for many quarters. I thought we were going to catch up, we are not. Demand is increasing. It is not increasing in just one place. It is increasing across many places. We're seeing usage increases in products. We are seeing new products launch that are getting increasing usage, and increasing usage very quickly. When people see real value, they actually commit real usage.

And I sometimes think this is where this cycle needs to be thought through completely is that when you see these kind of demand signals and we know we're behind, we do need to spend. But we're spending with a different amount of confidence in usage patterns and in bookings, and I feel very good about that. I have said we are now likely to be short capacity to serve the most important things we need to do, which is Azure, our first-party applications. We need to invest in product R&D and we're doing end-of-life replacements in the fleet. So we're going to spend to make sure that happens. It's about modernization. It's about high quality. It's about service delivery, and it's about meeting demand.

And so I feel good about doing that, and I feel good that we've been able to do it so efficiently and with a growing book of business behind it.

Yes. The only thing I would add to what Amy captured was, if you sort of look out, there are 2 things that matter, I think, and that are critical in terms of how we think about our allocation of capital, also our R&D. One is how efficient is our planet-scale token factory, right? I mean that's at the end of the day, what you have to do. And in order to do that, you have to start with building out a very fungible fleet. It's not like we're building one data center in one region in the world that's mega scale. We are building it out across the globe for inference, for pre-training, for post-training, for RL, for data [indiscernible] or what have you. So therefore, that's the fungibility is super important.

The second thing that we're also doing is continually modernizing the fleet. It's not like we buy one version of, say, NVIDIA and load up for all the gigawatts we have. Each year, you buy, you write the Moore's Law, you continuously modernize and depreciate it. And that means you also use software to grow efficiency. I talked about, I think, 30% improvement on both serving up GPT-4.1 and 5.0, right? That's software. That's sort of -- and by the way, it's helpful on A100s, it's helpful on GB200s, and it will be helpful on GB300s. That's the beauty of having the efficiency of the fleet.

So keep improving utilization, keep improving the efficiency. So that's what you do in the token factory. The other aspect, which Amy spoke to is we have some of the best agent systems that matter in the high-value domains, right? It's in information work. That's the Copilot system. Coding, I mean, I should also say one of the things I like about Copilot is, I mean, Copilot ARPU is compared to M365 ARPUs, right? It's expansive. The same thing that happened between server and cloud like we used to always say, well, is it zero-sum, it turned out that the cloud was so much more expansive to the server market.

The same thing is happening in AI because first, you could say, hey, our ARPUs are too low when it comes to M365 or you could say we have the opportunity with AI to be much more expansive. Same thing with tools, right? I mean, tooling -- the tools business was not like a leading business, whereas coding business is going to be one of the most expansive AI systems. And so we feel very good about being in that category. Same thing with security, same thing with health. So we have -- and in consumer, one of the things is it's not just about ads, it's ads plus subscriptions that also opens up opportunity for us.

So when I look at the entirety of these high-value agent systems and when we look at the efficiency of and fungibility of our fleet, that's what gives us the confidence to invest, both the capital and the R&D talent to go after this opportunity.

The next question comes from the line of Karl Keirstead with UBS.

Okay. This one is for Amy. Amy, I certainly don't want to take you down too complex an accounting path with this question, but the investment in OpenAI that sits in other income at $4.1 billion is so large that I think the audience could -- listening in could benefit from a little bit more color about what that is. It feels like it's so much larger than you were running through other income in prior quarters that it mustn't just be your share of the OpenAI losses. So could you just describe that? And what we can expect in subsequent quarters? And whether this signals any kind of accounting change?

The Q1 number was not impacted at all by the new agreement that was put in place. Let me first say that. Secondly, that increased loss was all due to our percentage of losses in OpenAI due to the equity method. So just to be very clear. So there is not anything there that is not the increased losses from OpenAI.

The next question comes from the line of Mark Murphy with JPMorgan.

So we seem to be entering into a new era where the contractual commitments from a small number of AI natives are just incredibly large, not only in absolute terms, but sometimes relative to the size of the companies themselves. For instance, contracts worth hundreds of billions of dollars that are 20x their current revenue scale. Philosophically, how do you evaluate the ability of those companies to follow through on these commitments? And how do you think about placing guardrails on customer concentration for any single entity?

Yes. Maybe I'll start and then Amy, you can add. I mean it goes back a little bit, Mark, to what I said about building first, the asset itself such that it's most fungible. And then to recognize the strength of even sort of our portfolio, we have a third-party business, we have a first-party business, we have third-party also spread between enterprise, digital natives, I always felt that we need a balance there because it may start with digital natives. They're always going to be the early adopters. You always have the hit app of the generation. And then -- essentially then it spreads throughout. The enterprise adoption cycle is just starting and so therefore, having the -- over the arc of time, I think that third-party balance of customers will only increase.

But it's great to have the hit first-party apps in the beginning because you can build scale that then if it's a fungible and that's where the key is. You don't want to build for a digital native in -- as if you're just doing hosting for them. You want to build. That's where -- I think some of the decision-making of ours is probably getting better understood. What do we say yes to, what do we say no to. I think there was a lot of confusion, hopefully by now, anyone who switched on would figure this out. And so that's, I think, one thing we're doing on the third party.

But the 1 -- first party is probably where a lot of our leverage comes and it's not even about one hit app on our first-party even. Our portfolio of stuff which I just walked through in the earlier answer, gives us, again, the confidence that between that mix, we will be able to use our fleet to the maximum. And remember, these assets, especially the data centers and so on are long assets, right? There will be many refresh cycles for any one of these when it comes to the gear.

So I feel that once you think about all those dimensions, the concentration risk gets mitigated by being thoughtful about how you really ensure the build is for the broad customer base.

And maybe just to help with another angle that I think, Satya helped a lot is that when you think about concentration risk or delivering to any customer, you have to remember that because we're talking about this very large flexible fleet that can be used for anyone and for any purpose, 1P, 3P, and including our commercial cloud, by the way, which I should be quite clear on, it is pretty flexible in every regard, you have to remember that the CPU and GPU and the storage gear, doesn't come into play until the contracts start happening. And so you're right, some of these large contracts have delivery dates over time.

So you get a lot of lead time in being able to say, "Oh, what's the status?" And so I think we're pretty thoughtful around what's always gone in our RPO balance, and then considerate of that. There's always been that taken into account when we publish that bookings on brand, publish the RPO balance.

The next question comes from the line of Brad Zelnick with Deutsche Bank.

And I'll echo my congrats on an amazing start to the year. Amy, is there any way to quantify or frame the revenue impact of Azure being short on capacity? And while I appreciate the constraints you face are broad across the industry, is there risk of workloads going elsewhere? And how do you mitigate that?

Yes, Brad, it's a great question. It's always hard to quantify precisely what would have been the revenue impact in quarter. But I would offer a way to think about it is Azure probably does bear most of the revenue impact. Because when you think about real priorities that you have to fill first, it's obviously the increasing usage and adoption and sales we've seen of M365 Copilot and the usage of Copilot chat, which we've seen very different patterns, which we're encouraged by. It's the adoption of security features. It's the GitHub momentum.

And so when you're thinking about it, that is where and it is a priority for us to allocate resourcing there first. And so you are right to ask how do I think about that. We've worked very hard to try to mitigate it as best we can, but we have been short in Azure, and we've been clear on it. And I would say the other 2 priorities that I haven't mentioned maybe as much before is also just making sure our product teams and the AI talent that we've been able to hire into the company really over the past 1.5 years have access also to significant capacity because we're seeing it make the product better in a loop that is adding great benefit today into products people are using today for real-world work.

And so we are making that a priority to make sure our research teams have that as well as our product engineering teams. And yes, it does impact Azure directly. That is the place where you see that prioritization. But I think it's probably hard for me to give an exact number, but it is safe to say that the number could be higher.

The last question will come from the line of Kash Rangan with Goldman Sachs.

Amy, I just wanted to congratulate you, I think you said before that it is possible to accelerate Azure growth while getting efficient with margins and you've done it. Congrats on that.

I have one for you, Satya. With respect to the elephant in the room, following just being a little more direct, following up on Keith Weiss' question. There's talk that another hyperscaler came in and took away the business that was rightfully Microsoft's. I'm sure that there is a different point of view here.

I'm wondering if you could offer some perspective on your criteria to -- is it about a certain volume of business that you wish to execute on the Microsoft paper? Or is it something broader than that, that I don't think maybe people fully appreciate the terminal value that Microsoft will have on its balance sheet at the end of these contracts, which I think is probably being underestimated as you have a full stack and you've got the multiple vectors to monetize, be it databases, Foundry. And to your point that you are a platform company, not just a hyperscaler. Maybe that's what it is all about, or maybe there's another story about you letting the other hyperscaler company coming from nowhere and claiming a big piece of that 4- to 5-year puzzle. And congratulations.

Well, thanks, Kash. I mean for us, again, just always goes back to, I think, the core principle, which is build a fleet that is fungible across the planet and works for third-party and first-party and research. So that's essentially what we have done.

And so when some demand comes in shape, that don't fit that goal, where it's too concentrated, not just by customer, by location, by type of skewing, right? I think Amy mentioned some very key things. When you think about the margin profile of a hyperscaler, you've got to remember this, the AI accelerator piece, but there's compute, there's storage. And so if all of the demand just comes for just one [ meter ] that's really not a long-term business we want to be in. That's even from a third party. We have to balance it with all of our first-party stuff because that's after all a different margin stack for us. And then we have to fund our own R&D and model capability because in the long run, that's what's going to differentiate us.

And so I look at all of those. We sort of use all of that to make sure we are saying yes to all the demand that we want, we say no to some of the demand that may be something that we could serve, but it's not in our long-term interest. And so that's sort of the decision-making we have done, and we feel very, very good about the decisions. In some sense, I feel even each time we say no to, the day after, I feel better.

And just Kash, I think this is our last call with you. And I just want to say thanks and congratulations. It's been a privilege to work with you and best of luck.

Let me add to that, Best of luck, Kash.

Thanks, Kash. That wraps up the Q&A portion of today's earnings call. Thank you for joining us today, and we look forward to speaking with all of you soon.

Thank you all.

Thank you. This concludes today's conference. You may disconnect your lines at this time, and thank you for your participation.