Innodata Inc. Stock

Hello, everyone. Thank you for your patience, and thank you for joining us. Welcome to the Innodata Q3 2026 Earnings Call. [Operator Instructions]

I will now hand the conference over to Amy Agress, General Counsel. Amy, please go ahead.

Thank you, operator. Good afternoon, everyone. Thank you for joining us today. Our speakers today are Rahul Singhal, President and Chief Executive Officer of Innodata; and Jayant Chauhan, Chief Financial Officer. Also on the call today is Aneesh Pendharkar, Senior Vice President, Finance and Corporate Development. We'll hear from Rahul first, who will provide perspective about the business, and then Jayant will provide a review of our results for the third quarter. We'll then take questions from analysts.

Before we get started, I'd like to remind everyone that during this call, we will be making forward-looking statements, which are predictions, projections and other statements about future events. These statements reflect management's views as of today and are based on current expectations, assumptions and estimates as of today, and as such, the statements are subject to risks and uncertainties. Actual results could differ materially from those contemplated by these forward-looking statements. These statements include, but are not limited to, statements regarding our future financial and operational performance, business strategies, growth initiatives, expectations, projections and guidance.

Factors that could cause these results to differ materially are set forth in more detail in today's earnings press release and in our most recent annual report on Form 10-K, including the sections entitled Cautionary Note Regarding Forward-Looking Statements and Risk Factors, as well as our subsequent quarterly reports on Form 10-Q and other reports and filings with the Securities and Exchange Commission. These documents are available through the commission's website and our website. We undertake no obligation to publicly update or revise any forward-looking information.

In addition, during this call, we may discuss certain non-GAAP financial measures. In our earnings release filed with the SEC today, as well as in our other SEC filings, which are posted on our website, you will find additional disclosures regarding these non-GAAP financial measures, including reconciliations of these measures with comparable GAAP measures.

Thank you. I will now turn the call over to Rahul.

Thank you, Amy, and good afternoon, everyone. Before I turn to the quarter, I want to take a moment to acknowledge the occasion. This is my first earnings call as Innodata's CEO. I'm honored by the confidence that Jack and the Board have placed in me, and I'm mindful of what I have inherited, a company with extraordinary momentum, an extraordinary team and an extraordinary opportunity in front of it. I could not be more excited to discuss our results and strategy.

I also want to acknowledge Jack. As Executive Chairman, he remains deeply engaged, partnering with me and the research team to build differentiated capabilities and to bring them to the federal government and the enterprise. Our strategy is shared, and the research-led engine that has been driving our results is the one I intend to build on.

I said last quarter that I would repay the confidence placed in me with results, and I'm pleased to be starting with a quarter that lets me do exactly that. Q3 was another record revenue quarter for Innodata. Revenue was $95.6 million, up 53% year-over-year, exceeding analyst consensus of $83.3 million by approximately $12.3 million or 15%. This was our 13th consecutive quarter of year-over-year growth.

We exceeded analyst consensus on other key metrics as well. Adjusted gross margin was 40%, in line with our publicly stated 40% target. Adjusted EBITDA was $17.7 million or 19% of revenue, exceeding consensus by approximately $2.3 million. Net income was $10.3 million, up 24% year-over-year and fully diluted earnings per share were $0.29 compared with consensus of $0.17.

Much as we hoped, Q3 and Q4 to date have been strong in new business wins. The customer we referred to last quarter as our second largest customer became our largest customer in Q3 and continues to expand with us. In the quarter, we signed a new program with this customer for pretraining data with a contract value of $29 million. So far in Q4, we have signed an additional program with this customer valued at approximately $18 million, and we have a robust pipeline with them going into 2027.

We also have won 5 new programs at another customer, a large hyperscaler. Just to give you a sense for the momentum and success we have already achieved with this customer, in the trailing 12 months, we did approximately $3.1 million of revenue with this customer. But based on these new programs, we are now forecasting this customer may contribute more than $20 million in revenue in 2027. And beyond that, there's significant additional pipeline in discussion.

We also added 5 new logos in Q3, including model builders and an enterprise data platform. And in the last few weeks, we signed an enterprise-focused foundation model company. Beyond that, we are in promising conversations with the leading frontier model developer for a major expansion, and we just opened discussions with a leading social media company. The business we are winning reflects how well we continue to align with the needs of frontier model builders even as those needs grow more sophisticated across virtually every dimension.

In just the last 18 months, model training emphasis has shifted from human-based evaluations to agentic traces within sandbox replicas of real-world environment, commonly called RL gyms, and from synthetic data to a renewed appetite for real-world data. Each of these shifts has expanded both what is possible and what is required, and each has created new opportunities for us.

Our significant new wins are also enabling us to meaningfully drive customer diversification. One dynamic worth noting, several times, researchers we have worked with at the leading labs have moved to other labs or to well-funded start-ups. And when they land, they call us. They knew our capabilities and our people, so they brought the relationship with them. This has helped us, and we believe it will continue to help us grow our pipeline at a rate that outpaces SG&A growth. Today, we are raising our guidance from 40% to 42% for 2026.

Given the magnitude of the business we are winning, our raise could have been more substantial but for the fact that in mid-October, our traditionally largest customer decided to end one of its programs with us. The good news is that we are expecting this revenue to be replaced by new programs at this customer by Q1, but we expect it to cost us about $12 million in Q4 revenue.

I'd like to emphasize again, as we have in the past, that within our large customers at any point in time, there are programs ending, other programs starting and still other programs in discussion. This reflects the dynamic nature of our business and represents our business running as designed.

I'd now like to take a few minutes to drill down on 6 specific growth drivers that we believe will propel our growth over the next 12 months. I'll start with driver #1, physical AI. Capital is pouring into physical AI faster than into any category since the LLM boom began. Physical AI is sometimes referred to as embodied intelligence or simply AI robotics. In the first half of 2026, global venture funding in the space totaled $47.4 billion across 521 deals. This exceeds the amount invested in physical AI from 2022 through 2024 combined. The bets are concentrated in robot foundation models and companies looking to build consumer and industrial robotics.

Recent research confirms that robotics foundation models improve with the data the same way LLMs do. But because no Internet scale corpus of physical interaction exists, the data must be manufactured through teleoperation, egocentric human video and simulation, all work we've begun doing. And as models move from lab demos to deployment in factories, warehouses and homes, we anticipate demand expanding from one-off pretraining corpora to continuous task-specific data for fine-tuning and evaluation.

As a data provider, we become recurring infrastructure. We've been working hard to capture the opportunity and have made significant strides across robotic data collection, including egocentric data, UMI gripper data and motion capture data, the creation of large-scale AI-assisted action annotations for VLM training and robot safety testing and evaluation.

Our physical AI data business is scaling materially. We operate large egocentric data collection programs, stereo camera headset capture in home environments across the United States and Latin America and 6-camera headset capture of complex industrial tasks across hundreds of businesses. Our customers are leading humanoid and frontier robotic labs with big ambitions. We are seeing just one of our recently acquired humanoid customers representing a potentially multimillion-dollar opportunity for us next year.

In September, we launched our state-of-the-art Vicon motion capture lab to capture high-fidelity 3D human motion for world model training and humanoid robot training, and we have demonstrated an end-to-end real-to-SIM-to-real pipeline, bringing captured human motion into Isaac Sim, retargeting it to humanoid kinematics and executing our own Unitree G1. Beyond collection, we are quickly becoming a trusted annotation partner to physical AI labs, delivering more than 620,000 robot executable labeling jobs this year so far. This is a new focus area for us, which I expect to become a big growth driver for 2027 as robots move from labs to production.

We are launching an annotation pilot with a leading robotics lab, which if scaled, could be a sizable opportunity. We also won a pilot with a leading frontier lab. If the pilot is successful, we anticipate it could yield up to $7 million of run rate revenue in 2027. We're also working with a major technology company on safety red teaming data for robot foundation models.

I'll next talk about driver #2, RL environment for personal and knowledge work agents. I mentioned earlier the shift towards training in RL gyms. The reason for this shift is simple. An agent learns by acting and seeing the consequences, and labs can't let agents learn on live systems for safety and privacy reasons and because many actions can't be undone. So the agent trains inside a sandbox replica where it can explore, make mistakes and reset thousands of times.

Innodata is now producing at scale the kind of high-quality RL environments that we believe will dominate model training for the foreseeable future. We anticipate this movement, and we have built strong research team with deep expertise in environment design and reinforcement learning. We are particularly focused on 2 use cases, personal agents and knowledge work agents, and we intend to lead in these areas.

On the personal agent side, we've been working for some time with one of our largest customers on personalization for its consumer agents. If you follow the consumer agent launches this year, you saw what personalization at that level looks like in the hands of hundreds of millions of users. We are a part of the work behind that. Our data, our environment and our methodologies are embedded in how agents of that kind learn what a person actually wants. We believe personal agents are the next competitive battleground, and the winners will be the companies whose agents truly know you and can tailor the experiences to you.

The sophisticated RL gyms that capture that kind of depth are a key differentiator in that race. And our customers have told us the methodologies and tooling we have developed for extracting the data behind the personal agents are unlike anything they have seen. We expect to be building more of these personalization agent gyms over the next several quarters.

On the enterprise side, the battleground among model builders has shifted from coding agents to agents that perform everyday knowledge work across desktop and browser-based workflows that span multiple systems, start from unstructured inputs and require judgment at every step. Models are now being asked to do the expert work that rules-based automation never could. We are building sophisticated RL gyms to train these models, and we are building what goes inside them as well. We are in the early innings of this work, and we believe our team and the technology position us well, and we intend to lead in training these new generation of agents.

I'll next talk about driver #3, trust and safety. And I'll start with alignment, by which we mean the work of making AI systems behave safely and in accordance with human intent and policy. We believe the need for this work is increasing for a couple of very straightforward reasons. First, as AI moves from one-shot answers to multistep reasoning and from chatbot assistance to autonomous agents, the cost of alignment failure increases and evaluating safety becomes more demanding.

Secondly, models are being deployed in sensitive domains such as health and finance, where failure carry real consequences. As we have all seen in recent months, as models become more capable, the attendant risk become greater. The labs are responding by shifting more efforts towards alignment post training and specifically humans in the loop data. This presents a significant opportunity for Innodata.

Our alignment work today spans several high-stakes areas. For example, we are finalizing a new statement of work valued at approximately $6 million of potential revenue to perform chemical, biological, radiological, nuclear and explosive, or CBRNE, risk evaluation agent safety work. This is with a large hyperscaler that we announced on our first quarter call has selected us as a global trust and safety partner.

In addition, the customer has verbally committed to a program with us under which we will test agentic safety at some of their platforms. We believe this program has the potential to yield approximately $2.5 million of revenue next year. We are also in pilots on several mission-critical safety programs in health, cybersecurity and robotics. We believe these pilots will be successful and that the programs will scale significantly. Several of these pilots use what we call temptation gyms, environments built to reveal whether the model will pursue a legitimate goal by illegitimate means or cut corners to get there. I'll say more about these in a moment.

Across trust and safety, as across our agentic and RL work, we are gaining ground because of how our customers see us as a trusted adviser, a serious applied research organization and a reliable delivery partner. That combination makes us sticky within these accounts, and it lets us help shape demand, not just respond to it.

I'll now turn to driver #4, end-to-end training pipelines that span pretraining to agentic AI. Here is the context. The open web is running dry. Epoch AI estimates that labs will exhaust the supply of public human written text sometime between 2026 and 2032. Every stage of model development now depends on data that must be deliberately created, such as specialized pretraining corpora and RL environments and evaluations. I believe Innodata is one of the few companies in the world doing this work end-to-end and bringing deep domain expertise to all of it.

On the pretraining side, we are going beyond standard web-based sourcing, instead securing unique differentiated data sets. We see demand for this in several areas. Our largest opportunity is a potential program we are discussing with a leading frontier lab. It is intended to span a wide variety of unique data types at a run rate in the millions of dollars per month. Video is a good example of data that can't simply be scraped at the quality and with the rights that labs need. We are in conversation with one customer on a high-end cinematic video opportunity valued at approximately $2 million.

Charts are another area we are focused on. Our teams are building complex charts and graph data sets to strengthen visual reasoning in models, and we have secured several new customers for these data sets.

We are also expanding our real-world business data licensing practice. Frontier labs want raw real-world data as well as agentic trajectories with verifiable outcomes required to train their models. Because we work so closely with our customers, engineering teams, we have an advanced understanding of which data sets will deliver the performance lift they're looking for. And that same knowledge makes us a natural partner for the owners of that data who rely on us to connect it to the problems it can solve.

Human evaluation continues to progress from generalist crowds to expert pools. We recently won a new contract value at approximately $1.2 million with a leading global provider of legal and regulatory solutions. Under this agreement, our legal experts will provide human-in-the-loop validation and the expert feedback used to fine-tune the customers' specialized legal AI models. The win highlights our strength in high-stakes workflows and establishes a scalable blueprint for high-margin multiyear growth in enterprise AI.

We're applying the same expert model to agentic AI using knowledge workers to build agentic gyms, environments where agents perform real computer use and knowledge work tasks against realistic data. As agents take on more autonomous work, labs need to know not only whether an agent can complete a task, but whether it completes it honestly. Our temptation gyms are exactly built for that. These environments are deliberately designed with opportunities to cut corners, gaining a reward, fabricating a result, misrepresenting what was actually done, and they are instrumental so labs can see whether an agent takes it.

That lets lab catch behavior that in real deployment would amount to a misrepresentation of fraud. This kind of integrity testing is becoming essential and requires both domain expertise and environment building capability. Temptation gyms are a new category, and we believe being early gives us capability that will differentiate us.

Next, I'll address driver #5, enterprise AI. As we have said previously, implementation of generative AI models in the enterprise is a very large opportunity for us. We have been building a practice for the last 3 quarters, and we continue to gain momentum. Just last quarter, we added approximately $5 million of revenue run rate of agent implementation for a large customer. We also initiated 2 new enterprise relationships with a global technology hardware company and a major financial services firm. And we expanded our work with the social media customer we have served since last year to about $2.5 million a year.

We believe these accomplishments give us a strong foundation to accelerate from in 2027. We have built a 3-tier partnership strategy that enables us to help our enterprise customers at the foundation model layer, the data layer and the infrastructure layer. Our partnership with Anthropic, a leading foundation model provider, helps us drive enterprise agentic solutions into production. Our partnerships with Databricks and Snowflake, leading data platforms help us build the AI-ready data foundations that every successful deployment depends on. And our partnership with Google Cloud enables us to deliver and scale solutions on enterprise-grade cloud infrastructure. Moreover, these channel partnerships gives us additional growth paths and joint opportunities.

We believe our experience in training frontier models puts us in a unique position to understand how to create golden data sets, benchmark models, build custom rubrics and deploy end-to-end data engineering at scale to create robust production-ready solutions for enterprises and gives us a unique advantage compared to traditional systems integrators.

Lastly, I'll talk about our sixth driver, our federal practice. We launched our federal practice about a year ago. Since then, it has built a position in government AI programs that compound each award. In its first year that produced 2 programs and a third committed through our prime partners. The opportunity we described as a finalist in May, an AI development environment for the Department of War, is in Phase 1 delivery, with a production decision expected in the coming months. We were awarded advanced data labeling on our U.S. Navy autonomy program, the first phase of a competitive down-select, selected on the computer vision capability we demonstrated to the government this year.

And we have been identified for a subcontract role on the second multiyear department award program. Each of these is staged, with a larger award decided after the current phase performs, which is where we have chosen to compete. We have a verbal win for one contract of $5 million per year for 5 years with signing expected by end of the fourth quarter. We hope to continue momentum across all programs.

We are ending the year with real momentum, a clear strategy and tailwinds from each of the 6 growth drivers I've just walked you through. I've never been more confident about where Innodata is positioned. The opportunities in front of us are both large and they are close at hand. We are investing to capture them, and we expect those investments to pay back quickly and to build durable advantage. The next 12 to 18 months will be about execution, and that is where Jayant and I are focused.

Let me turn over the call to Jayant, our CFO, to share his perspectives about the business and our Q3 financial results.

Thank you, Rahul, and good afternoon, everyone. Let me walk through our third quarter results and also share our thinking on investments and capital allocation. Revenue for Q3 2026 was $95.6 million, up 53% year-over-year and 4% sequentially, our 13th consecutive quarter of year-over-year growth. This exceeded analyst consensus by $12.3 million or 15%.

Adjusted gross profit was $38.3 million, representing adjusted gross margin of 40%. This is in line with our 40% target. In any quarter, our gross margin is driven by program mix. Adjusted EBITDA was $17.7 million or 19% of revenue, up 9% year-over-year. This was ahead of consensus of $15.4 million by approximately 15%. Net income was $10.3 million, up 24% from the $8.3 million we reported in Q3 last year. Fully diluted earnings per share was $0.29 compared with consensus of $0.17, a beat of approximately 70%. Our year-to-date effective tax rate was approximately 12% compared to our long-term target range of 23% to 25%. The lower tax rate was driven by tax benefits recognized this quarter.

Turning to the balance sheet. We ended the quarter with approximately $164 million in cash and cash equivalents. Excluding approximately $10.3 million of customer prepayments, which are a pass-through, our cash was $153.7 million. This is up $19.3 million from $134.4 million on the same basis at the end of Q2. We have no debt. Our $50 million Wells Fargo credit facility remains undrawn. We did not use our ATM program during the quarter.

With the full quarter behind me, let me share some perspectives on investments and capital allocation. Given the opportunity set Rahul described, we will continue to invest in areas we believe will become meaningful growth drivers in 2027 and beyond. We intend to fund these investments from operating cash flow, and we expect them to pay back quickly. On capital allocation, with approximately $154 million of cash, net of customer prepayments and no debt, we have real flexibility. Our first priority is organic investment in the growth drivers. Beyond that, we will be selective about M&A that we believe can accelerate our strategy, and we intend to make these investments while maintaining a strong balance sheet.

To sum up, Q3 delivered record revenue, adjusted EBITDA ahead of expectations at our 40% gross margin target and $19.3 million of underlying cash generation, while we continue to invest in the drivers we believe will enable our next phase of growth.

With that, I'll turn it back to the operator. Operator, we are ready for questions.

[Operator Instructions] Your first question comes from the line of George Sutton with Craig-Hallum.

Can you explain the process of your labor pool and how that works? So you have a program that ends and this happens all the time. I think typically, you would like to deploy those folks into other jobs immediately. You mentioned Q1 could be a similar kind of a deployment. Do these require different skill sets? Do you then rehire from that pool? Can you just give us a picture there?

George, thank you for your question. So yes, I think the way the program start and stop is dependent on the types of tasks that these frontier model builders are working on. Sometimes the skills that they're working on could be redeployed. And sometimes those skills cannot be redeployed.

The other thing to note is when these different frontier model builders are in different cycles of their training, and they are in different training -- training the different tasks. So I think it becomes a matter of what's available, what's usable, reusable at that time from a labor pool perspective. And I think when we talk about Q1, I think it's different programs that we are expecting us to start.

Understand. And you made an interesting comment when you were talking about the physical AI, I think specifically the robotics work that you look to become recurring infrastructure, which is quite different than the normal programs that you've traditionally worked on. Can you just explain that a bit further?

Yes. Absolutely. I think we are very excited about physical AI. So if you think about the investment that is going in physical AI, it is, I believe, going to outpace the investment that happened in LLM. Now one of the most interesting things about physical AI or embodied intelligence is the fact that you can't scrape data from the web as the LLM started to. So the pretraining data or the pretraining corpora doesn't exist. The way these models are getting trained is it starts by doing something called an egocentric data, where you are using cameras to collect data for different tasks.

So a task could be you're collecting data for loading a dishwasher or you could be collecting the data for folding your laundry, right? Those are egocentric data. Then you have something called as UMI data sets, and then you have something called as motion capture data sets. Now these are different modalities, different ways, different technologies and techniques that are being used. And each one progressively becomes a lot more interesting, a lot more complex and the models start to become better. This is where the world model start to play in.

Because there is not a single layer of physical corpora that people can use, you have to continuously be collecting, creating, annotating the data so that these robots can actually execute around that. So I think it's a very interesting and exciting opportunity for us.

Super. And then just one for Jayant. So relative to the guide for Q4, historically, what you've done is your guide includes only what you have as booked business. So if you get additional business during the quarter, that creates some upside potential. Can you just talk to the thought process with the Q4 guide? And then also reference the pipeline of opportunities and kind of speed of deal cycles that could result in more revenues in Q4?

George, thank you for the question. In terms of how we think about guidance, that approach is still consistent with the past -- what in the past, what we have done. We continue to be prudent about it. And you're right, we generally factor in only revenue from active programs, new awards or smaller pipeline opportunities where we actually have a clear line of sight. We do not factor large pipeline opportunities until they are awarded and until we know the timing of that revenue recognition. So continue to be consistent with that approach as of now and going forward as well.

And on to your second question on -- of course, on your second question of what we are seeing in the pipeline, I believe Rahul talked about a lot of those opportunities which we are seeing today in terms of guiding for 2027, as you're aware, we'll announce that along -- or rather we intend to announce that along with our Q4 results. So we'll hear more about that over the next quarter.

Your next question comes from the line of Allen Klee with Maxim Group LLC.

Yes. You spoke about how your adjusted gross margin hit your target of 40% plus for the quarter. In the prior 2 quarters, it kind of overearned. It was above. And could you just give us a little more color about what happened with the mix? And is there a reason why you think it might go in a certain direction going forward?

Allen, so I think as we have consistently said that our target gross margin is 40%. And in any quarter, the program mix, the margin mix on those programs change. And we also tend to do a lot of programs where we believe that they could be big growth drivers for us in following quarters or they are building capabilities, which we tend to take at lower margin. So we have consistently guided to 40% gross margin, and we will continue to guide to 40% gross margin because I think this business requires us to ensure that we are investing in different programs, building our capabilities and investing going ahead in a way that we can continue to lead the market.

Okay. And then for the project that was shut down in October, and then you -- I think you said that's going to have a $12 million impact in the fourth quarter, but that the same former largest customer is going to add that amount in the first quarter. Is those -- the margins of the business shut down, is there a way to think -- is that materially different than the margins of the new business?

Let me take that, Rahul. Well, look, on the new build, as you know, we don't really guide or we don't talk about any project level margin. So I would just point to the fact that we'll talk about 2027 and what that looks like during our Q4 results.

Okay. And then the last thing about capital allocation where you were talking about organic growth. Were you trying to imply that the amount of money you might spend on organic growth to fund that will increase like I guess relative to where it had been? Or what were you trying to get across to us?

Yes, I'll take that, Rahul. Allen, yes, as we've been talking to you guys in the past and to our investors as well, there are a lot of questions on how are we thinking about capital allocation given the balance sheet strength. So just basically clarifying that organic investments continue to be our priority, and that's where we'll focus as we invest behind the various growth vectors and opportunities Rahul laid out. So that continues to be the priority.

And if there is opportunistically, we see M&A opportunities, which we can -- which can accelerate our growth strategy and create value for our shareholders, we will continue to look at that as well, but priority is organic investments.

Your next question comes from the line of Hamed Khorsand with BWS Financial.

Just one question is about this large customer moving out one project to another. Is that other project won by you? Or do you have a confirmation that's happening in Q1?

Yes. So Hamed, we don't comment on exact timing. We have -- we anticipate winning more work, and we are in late stages of several programs that we can -- we anticipate will help us fill the gap. So I think the way to think about it is program start and stop. New programs are starting. We are working with multiple stakeholders, multiple business units within our largest customer, and this business moves really, really fast. So we anticipate filling the gap.

And then it sounds like you have a new large customer. What's the percentage of revenue?

Yes. So you're talking about the third large customer, Rahul talked in his comments earlier?

Yes.

Yes. So at this point in time, we are expecting to do about $20 million in revenue from that customer over the next year.

Okay. And then as far as what you're seeing from your largest customer, has that changed at all? Or has that been pretty consistent as far as revenue is concerned?

Yes. I think the -- what we're seeing from our largest customer is this market is moving significantly faster and the investments go in a variety of different directions. The customers tend to invest in things like VLMs, which is vision language models. They're investing in robotics. They're investing in LLMs. They're investing in agents, right? So the business is moving fast and you have to -- and we are working with a variety of different business units across the gamut within the largest customers in trying to address their needs and work with them.

We are spending time with them on reinforcement learning gyms where we are creating a cutting-edge gyms that allow the agents to be trained and giving them our expertise and our thought leadership in helping them trying to dominate that market. So I think different places is where we are -- we continue to work with them in a variety of different programs.

There are no further questions at this time. I will now turn the call back to Rahul Singhal for closing remarks.

Thank you, operator. To wrap up, Q3 2026 was another record revenue quarter for Innodata and another across-the-board beat. We delivered 53% revenue growth, our 13th consecutive quarter of year-over-year growth, adjusted gross margin of 40%, adjusted EBITDA ahead of consensus and strong net income, all while continuing to invest in the research and innovation that is driving the business.

Our diversification continued. A customer that represented less than 5% of our revenue a year ago is now our largest customer, giving us 2 customers of comparable scale. Another large hyperscaler customer with which we did only $3.1 million in revenue in the trailing 12 months is on track to be a $20 million-plus customer next year. We also have won notable new customers across frontier labs, enterprise data platforms and robotics. We are raising our guidance to 42% or more growth for 2026. Given our wins, the raise would have been a lot more, but for an air gap in Q4 revenue from our historically largest customer that we forecast getting filled by Q1.

Most importantly, we believe we are positioned at the epicenter of the most important transition in AI right now, the move to the next S-curve of agents, reinforcement learning and embodied intelligence, and at the center of the trust, safety and alignment work that will determine whether that transition succeeds. These are primarily data challenges and they're precisely in our wheelhouse. I believe 2026 will finish as a strong year for Innodata and 2027 will prove equally or even more exciting.

It is a privilege to be leading this company. I thank our team and customers around the world and all of you for continuing to be on this journey with us. We look forward to speaking with you next quarter. Thank you.

This concludes today's call. Thank you for attending. You may now disconnect.

Thank you for standing by. At this time, I would like to welcome everyone to the Innodata Q2 2026 Earnings Call. [Operator Instructions]

I would now like to turn the call over to Amy Agress. You may begin.

Thank you. Good afternoon, everyone. Thank you for joining us today. Our speakers today are Jack Abuhoff, Chairman and CEO of Innodata; Rahul Singhal, President and Chief Revenue Officer; and Jayant Chauhan, Chief Financial Officer. Also on the call today is Mariz Espineli, Chief Accounting Officer and Aneesh Pendharkar, Senior Vice President, Finance and Corporate Development. We'll hear from Jack and Rahul first, who will provide perspective about the business and then Jayant will provide a review of our results for the second quarter. We'll then take questions from analysts.

Before we get started, I'd like to remind everyone that 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, assumptions and estimates and are subject to risks and uncertainties. Actual results could differ materially from those contemplated by these forward-looking statements. Factors that could cause these results to differ materially are set forth in today's earnings press release 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 undertake no obligation to update forward-looking information.

In addition, during this call we may discuss certain non-GAAP financial measures. In our earnings release filed with the SEC today as well as in our other SEC filings, which are posted on our website, you will find additional disclosures regarding these non-GAAP financial measures, including reconciliations of these measures with comparable GAAP measures. Thank you. I will now turn the call over to Jack.

Thank you, Amy, and good afternoon, everyone. Q2 was another record quarter for Innodata. Revenue, adjusted gross profit, adjusted EBITDA and cash all reached new highs, and we exceeded analyst consensus on all key metrics. Revenue was $92.1 million, up 58% year-over-year, exceeding analyst consensus by approximately $5.8 million, or 7%, and making Q2 our 12th consecutive quarter of year-over-year growth.

To put that in perspective, in Q2, as in Q1, our quarterly revenue exceeded our annual revenue of just 3 years ago. Our adjusted gross margin, meanwhile, was 49%, up 2 points sequentially and 9 points above our 40% publicly stated target. Adjusted EBITDA was $25.4 million, up 92% year-over-year, exceeding analyst consensus by approximately $8.5 million, or 50%. Fully diluted earnings per share were $0.41 per share, nearly double analyst consensus of $0.21 per share. Again, this quarter, we delivered growth, margin expansion and cash generation together, while investing in innovation that converts to revenue within quarters, not years. That is the business model working as designed.

Last quarter, we told you to expect our largest customer to represent a smaller percentage of total revenue. In Q2, our largest customer represented 37% of revenue, down from 56% of revenue in Q1, while the big tech customer we announced last quarter scaled from 17% of revenue to 34% of revenue, becoming our second-largest customer. While our largest customer contributed less revenue in Q2 than in Q1 as a result of a change in the quarter to program structure and service mix, we continue to expect it to grow year-over-year for the full year. We also landed an important new customer in the quarter, one of the fastest-scaling frontier labs. The upshot is our base continues to broaden in both customers and customer programs.

Now before turning to guidance, I want to share an important announcement about Innodata's leadership. Effective September 30, Rahul Singhal will become President and Chief Executive Officer of Innodata and will join our Board, and I will transition into the role of Executive Chairman. This is a planned transition made from a position of strength. And for me, it is also a personal one. Many of you know Rahul from these calls, from investor conferences and from the work he has led over the past several years as a principal architect of Innodata's transformation into a strategic partner to the world's leading AI builders. He knows our customers, he knows our technology and he knows our people. Rahul has been central to every element of the strategy behind the results you have seen quarter after quarter. The Board and I didn't have to look far for the right leader. Rahul earned this role, taking on expanding responsibility year after year and delivering every time. This is how we build this company: we grow our capabilities, and we promote our own people.

As Executive Chairman, I will remain deeply engaged, focused on partnering with Rahul to build capabilities enabled by our research team. Bringing these capabilities to the federal government and to the enterprise, I believe, is where I can best contribute to creating significant shareholder value. And as one of the company's largest shareholders, that is exactly what I want to be doing. Our work with the Mag 7 and the leading AI labs is on a firm path to greater heights and greater diversification. Our enterprise AI and federal strategies, built on the differentiated technology we develop for the frontier labs, represent opportunities for potentially driving high-quality recurring revenue that results in significant value creation.

We are building Innodata to be a generational company. And with that same aspiration in mind, we were pleased to have announced recently that Jayant Chauhan joined Innodata as Chief Financial Officer. Jayant's abilities round out an already strong finance team, with Mariz Espineli stepping into the role of Chief Accounting Officer. Beyond the traditional CFO mandate, Jayant will work strategically on capital allocation and capital markets, customer partnerships, M&A that can accelerate our strategy and investor communications while scaling the financial infrastructure of the company we are becoming.

Before I turn the call over to Rahul, let me address guidance. We are reiterating our guidance of 40% or more year-over-year revenue growth. We have some large new potential engagements in our pipeline with both existing and new customers that we believe are likely wins, but we have not yet factored them at all into our forecast at this point. As a matter of prudence, we will only factor them into our forecast when we know they're 100% won, and we can forecast the timing of revenue recognition.

I will now turn the call over to Rahul to discuss the market, our strategy, and the execution milestones that we believe prove the strategy is winning.

Thank you, Jack, and good afternoon, everyone. Before I begin, a personal note: I'm truly honored by the confidence both Jack and the Board have placed in me, and I intend to repay it with results. Innodata has extraordinary momentum, an extraordinary team and an extraordinary opportunity in front of it. I intend to build on all 3.

One of the most significant developments of the past 18 months is the increasingly pivotal role that research and innovation are playing at Innodata. It is not overstating the case to say that research has become a growth engine and the means by which we increasingly differentiate, expand existing partnerships and forge new customer relationships. Our growth is increasingly driven by research and innovation across the full model training life cycle from pretraining and posttraining to model evaluation and benchmarking. Our innovation is producing intellectual property and differentiation that is generating demand.

Several quarters ago, we talked about how we were benchmarking frontier model performance: isolating weaknesses, building remediation datasets to address those weaknesses and proving the efficacy of those datasets by training small models that were architecturally similar to the big ones. Today, we are doing much more than that. I'd like to share a few examples of what we are doing now, because the work is fascinating in its own right and because it gives you a sense of where we intend to take Innodata over the next several years. Through our research efforts, we established an early position in agentic reinforcement learning, one of the most important frontiers in AI development.

With a large lab, we won a significant new program covering personalization of long-horizon agents, which is now scaling. And we have also been awarded a second program covering reinforcement learning environments for desktop computer-use agentic tasks. In the enterprise, we see companies quick to develop AI agents but struggling to deploy them in production with confidence. We believe combining our trusted observability platform and our innovatively architected reinforcement learning gyms enable us to position ourselves as the AI deployment assurance layer. We see this as opening a huge opportunity, and this is what Jack alluded to a few minutes ago.

In the quarter, we deepened delivery of these capabilities with one big tech customer and began delivery with another. This innovation has also opened active insurance and banking conversations that we expect to convert to pilots. Frontier model builders have also become intensely focused on dynamic, long-horizon agentic evaluation. In the quarter, we released 2 public benchmarks, including one that tests how well models perform on multi-turn, long-context and multi-model interactions. A benchmark is an assembly of expert-authored prompts, rubric constraints and LLM judges configured to test frontier models. Both are designed to surface failure modes that standard leaderboards miss, things like rounding drift and instruction forgetting, precisely the failure modes frontier labs are working to improve. Each benchmark engagement results in a data strategy recommendation and sets us up to deliver scaled data generation to improve the model.

In the quarter, we also expanded our capabilities in generating training data that extends the reasoning capabilities of the state-of-the-art models, delivering across 5 frontier labs and 5 domains. As AI moves from digital tasks to embodied intelligence, we are building the required data and measurement layer. This quarter, we signed 2 research agreements with a leading university and committed to a motion capture lab that we expect to come online in the next few months, capable of collecting sub-millimeter precision data for training robots and physical AI foundation models.

In the quarter, we ran successfully egocentric data collection pilots with leading robotics companies and our data collection practice shifted from individual pilots to scoping enterprise-scale multimodal programs, including a multilingual speech program spanning 7 languages and a roughly 2 million hour egocentric program that we hope to be awarded based on successful pilot results.

Data, data engineering and data science are central to improving AI and to making it safe and trustworthy. That centrality is what enables our research to deliver capabilities across many different spheres. Data engineering innovations can solve big AI challenges, including in domains where you might not expect to find us. We mentioned one such domain in our Q4 call: how we had developed an AI model for drone and other small object detection that exceeds prior state-of-the-art benchmarks by 6.45% and how in a field where progress is often measured in fractions of a percentage point, a 6.45% improvement is a material advance. We are now working on demonstrating that capability to the government.

Another example: as we announced earlier this week, we released the first stage of what we are calling our AI Cyber Training Suite -- 12 datasets and evaluation systems that train AI coding agents to write secure code and to repair vulnerabilities in the company's existing software. When we tested leading open-weight models on their ability to repair verified flaws, the repair rate more than doubled after a single round of fine-tuning on just a portion of our data. Given that AI now writes a growing share of the world's code, the inability to trust that code without the security team reviewing everything it produces is a real blocker to enterprise adoption. We believe our suite has the potential to remove that blocker. Across frontier labs, federal and the enterprise, the pattern is the same: research and innovation are creating differentiated capabilities that win programs and compound into durable customer relationships. We couldn't be more excited about the opportunity ahead of us.

Jack, back to you.

Thanks, Rahul. I also want to take a few minutes to connect this quarter's results to the structural economics of our business and to spend a few minutes talking about the broader market dynamics. First, operating leverage: revenue grew 58% year-over-year, while adjusted EBITDA grew 92%, roughly 1.6x faster. Each incremental program builds on the same core operating infrastructure, so the marginal cost of the next program is meaningfully lower than the cost of building that capability from scratch.

Second, margin quality: adjusted gross margin of 49% is 9 points above our publicly stated target. The expansion is driven by mix and bolstered by the high-value pretraining programs and off-the-shelf datasets, where we retain IP and monetize the same asset across multiple customers. These are the software-leveraged economics we have been deliberately building toward.

Now, turning to the broader market dynamics. There are debates about whether we are at a peak AI CapEx, whether competition will commoditize models, and what the recent security incidents mean for the industry. Now, these debates play out against extraordinary numbers. Hyperscaler capital spending is guided to roughly $700 billion this year, nearly double last year, with estimates revised upward throughout the year. But we believe each of these debates resolves in favor of the data evaluation and assurance layer we provide.

Let me explain. If CapEx comes under pressure, monetization pressure rises and monetization runs on deployment, fine-tuning and assurance -- our business. If inference commoditizes, 2 things follow: labs engineer for use-case-specific differentiation, which requires specialized data and AI becomes more accessible to the enterprise, which requires more assurance, not less. Again, our business. If security incidents multiply, they prove the need for exactly the engineering we announced this week. Yet again, our business. However, the market moves, we believe it moves toward the work that we do.

I will now turn the call over to Jayant, our new Chief Financial Officer, to walk through the financials.

Thank you, Jack, and good afternoon, everyone. I'm Jayant Chauhan, Innodata's Chief Financial Officer. And as you know, this is my first earnings call since coming on board in July. Mariz has transitioned into the role of Chief Accounting Officer, and I'm thankful to her for her partnership in getting me up to speed quickly. I've spent the past several weeks getting to know the business and meeting our teams here in the U.S. and around the world. I'm energized by what I've found. I look forward to getting to know many of you on this call and afterwards. With that, let me walk through our second quarter results.

Revenue for quarter 2 2026 was $92.1 million, up 58% year-over-year and 2% sequentially, our 12th consecutive quarter of year-over-year growth. This exceeded analyst consensus by $5.8 million, or 7%. Adjusted gross profit was $45.4 million, representing adjusted gross margin of 49%. That was 2 percentage points higher than Q1 and 9 percentage points above our externally communicated 40% target. The improvement was driven by the mix shift towards higher-margin programs. Adjusted EBITDA was $25.4 million, or 27.5% of revenue, up 92% year-over-year. This exceeded analyst consensus of $16.8 million by approximately 50%. Net income for the quarter was $14.4 million, double the $7.2 million we reported in Q2 last year. Fully diluted earnings per share was $0.41, exceeding the consensus estimate of $0.21 by approximately 95%. Our effective tax rate for the quarter was approximately 18%, compared to our long-term target range of 23% to 25%. The lower tax rate was driven by tax benefits recognized this quarter.

Turning to the balance sheet. We ended the quarter with $250.4 million in cash and short-term investments. Excluding customer prepayments, which are a pass-through, our cash and short-term investments position was approximately $134 million, up $37 million sequentially. We remain undrawn against our Wells Fargo credit facility.

Lastly, after market close today, we will file a prospectus supplement establishing an at-the-market equity program with Goldman Sachs as lead agent, alongside a broader syndicate. The program provides an efficient supplemental capital markets tool that we can use selectively and opportunistically. Our balance sheet is strong, with cash and short-term investments of approximately $134 million, net of customer prepayments and has no debt outstanding at end of Q2. The program preserves optionality to support future growth initiatives, potential strategic opportunities and continued balance sheet strength as we scale. With that, let me close. This was a good quarter for me to step into, and I'm looking forward to building on the growth and financial discipline this team has already established.

With that, I'll turn it back to the operator. Operator, we are ready for questions.

[Operator Instructions] Your first question comes from the line of George Sutton with Craig-Hallum.

First, congrats to Rahul and welcome to Jayant. Jack, I still hope to harass you with questions regularly. So I'm curious if we can talk about the things that are not in your guidance. You mentioned some opportunities that aren't necessarily 100% booked yet, thus not in guidance. Can you give us any bigger picture in terms of what some of those opportunities look like? And will you give us more regular updates, perhaps, than just the quarterly announcements?

Sure, George. Thank you. And needless to say, I look forward to your questions as often as you'd like to bring them to me. Yes. No, we were thrilled with the quarter. Really, I think there were a lot of proof points laid down in the quarter and some of the things that we're learning as we go forward are as important to us as the financial signal that you're seeing today. The innovation that we're accomplishing, that we're producing, is teaching us -- is laying out the direction for us. It's showing us that reliability in agentic enterprise AI can be engineered. It's showing us that the kinds of innovations that we're capable of creating and the difference that we can make by operating at the data engineering layer in kind of very random things: drone detection, cybersecurity, these are just 2 examples.

When I look at the set of opportunities we have, they run the gamut. And now I'm responding to your question about the things that are significant, some quite large things that are not in our guidance today. They run across our capabilities. There are things that are on the government side and the enterprise side. There are things that are on the frontier model side. A lot of the capabilities that we're demonstrating now in agentic AI, both deployment and training, are prominent in our pipeline. We're excited about it. But from a methodological perspective, we've maintained the discipline to count our chickens only once they're hatched. So we're looking forward to sharing more as we proceed through the second half of the year. We think it's going to be exciting.

So the security incidents that we're starting to see in AI are obviously concerning and seem to have created a very nice new opportunity for you. I wondered if you can just walk through that and, obviously, if you can bring to bear the press release from a couple of days ago with some of your capabilities. What does that mean in terms of opportunity for you?

Yes, good question. So I think when we look at the problems that the enterprise is having, they want to embrace agentic AI, but can they trust it? What are the reasons that they may not be able to trust it? And certainly, when they're reading about models escaping their sandboxes or gaining elite cyber capabilities when they escape containment and things like this, that becomes a real concern.

Now one of the reasons that concern exists is a lot of the frontier models that are capable of these cybersecurity disruptions were themselves built on training data that contained unpatched code. It's fascinating. So basically, if you can identify the things that went into their training data mix and you can build an agent that can detect those code aberrations, can detect the code that's been introduced even when patches were subsequently introduced. And then from that, if you can enable that AI to generalize to new novel threats, things that it hasn't seen and identify threats that are in the existing software, you've got a very capable set of technologies that enable the enterprise to more safely adopt AI. So we're having some interesting discussions about that. We think it's another example of the kinds of innovation that we're increasingly capable of.

And just one other question. Obviously, we're seeing more federal government testing of models before they are released and a lot of it through red teaming. Can you just give us a sense of your involvement in the broader federal area?

So there are a couple of things there. I think we're having a lot of interesting discussions with players in the government about how we can partner with them and where we can cooperate with them. We're also discussing things with agencies, the ability to be represented in the TradeWinds marketplace is an accepted solution for different things is a huge opportunity, a huge advantage that we now have. I think from a perspective of what will be the federal government's relationship with AI, there are 2 things there: there's first, they're very much accelerating their ability to procure AI solutions and get the best. The other thing that we're seeing is the frontier model companies are inviting the government proactively to help them regulate the agency. So when you look at what the eventual need will be for things like benchmarks and evaluations and red teaming, we released 2 benchmarks this quarter that we think are very novel and very useful to deliver those kinds of things and evaluation work on behalf of the government is an opportunity that we're tracking.

Your next question comes from the line of Allen Klee with Maxim Group.

You mentioned one of the positives this quarter was a higher mix of higher-margin projects. I was wondering, should we think of this as a trend towards that? Or maybe that was just the mix this quarter and it may revert back to where it's historically been?

Yes. So it's a very good question. I'm going to answer it in the following way. I think it's both. And now let me explain what I mean by that. We do bid on work that has a lower gross margin than the one that you're seeing today. Some of those projects could be large. We would intend to take those on. If we win those, I think the cash flow from them will be -- we would anticipate to be quite compelling. Would that mean the gross margin on a weighted basis would decline somewhat? It would. On the other hand, from a strategic perspective, the things that we're working on, the things we're innovating, will likely have a higher revenue quality. And we measure revenue quality at or we think of revenue quality as a function both of gross margin and the recurring nature of that revenue. So I think over time, strategically, it's going to trend upward. I think on a quarter-by-quarter basis, it will depend on product mix.

Then also you talked about using off-the-shelf datasets more often to do the training. I'm just trying to -- could you explain a little of like do you own the data or you get to use it and use it multiple times? And if you don't do that, how you're accessing the data?

Sure. So the off-the-shelf datasets up until now, and I'll come back as to why I said that. For the most part, up until now, our datasets that we engineer, and we engineer them around model deficiencies that we detect in our benchmarking. So when we see that there is a deficiency or when we see that -- or when we identify a capability that the frontier models are looking to create, and we can engineer a dataset that helps them get there, rather than waiting for them to request that of us, we build that dataset, we maintain or we retain the IP associated with that dataset, and we enable them to use those datasets for training their models.

It's good for everybody, right? It's good for our customers and it's good for us. And that's one of the contributors to higher margin profiles. There are also times when on behalf of someone else who owns a dataset, we will represent them. We will perhaps do some engineering to that data. We will configure it so that it's ready for models to be trained on it. And then we will invite our customer partners to utilize that data as well. But most of what you're seeing today is data that we've figured out how to assemble around particular model needs and frontier model capabilities.

My last question is, in the most likely case scenario, is there any reason that it would be likely that there would be a sequential decline in revenues in the third or fourth quarter?

So within the constraints of our business model, it's certainly possible. And if it were to occur, I don't know that I would particularly care. So what I care mostly about is where we're taking the company and where it's going, not quarter-to-quarter performance. The kinds of innovations that we're producing today, the track record we're getting, the new customers that we're winning, I think over time, will continue to inure to our benefit. And I think that we're going to continue to grow this company in a very significant way over the next several years. If we were to win a very large onetime project that were delivered in 2 quarters and then there were an air gap after a third quarter, would I consider that a failure? Not at all. What I would consider a failure is if we're not maintaining the relevance that we are right now to our customers and if we weren't identifying huge market opportunities that I believe we'll be able to exploit over the next several years.

This concludes our question-and-answer session. I will now turn the call back over to Jack Abuhoff for closing remarks.

Thank you very much. So to wrap up, Q2 2026 was another record quarter for Innodata. It was an across-the-board beat. We delivered 58% revenue growth, 49% adjusted gross margin, 92% adjusted EBITDA growth and significant cash generation. It was our 12th consecutive quarter of year-over-year growth as well. We're seeing that diversification is happening in practice. Our largest customer declined to 37% of revenue while our overall business grew. And the customer that generated essentially no revenue a year ago is now our second-largest customer.

We announced a planned leadership transition. Rahul will become our President and CEO on September 30. I'll become our Executive Chairman. I'll be focused on building long-term differentiating capabilities across our enterprise and federal markets. Meanwhile, Jayant Chauhan has joined as CFO, further strengthening our financial leadership and enabling me to do some of the things that I want to do. As one of the company's largest shareholders, I believe this is a tremendous path forward to very significant shareholder value creation.

As we've discussed, our growth is increasingly research-driven and innovative, from novel benchmarks and reinforcement learning environments to capabilities in agentic deployment assurance, physical AI. I think we're at the very early stages of many of this. So we're very excited about what lies ahead. We're very confident that 2026 can be a tremendous year for Innodata, and I thank all of you for continuing to be on this journey with us.

Ladies and gentlemen, that concludes today's call. Thank you all for joining. You may now disconnect.

Well, good day, everyone, and welcome to the Innodata First Quarter 2026 Results Conference Call. Just a reminder that this call is being recorded. At this time, I will hand things over to Ms. Amy Agress. Please go ahead.

Thank you, operator. Good afternoon, everyone. Thank you for joining us today. Our speakers today are Jack Abuhoff, Chairman and CEO of Innodata; Rahul Singhal, President and Chief Revenue Officer; and Marissa Espineli, Interim CFO. Also on the call today is Aneesh Pendharkar, Senior Vice President, Finance and Corporate Development. We'll hear from Jack and Rahul first, who will provide perspective about the business, and then Marissa will provide a review of our results for the first quarter. We'll then take questions from analysts.

Before we get started, I'd like to remind everyone that 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, assumptions and estimates and are subject to risks and uncertainties. Actual results could differ materially from those contemplated by these forward-looking statements. Factors that could cause these results to differ materially are set forth in today's earnings press release 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 undertake no obligation to update forward-looking information.

In addition, during this call, we may discuss certain non-GAAP financial measures. In our earnings release filed with the SEC today as well as in our other SEC filings, which are posted on our website, you will find additional disclosures regarding these non-GAAP financial measures, including reconciliations of these measures with comparable GAAP measures.

Thank you. I will now turn the call over to Jack.

Thank you, Amy, and good afternoon, everyone. Q1 was a record quarter for Innodata, and it was record-setting by a wide margin. Revenue, adjusted gross profit, adjusted EBITDA and cash all reached new highs. Revenue was $90.1 million, up 54% year-over-year, exceeding analyst consensus by approximately $13.6 million or 18%. Adjusted gross margin was 47%, a 6-point sequential improvement and 7 points above our 40% public target. Adjusted EBITDA was $25 million or 28% of revenue, exceeding consensus by 139%. We ended the quarter with $117.4 million in cash, up $35.1 million sequentially with no debt drawn against our recently expanded $50 million Wells Fargo credit facility. These are not incremental improvements. They are step change results.

Today, we have printed a quarter that has beaten our annual revenue of just 3 years ago. Just as importantly, our results demonstrate that the strategic position we have been building is now translating into scale, margin expansion and cash generation. With 1 quarter behind us and progressively increasing visibility, we are raising our full-year 2026 revenue growth guidance to approximately 40% or more. That is up from the 35% or more we guided to on our last call just 10 weeks ago.

We continue to view this guidance as prudent. There are several potentially large programs we have not included in our forecast. As timing and scope get finalized, we'll adjust our forecast accordingly. The fact is that the year is developing faster and across more customers and programs than our original plan contemplated.

Today, we are also announcing a new set of engagements with one of the world's leading big tech companies. We believe these engagements could potentially generate approximately $51 million of revenue this year. 12 months ago, in the first quarter of 2025, our revenue from this customer was 0, but this year, we expect it to become our second largest customer. Moreover, we believe this relationship will continue to expand over time. We see considerable headroom both within the current program and from additional programs that we're actively discussing with this customer.

For several quarters, we have told you that 2026 growth would come from a broader and more diversified customer base. Our Q1 results, together with our outlook for the year, demonstrate that the diversification we plan for is now happening in practice. This year, we expect our largest customer to represent a decreasing percentage of total revenue even as our absolute dollar revenue with that customer expands. With our largest customer, we continue to grow as we diversify into more organizations and more AI workflows and partner with them on their flagship next-generation AI program. At the same time, growth outside that account is accelerating even faster.

In Q1, revenue from our other big tech customers in the aggregate grew 453% year-over-year. We believe this represents one of the strongest forms of customer diversification the company can deliver. The largest account continues to grow in absolute dollars, while the rest of the customer base grows even faster.

I will now turn the call over to Rahul to discuss where we see the market going, how our strategy comports with our market thesis and how our execution milestones offer proof that our strategy is enabling us to win.

Thank you, Jack, and good afternoon, everyone. It's great to be with you today, especially in a quarter where we have so much progress to share. I'll start with the market in which we believe we today have our strongest strategic position, the AI Innovation Labs and Frontier model builders. We define this as roughly 20 organizations globally that are developing the most advanced foundation models, including the major U.S. labs and sovereign-backed assets.

We are seeing real accelerating momentum across this customer set. We believe this is because we are aligned with where Frontier AI is going. Our conviction is straightforward. AI is moving from text to multimodal from one-shot answer to multistep reasoning from passive assistance to autonomous agents and ultimately, from purely digital tasks to embodied intelligence and robotics, autonomous systems and physical AI applications. Each step along that trajectory makes data engineering more specialized, evaluation more demanding and expert judgment more important.

That is exactly the work Innodata has been preparing for. We have deliberately moved up the stack towards high-quality pretraining data, expert graded reasoning data, agent trajectories, evaluation infrastructure and trust and safety services. The clearest evidence that this strategy is working is now showing up in our revenue.

I'll start with the major Q1 set of new engagements Jack just described. This customer is using us across the life cycle of frontier model development. We are producing high-quality text-based pretraining data at scale, including STEM data sets across physics, mathematics, chemistry, engineering, biology. These are the kinds of expert grade data used to teach models to reason at graduate and PhD levels.

On post training, we are working on data sets for advanced reasoning, creative writing and agent improvement. This customer chose us because our delivery infrastructure combines deep subject matter expertise, a global expert network, leading data scientists and engineers and secure physical infrastructure that allows us to operationalize large complex data requirements. That combination is hard to assemble, harder to scale and increasingly central to what Frontier Labs need. We are seeing the same thing playing out across the broader Frontier Lab customer base.

We are pleased to announce that a large hyperscaler just selected us to become its global trust and safety partner for evaluating models before they're released into production. We were selected because of a differentiated view of how frontier models should be tested holistically for safety, reliability and real-world readiness. We anticipate that our initial statement of work will lead approximately $3 million of potential annual run rate revenue with likely further expansion.

At another company, one of the world's largest cloud and commerce companies, we have moved from execution partner to strategic partner. We believe we have line of sight on approximately $7 million of total contract value across the customers' trust and safety and responsible AI programs, most of which we believe will start later this year and on more than $8 million of total contract value across AI and safety, scale data generation, global responsible AI testing and physical AI.

Physical AI is an important element of our broader thesis. As AI moves into the real world, the data, testing and safety requirements become more complex and more mission-critical. We will talk more about this later in today's call. We're also seeing strong traction in potential 7-figure opportunity with several of Asia's leading tech companies and a major European frontier AI lab. Our customer base is broadening and the pattern is consistent. Relationships start with a focused initial use case, we execute well and work expands and becomes more specialized. We read every day about the significant AI capital investment our customers are making towards physical infrastructure, data centers, networking and compute.

Infrastructure alone does not create usable AI systems. AI labs also require model training, evaluation, safety and continued improvement throughout the AI life cycle. This is the work we do. It is iterative, deeply embedded and structurally compounding. With each new cycle, we learn more about the customer stack, evaluation [indiscernible], security posture and model improvement priorities. This institutional knowledge, we believe becomes an asset that compounds and makes us more valuable over time.

Reuters recently reported that Morgan Stanley now expects AI-related CapEx by the 5 major U.S. hyperscalers to top $800 billion this year and to reach $1.1 trillion next year. Goldstein meanwhile estimates cumulative AI infrastructure spend could reach $7.6 trillion by 2031. While those estimates are not our revenue forecast, they underscore the scale of the ecosystem being built around AI and speak to the scale of the specialized data, evaluation and safety infrastructure that will be required to make that capital productive.

The Frontier Labs ambitions increasingly extend into robotics, intelligent devices, complex reasoning and real-world scenarios, all of which create more complex data and evaluation requirements. In fact, that same trajectory thesis also explains why we are investing in both federal and enterprise markets. As the application of AI moves from chatbots to digital agents to embodied intelligence, we expect federal and government-aligned customers to become meaningful long-term growth vectors.

On the strength of our conviction, we launched our federal practice last September, and it continues to gain market traction. Our engagement with Palantir is generating strong customer feedback in computer vision, and we have initiated work with a major federal systems integrator. We were also just selected as a finalist for potentially significant award. We believe making it this far in the selection validates the suitability of mission-critical regulated AI work.

In Q1, Innodata Federal in concert with the robotics and computer vision practice gained traction with several U.S. government research agencies and specialized AI vendors. As we previously reported, we were awarded a prime contract position under the Missile Defense Agency's Shield program, part of the broader golden dome strategy, positioning us to compete for future task orders as programs scale. We believe these are early proof points showing that the embodied AI portion of our thesis is already beginning to monetize in the federal market. We are encouraged by the White House AI Action Plan released in July 2025 that identified more than 90 federal policy actions to accelerate AI adoption, infrastructure, evaluation and government use.

The same thesis applies to enterprise AI. In enterprises, we anticipate an exploding need for data engineering. This quarter, we had active programs across major hyperscalers, networking and consumer Internet customers, covering use cases across customer service, data center operations, financial operations, legal workflows and intelligent content delivery.

Much of the work we are doing involves building and deploying agents, and we see firsthand the huge business impact these autonomously acting agentic systems will likely have for our customers. At the same time, we observed the gap that exists between the business value they want to extract with agents and the means by which they gain confidence that the agents are working as intended. To address this gap, we have built an evaluation and observability platform, which we released this quarter in beta.

Our platform is a control plane for agent systems. It helps enterprises evaluate agent behavior, inspect traces, monitor live performance, that regressions early and maintain audit trails and production. Over time, it allows experts to supervise larger and more complex workloads with fewer resources and to optimize agent token consumption.

I'm thrilled to report that just last Friday, we signed our first major platform opportunity, a $1 million engagement with one of our hyperscaler customers. We also now have 15 other companies actively evaluating the platform. Equally exciting, we are in discussions with 2 leading hyperscalers about becoming channel partners to distribute our platform to their customers. This could be a game changer, potentially enabling us to scale the platform in a manner that would not be possible with a direct sales force alone.

External market data supports the enterprise thesis. Citigroup recently raised its global AI market forecast to more than $4.2 trillion by 2030 with roughly $1.9 trillion tied to enterprise AI. Before I turn the call back over to Jack, I want to emphasize something. Each of these 3 vectors, Innovation Labs, federal and government-aligned customers and enterprise AI is a multi-customer business with its own structural tailwinds. Together, they form a diversified growth thesis and gives us confidence to anticipate both additional upside as 2026 unfolds and continued growth in 2027 and beyond.

Okay, Jack, I'll turn the call back to you now.

Thanks, Rahul. I'm going to take the next few minutes to connect the progress Rahul just described to how we believe our business model can flex over time at both the gross margin line and the adjusted EBITDA line.

On gross margin, we see the opportunity for expansion as we develop capabilities that decouple revenue growth from linear headcount growth. One example is off-the-shelf data sets, where we retain the IP rights, enabling us to resell the same data set to multiple customers. We are increasingly using this model for data sets that have proven particularly effective at solving specific model training goals. The economics can be attractive, advancing our long-term objective of adding more software leveraged offerings to the mix. Our Q1 margins benefited from this offering, and we expect our Q2 margins to benefit as well.

A second example is platforms. Rahul discussed the important milestone we achieved in Q1 with the launch of our agent observability platform. Beyond that, we have built platforms that generate data pipelines for agent optimization and adversarial simulation. These are proprietary technologies for generating synthetic data in a highly novel way, enabling scaled human judgment to be applied more efficiently, more consistently and across larger workloads, translating to more revenue for us with fewer people.

Turning to adjusted EBITDA. Our results show that operating leverage is inherent in our business. In Q1, revenue grew 54% year-over-year, while adjusted EBITDA grew approximately 96%. Put differently, adjusted EBITDA grew roughly 1.8x faster than revenue. That is operating leverage by definition. Now the reason is structural. Each incremental program builds on the same core operating infrastructure, so the marginal cost of adding the next program is meaningfully lower than the cost of building that capability from scratch.

As revenue growth accelerates, we expect this operating leverage to remain an important feature of the model. The reinvestment we are making in the business supports both of these leverage points. On go-to-market, we are adding talent to improve account penetration and market reach and putting in place compelling channel partnerships. On product and research, we have meaningfully expanded our internal research team over the last several quarters, attracting senior scientists and engineers from leading AI labs and top universities. This investment helps us continue to differentiate as we move up the value chain toward evaluation, agent reliability, alignment, risk sensitive control and synthetic data.

I want to highlight one specific milestone that captures the kind of research organization we are building. One of our researchers, Esther Derman, recently had 2 papers accepted at the 2026 International Conference on Machine Learning, or ICML. ICML is one of the most prestigious AI research venues in the world. One of Ester's papers received the so-called Spotlight designation, which places it at the very pinnacle of AI research.

Put that in context, ICML reported that 23,918 submissions entered review for 2026, which interestingly was twice the number from the year before. Of this close to 24,000 papers, just 6,352 or 26.6% were accepted, and of that, a mere 536 or 2.2% were selected as spotlight papers. Esther's accepted papers focused on model-based off-line reinforcement learning and risk-sensitive reinforcement learning. The spotlight paper is on risk sensitive reinforcement learning.

Both areas map directly to problems our customers are working to solve, how to train AI systems efficiently and how to make AI systems behave reliably in environments where the cost of failure is high. We are excited about Esther's accomplishment, and we expect more achievements like this from the team in the quarters ahead. The depth of research talent we are building is becoming a meaningful competitive advantage.

In our last call, I said we are entering a golden age of innovation at Innodata. Today, I'll reiterate that even more strongly. We are building proprietary technologies that allow us to construct unique data sets, measurably improve model performance and bring Agentic systems to production readiness. Raul and I are focused on some highly creative ways to translate this innovation into the strongest possible economic outcome for Innodata and its shareholders. We expect to provide additional updates on this as the year progresses.

I will now turn the call over to Marissa, who will walk through the numbers.

Thank you, Jack, and good afternoon, everyone. Revenue for Q1 2026 was $90.1 million, up 54% year-over-year and 24% sequentially from $72.4 million in Q4 2025. This exceeded analyst consensus of $76.5 million by approximately $13.6 million or 18%.

Adjusted gross profit was $42.6 million, representing adjusted gross margin of 47%, that was 6 percentage points higher than Q4 and 7 percentage points above our externally communicated 40% target. Adjusted EBITDA was $25 million or 28% of revenue. This exceeded analyst consensus of $10.4 million by approximately 139% and represented a 6-point margin expansion from Q4.

Net income for the quarter was $14.9 million. Fully diluted earnings per share was $0.42 compared with consensus of $0.08. Our effective tax rate for the quarter was approximately 14%, below our long-term target range of 23% to 25%, primarily reflecting tax benefits recognized during the quarter.

We ended the quarter with $117.4 million in cash, up $35.1 million from $82.2 million at year-end 2025. The increase reflects continued strong profitability, disciplined working capital management and customer prepayments related to our pretraining programs. We remain fully undrawn against our Wells Fargo credit facility, which we successfully renewed and expanded during the quarter from $30 million to $50 million on the 3-year term. We believe the expanded facility reflects our increased scale, profitability and balance sheet strength.

As Jack noted, we are raising our 2026 revenue growth guidance to approximately 40% or more. We continue to view that guidance as prudent. As Jack mentioned, there are several potential large programs we have not included in our forecast. As timing and scope gets finalized, we'll adjust our forecast accordingly.

One reporting note, effective this quarter, we are reporting our financial results as a single operating segment. We previously reported 3 operating segments: BDS, Agility and Synodex. The shift to single segment reporting reflects the transformation of our business, strategy and operating model, driven by our focus on Agentic AI technologies and by the increasingly integrated way we manage and deliver our services.

Thank you, everyone, for joining us today. Operator, please open the line for questions.

[Operator Instructions]. We'll take the first question from George Sutton, Craig-Hallum.

Great results, guys. I did miss the first few minutes, Jack, so I apologize if this is redundant, but I wondered if you can go in a little more detail on the $51 million contract that you announced today. Just give us a sense of the timing of that, the potential broadening of that over time or into next year, for example?

Sure. Thank you, George. Very excited about that win. It's a very significant win for us from a dollar value perspective, but in addition to that, what's even more exciting is that we now believe we've got another growth partner of significance. It's pretty clear to us that we expect this customer to be our second largest customer this year, which is very meaningful. There are active conversations going on with the customer about things that are not in that $51 million, other things that we can be doing with them.

The work that we're doing goes across pretraining, mid-training, post-training activities as well as evaluation. They're seeing us as a full-service shop, and they're very much leaning into several of our later or latest innovations, which is also tremendously exciting. They're a very large company. They're one of the big techs, and we're excited about the partnership.

I wondered if we could just think through even 12 months ago, 18 months ago, when the vast majority of your work seemed to be on the post-train side. Now we're talking a much broader set of use cases. You're talking about trust and safety and robotics and federal and the new platform evaluation and observability. Can you just give us a sense of how different the scope of what you're working on is today versus then? I assume that could only increase from here.

Yes. I mean -- so -- and we mentioned the term a couple of times in the prepared remarks, we talked about our strategic trajectory. I think that's like really super critical. Our hypothesis all along has been that these tools are going from one-shot answers to multistep reasoning engines that's giving way to autonomous agents, which are giving way to embodies intelligence.

What's critical is along that categorical vector path, if you will. The thing that will propel that along and what will become -- where companies will have even -- we predict even more voracious appetites for data is making that journey across that trajectory. At the same time, on the other axis, you think of it like a quality vector. It will be the data mixes and the quality of data that determines within any one of those categories, how well the AI is performing.

Strategically, we're working on 2 things. We're working on what are the data sets that are going to be required, what are the data capabilities that will be required in order to move along that vector of capabilities and then what does the data look like? How do we create more interesting data mixes and higher quality data that helps our partners achieve the quality that they're seeking within any one of those categories. Whether it's pretraining, mid-training, post-training evaluations, safety, to us, it's what is required in that category and what's required at the point in time as determined by research in order to achieve the best results.

One last question. Obviously, a quarter ago, we built in a fair amount of investment that you were making in sales and marketing and R&D, and you meaningfully exceeded any expectations we had on the EBITDA line. This was not the quarter we were expecting a good EBITDA progress. Can you just talk about what those investments yielded you, what they might yield you going forward?

Yes. We talked a little bit about today the potential of channel partnerships with our observability platform. We talked about other platforms that we have that help make agents perform better and make them safer. We talked about off-the-shelf data sets. Those are all things that we've been working on within our R&D labs and that we're continuing to work on.

Then there are some other things that we're starting to work on, some things that I think we'll be announcing maybe as early as next quarter, actually. We see a tremendous ROI that we're getting from our R&D organization. We're thrilled with the people that we've got. We're thrilled with the output that we're getting. What we're seeing is that's enabling us to move along the trajectory that I described to be a little bit ahead of where our customers need us to be and to increasingly be a thought partner to our customers to bring them new ideas, to encourage them to come to us with their problems, not just their orders. That's huge for our business.

Next up is Allen Klee from Maxim Group.

Congratulations. In terms of following up a little bit on one of the last questions of the investments that you're making to grow, you did talk about how you're going to get some better margins from certain things you're deploying. Is there a way to think about like as we go through the year and specifically next quarter, should we think that there -- for some reason next quarter, there would be a more than normal jump in investment expenses? Or is there any reason why there might be a timing that revenue might not be what would normally track?

Yes, I don't think that you should anticipate a step change in investment at this point. We're comfortable and we're getting a great return on what we're doing today. It will increment that up. We certainly don't see it flatlining. It will continue to increase. I think the enormous operating leverage in the model will enable us to do that without having to take a big hit on profitability. I think that we're able to really pull off the hat trick here, both revenue growth, margin growth and innovation growth as we move along the trajectory of helping models get smarter and helping them achieve extraordinary levels of intelligence.

I might have missed something that was said when there was a discussion on the segment. Are you still breaking out the 3 segments? If you are, could you provide what the revenues were for each one? Or is this all getting combined now?

It's all getting combined now. We're reporting on a consolidated basis. We ran the tests for segment reporting and made the determination that it's more -- that it's appropriate for us now to be reporting on a consolidated basis. Within the Synodex and Agility platforms, we're doing some really interesting things helping to think through -- and everybody has probably been reading about where software going, is software becoming service.

We're doing some things to enable that to take place. We see enormous opportunities for Agentic technologies within those businesses and potentially the ability to transform them. We're managing them differently. We're not thinking about small incremental improvements in revenue. We're thinking about fundamental step changes in the purpose of those businesses and what they can achieve for customers.

Then when you were talking about the Frontier lab, could you maybe just give an example of what is being provided?

Sorry, Frontier labs generally or any specific frontier lab? I'm not sure I'm following the question.

I'm just trying to understand a little more of like what specific area of what you provide that this is adding to.

Sure. If you take some of the wins that we were describing on our call today for the large $51 million contract, we're providing what's called pretraining, mid-training and post-training data. Soon we anticipate providing evals as well. You can think of those as all classifications of data that's required in order to train and fine-tune large language models. In terms of -- one of the other customers we talked about, we're providing trust and safety services. We're evaluating models. We're testing them. We're isolating areas where they're underperforming. We're prescribing the data mixes that are required in order to mitigate that performance.

Similarly, on another one of the wins that we talked about or soon to be wins, scaled data generation, large-scale data to train and improve models, testing for alignment with responsible AI. We're getting into creating data sets that are required for physical AI. You can think of physical AI as embodied intelligence or robots. It's really along the full spectrum of capabilities that are required by the foundation model builders from a data perspective in order to support their products.

Up next is Hamed Khorsand from BWS Financial.

Just first question is, was there anything of onetime nature in the first quarter results as far as the revenue is concerned? Or should we expect this to be a good baseline going forward?

I'd say both. There are things that we're doing that we won't be doing next quarter. There are things we're going to be doing next quarter that we're not doing this quarter. I think that it was a strong quarter. I think next quarter is going to be a strong quarter. I think the quarters after that are going to be good. We're providing -- we're not providing quarter-by-quarter revenue guidance because the fact is that things do start and stop. When we talk about the phases of training a model, those don't necessarily dovetail perfectly, but we've got more and more things going on, and that tends to even things out. We're doing some things now increasingly that are of an ongoing nature.

No, I don't think you should think of the quarter as aberrational at all. I think that as we move through the year, there are going to be things that we're doing increasingly that are driven by innovation that are going to be margin accretive, margin supporting. Yes, we're excited about the year.

Then my other question was, has the composition of revenue changed at all? Or is it still -- the scope of work is still the same? I mean you're talking about something that might happen in the future as far as the Agentic and the valuations and so forth.

No, these are things we're doing today. I mean, the thing that doesn't change is our mission for the company, and our mission is to be the data partner to foundation model builders and to be the intelligence infrastructure layer for enterprise. That's not changing. What does change is as the models and the capabilities seek to do more and perform better, the mix of what we do does change. That's our job to stay research-led and to ensure that we're a little bit ahead of where our customers need us to be.

Everyone, at this time, there are no further questions. I'd like to hand the call back to Mr. Jack Abuhoff for any additional or closing remarks.

Thanks, operator. Yes, to wrap up, Q1 '26 was a record quarter for Innodata across all the key metrics that we're reporting, revenue, adjusted gross profit, adjusted EBITDA, cash. We delivered 54% revenue growth. We expanded margins meaningfully. We generated significant cash without having to draw on a credit facility.

Based on these results and our forward visibility, we are raising 2026 revenue growth guidance to approximately 40% or more year-over-year. We continue to view this outlook as I'll use the term prudent. We see potential upside as additional programs that are not included in that forecast convert them scale. A big tech customer that generated no revenue for us 12 months ago is now on track to become our second largest customer this year. Our customer concentration is improving in the very best possible way, faster growth from the broader customer base, while our largest customer continues to grow in absolute dollars.

We're also continuing to innovate at an increasingly rapid pace. The strength of our research bench is showing up in customer outcomes and in external recognition like Esther's 2 ICML 2026 paper acceptances and her 1 Spotlight designation, really exciting stuff. We launched our valuation and observability platform in beta in the quarter, and no sooner did we launch, and we closed a $1 million opportunity with one of the world's largest hyperscalers around that platform. We're really excited about what lies ahead. We're confident that 2026 is going to be an exciting and tremendous year for the company. Yes, I thank everybody for being on the journey with us.

Once again, everyone, that does conclude today's conference. We would like to thank you all for your participation today. You may now disconnect.

Good afternoon, ladies and gentlemen, and welcome to the Innodata to Report Fourth Quarter and Fiscal Year 2025 Results Conference Call. [Operator Instructions] This call is being recorded on Thursday, February 26, 2026. I would now like to turn the conference over to Amy Agress, General Counsel. Please go ahead.

Thank you, operator. Good afternoon, everyone. Thank you for joining us today. Our speakers today are Jack Abuhoff, Chairman and CEO of Innodata; and Marissa Espineli, Interim CFO. Also on the call today is Aneesh Pendharkar, Senior Vice President, Finance and Corporate Development. Rahul Singhal, President and Chief Revenue Officer, is unable to be here today, but looks forward to joining us on our next call. We'll hear from Jack first, who will provide perspective about the business, and then Marissa will provide a review of our results for the fourth quarter and fiscal year 2025. We'll then take questions from analysts.

Before we get started, I'd like to remind everyone that during this call, we will be making forward-looking statements, which are predictions, projections and other statements about future events. These statements are based on current expectations, assumptions and estimates and are subject to risks and uncertainties. Actual results could differ materially from those contemplated by these forward-looking statements. Factors that could cause these results to differ materially are set forth in today's earnings press release 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 undertake no obligation to update forward-looking information.

In addition, during this call, we may discuss certain non-GAAP financial measures. In our earnings release filed with the SEC today as well as in our other SEC filings, which are posted on our website, you will find additional disclosures regarding these non-GAAP financial measures, including reconciliations of these measures with comparable GAAP measures. Thank you. I will now turn the call over to Jack.

Thank you, Amy, and good afternoon, everyone. Q4 was another strong quarter for Innodata. We generated $72.4 million in revenue, reflecting 22% year-over-year growth. This brought our full year revenue to $251.7 million, representing 48% year-over-year growth for 2025. Our Q4 consolidated adjusted gross margin was 42%, exceeding our externally communicated target of 40%. Our adjusted EBITDA totaled $15.7 million or 22% of revenue, also exceeding analyst consensus by $1.2 million. In fact, our results exceeded analyst consensus across the range of key metrics, including revenue, adjusted EBITDA, net income and EPS. We ended the year with $82.2 million in cash, up sequentially by approximately $8.4 million.

We achieved these results while making meaningful growth-oriented investments in both COGS and SG&A. In COGS, we carried capacity ahead of revenue ramp, which consistently proved to be the right move. And in SG&A, we invested in engineers, data scientists and customer-facing account leadership, which investments also proved prudent, yielding innovation that has expanded our opportunities. We believe our business momentum to be at an all-time high. We are seeing robust demand across the entire generative AI life cycle, spanning development, evaluation and ongoing model optimization. And we believe we are gaining traction with a broad and diversified number of large customers.

As a result of market demand and growing traction, we anticipate another year of potentially extraordinary growth in 2026. We currently estimate our 2026 year-over-year growth to potentially be approximately 35% or more. This estimate reflects active programs, recently awarded wins, late-stage evaluations and opportunities where we have clear line of sight. Because we are early in the year and because LLM initiatives spin up quickly, we believe there may potentially be significant upside to this range. However, we prefer to guide conservatively and adjust upward as visibility increases. At the same time, given the scale and complexity of the programs we support, timing variability in customer ramp schedules, budget approvals or shifts in research priorities could influence the pace at which revenue materializes.

Embedded in our outlook is the expectation that spend from our largest customer will increase somewhat in the year and that the remaining customer base in the aggregate will grow at a faster rate. We expect this other customer growth to come from a mix of the Mag 7, domestic AI innovation labs, sovereign AI initiatives and leading enterprises. We believe this will meaningfully contribute to customer diversification. Our customers are moving fast, driving shorter development cycles and responding faster to research breakthroughs. In 2025, we succeeded in this environment in no small part because we followed the research, anticipated customer needs and pivoted where required.

To illustrate, in the first quarter of this year for our largest customer, we deprecated a meaningful number of post-training workflows, which represented in the aggregate approximately $20 million of annualized revenue run rate, but replaced them with a combination of new post-training workflows and scaled pretraining programs, an area of recent focus and investment. From a revenue run rate perspective, the net effect turned out positive. Indeed, we believe continuous innovation is critical to achieving our ambitious plans for 2026 and beyond. The truly exciting news is we believe we are entering a golden age of innovation at Innodata as a result of investments we have made and intend to make in the future.

I'm now going to share some of our recent innovation initiatives. For competitive reasons, we'll be appropriately circumspect, but what we share will give you a meaningful window into how we're thinking, where we're investing, successes we're having and how we intend to capitalize on the opportunity ahead. I'll briefly walk through our recent innovation in three areas: generative AI model training, agentic AI and physical AI. Before I do, I want to underscore a unifying theme. Every innovation I am about to discuss is fundamentally a data innovation. Whether the goal is more capable LLMs, more reliable autonomous agents or more intelligent physical AI systems, data quality, data composition, data validation and data engineering at scale are at the heart of the matter. These are our core competencies.

We'll start with generative AI training. Historically, customers told us the kind of training data they wanted. Increasingly, however, they are asking us to diagnose model performance, design the right training data sets and demonstrate that those data sets will materially improve outcomes. Here's how that works. We begin by identifying performance gaps using our evaluation frameworks. We then engineer targeted data sets and validate their efficacy by fine-tuning either the customer's model or a structurally similar proxy model only after we measure and demonstrate performance impact do we scale. This shifts the discussion from how much is the data to how effective is the data. We believe this shift is being driven by two forces: the accelerating pace of AI research and the cost and time incurred to train ever larger models and conversations about data efficacy play directly to our strengths.

We are also advancing methods for creating data sets that improve long context reasoning and AI model's ability to absorb and reason over very large amounts of information at once. This remains one of the industry's most important technical challenge. Solving it requires not just architectural improvements, but advances in the creation at scale of very specific types of structured training data. Creating training data that improves long context reasoning is a nontrivial problem, but we have made and are continuing to make meaningful progress on it.

The second area of innovation is around evaluating systems of autonomous agents and improving them through targeted data set creation. We believe that autonomous agents may represent the most significant business innovation opportunity since the advent of electricity. But companies quickly discovered that many AI agents that performed impressively in controlled laboratory settings degrade in real-world production. The real world is chaotic. It's shaped by edge cases, conflicting constraints, unpredictable user behavior and adversarial conditions. Addressing this is fundamentally a data challenge. Agents must be continuously trained and rigorously stress tested with data sets that are realistic, diverse and complex.

For this, we have developed a set of three highly complementary hybrid solutions. The first is an agent evaluation and observability platform. Data scientists can use our platform during development to visualize and annotate agent trace data, to build LLM as a judge evaluators, to create business aligned evaluation rubrics, to generate golden data sets for aggression testing and to generate test data at scale. Then once the agent is deployed, our platform can be used to continuously monitor its performance, perform root cause analysis of performance issues and obtain mitigation data sets. We're pleased to share that we anticipate soon kicking off a managed services engagement with a hyperscaler in which we will use our platform to create test data at scale, perform automated evaluations and identify critical model vulnerabilities in order to improve performance of its customer-facing intelligent virtual assistant.

The second innovation is a managed agent optimization pipeline designed to systematically train for and therefore, neutralize the chaos of real-world deployment at scale. The pipeline generates realistic test scenarios, automates evaluation, rigorously measures constraint satisfaction and produces reinforcement learning data sets. Using this system, we have demonstrated improvements of up to 25 points in constraint satisfaction. Importantly, agents trained using conventional techniques tend to degrade significantly as task complexity increases. By contrast, agents trained through our pipeline sustain their performance under escalating real-world difficulty. In the most demanding scenarios, the performance gap between standard approaches and our system widens to more than 31 points. We currently have multiple AI innovation labs and enterprise customers actively exploring the system.

The third solution we've designed to support enterprise agentic AI is an adversarial simulation system that generates high-quality semantically diverse and scalable adversarial attacks to stress test agents. The system generates a full spectrum of attack types, direct jailbreaks, indirect prompt injection via RAG pipelines, multi-turn social engineering, steganographic payloads and compound attacks that combine injection techniques with domain-specific knowledge. Once vulnerabilities are identified, it generates highly targeted mitigation data sets to strengthen guardrails. We believe our system generates realistic adversarial attacks at scale in a meaningful way that exceeds existing alternatives. Many tools on the market produce simplistic or templated hostile content that lacks the nuance and sophistication of real-world threat actors, fails to scale across diverse scenarios or relies on generic tactics that models quickly learn to anticipate and overfit to.

But by contrast, our framework is designed to simulate adaptive multistep and strategically coherent attack patterns, including highly sophisticated model extraction, cybersecurity, cyber-crime and sovereign threat scenarios that better reflect how advanced adversaries operate and allow our partners to stay ahead of emerging threats. The result is adversarial training data that is both scalable and durable, forcing models to generalize rather than memorize and enabling more robust real-world resilience.

Our work is garnering interest from CISOs and security leaders at some of the world's premier AI and cybersecurity companies as well as relevant experts in government and has led to early-stage engagements with several of them. At a time when the cyber industry is experiencing significant disruption, these capabilities bolster our position in the emerging field of AI, trust and safety, an area where we are meaningfully deepening work with several hyperscalers.

We believe Innodata is well positioned to emerge as a leader in prompt layer security, protecting AI systems at the point of interaction rather than relying solely on traditional perimeter or endpoint defenses. Taken together, we believe these solutions position us not just as a data supplier, but as a life cycle partner in agent reliability. We believe 2026 will also mark the acceleration of physical AI, intelligent systems that perceive and interact with the physical world. While robotics provides the mechanical framework, physical AI provides the intelligence. The primary bottleneck in this domain is data set quality and scale. Manual annotation and static QA sampling simply do not scale to billion-sample corpora and continuously evolving environments.

We have developed a large-scale data engineering system that incorporates structural validation, distribution monitoring, temporal consistency checks and model-in-the-loop instrumentation. This enables us to identify and correct defects in data sets before they propagate into performance failures. We're already using components of this system in the high visibility engagements we recently announced with Palantir. We recently secured a significant engagement to create foundational data sets for next-generation robotic data sets, including egocentric data. Egocentric data captures the world from the robots point of view, what it sees and experiences in motion. We are also working with a leading robotics lab to create affordance data at scale. Affordance data teaches the system what actions are possible in a given setting, not just identifying objects, but understanding how they can be used.

Egocentric data and affordance data taken together form the cognitive scaffolding that allows machines to act intelligently in dynamic environments. This work also positions us to support the development of so-called world models, internal simulations that allow AI systems to anticipate outcomes, reason about cause and effect and plan several steps ahead. World models require richly structured data sets that capture interactions over time and the consequences of actions, precisely the type of data we are now engineering.

Finally, we recently developed an AI model for drone and other small object detection that exceeds prior state-of-the-art benchmarks by 6.45%. In a field where progress is often measured in fractions of a percentage point, a 6.45% improvement is a material advance. The model improves detection fidelity under real-world conditions where small size, speed, cluttered backgrounds and environmental noise make reliable perception extraordinarily difficult. We believe this advancement has compelling dual-use implications that we are now actively exploring with potential customers.

I'd like to underscore one of the important points I just made. For decades, Innodata has specialized in creating high-quality complex data sets. Today, these capabilities are central to unlocking the next generation of AI systems. Advanced LLM reasoning, agent reliability in chaotic environments and robotics perception in the physical world, all depend on engineered data ecosystems, and this is precisely where we operate. Our innovations in LLM training, agentic AI and physical AI are not separate initiatives. Rather, they are extensions of a single strategic advantage, our ability to engineer data that measurably improves model performance in real-world conditions.

We believe our innovation pipeline will be margin enhancing as well as revenue enhancing. We expect early 2026 adjusted gross margins to be in the 35% to 40% range as we ramp up new programs with normalization toward our target 40% or better adjusted gross margins as new programs ramp up and as innovation-driven workflow scale. Automation, synthetic systems and evaluation platforms all structurally increase our operating leverage. I'll now turn the call over to Marissa, who will go through the numbers.

Thank you, Jack, and good afternoon, everyone. Revenue for Q4 2025 reached $72.4 million, up 22% year-over-year. Sequentially, revenue increased 15.7% from Q3's $62.6 million. Adjusted gross profit for Q4 2025 was $30.1 million, an increase of 6% year-over-year and 9% sequentially with an adjusted gross margin of 42%. Adjusted EBITDA was $15.7 million or 22% of revenue and net income for the quarter was $8.8 million. To reiterate, this is net of significantly expanded data science and engineering efforts that are yielding the types of innovations Jack just spoke about.

We ended the quarter with $82.2 million in cash, up from $73.9 million at the end of prior quarter and $46.9 million at the year-end 2024, and we did not draw down on our $30 million Wells Fargo credit facility. As Jack mentioned, based on our current momentum, we presently forecast 35% or more year-over-year revenue growth in 2026. Thank you, everyone, for joining us today. Operator, please open the line for questions.

[Operator Instructions] Your first question comes from George Sutton of Craig-Hallum.

Jack, I feel like I just sat through an advanced AI data science class. So thanks for that. I wanted to step back a little bit because I think people have the assumption that some of what's working for you is somewhat temporary. And I think you've done an interesting job of kind of walking us through in past quarters from post-training as a start to then pretraining. And now there are dramatic other use cases, including things like robotics and autonomous agents. Can you just talk about the breadth of the things you're seeing and sort of where you see us in this continuum of data science opportunity for you?

Sure. Thank you, George. Thank you for the question. So as we look out near term, 2026, we see ourselves as being incredibly well set up by the innovations that we invested in, in 2025. And we see that innovation output as a flywheel. We're getting better. We're getting stronger. We're creating solutions that are solving problems that are the actual impediments that enterprises have when they're looking to integrate AI into their operations.

So when you look across the spectrum of current capabilities in AI and future capabilities in things like agentic systems, physical AI, robotics, all of this boils down to challenges in terms of data engineering. Of course, there are going to be continuous improvements in architectures. There'll be bigger models. There'll be narrower models for domain-specific challenges. But at the heart of it, in terms of making systems reliable, making them safe at an enterprise level, it's going to be about innovations such as the ones we're announcing today in data sets that are used for valuation, data sets that are used for training and improving safety and reliability of models. So we think that we're at the very beginning and that our relevance is by no means diminishing, but only increasing. It's increasing not just at the level of foundation model builders, but it's clearly extending through the enterprise. We're super excited about where we are right now and about the uptake that the innovations that we're creating are having and are going to be having over the next several years.

That's great. And then just one other question. Having lived through the last couple of years where you started the years with an expectation and you then ended up meaningfully exceeding those initial expectations. Is anything set up differently going into 2026 relative to what you see in your sights relative to what you're committing to today?

No, not at all. We're following exactly that same methodology. We're really limiting our -- or we're taking a conservative approach to forecasting growth based on opportunities where we have a very clear line of sight, but where we can't predict a close rate, where we can't feel pretty confident in something happening, we're just not baking that into our guidance. Our aspiration is to surprise and to beat expectations. When I look at this year, I think it will likely be another year of doing exactly that. We're seeing enormous opportunity with a much larger set of customers. We think that, that's going to result in growth. I think it's likely that we'll be increasing guidance as we move through the year. And I think it's going to be a year where we accomplish very meaningful customer diversification.

On top of that, as we already discussed, I think it's going to be a year where we're starting to see increasingly hybrid human/technologically-driven solutions. That spells or presents the promise, I believe, for increased recurring revenue. I think it promises greater margins over time, greater stickiness, a whole lot of things that will, over time, be, I believe, consistently improving revenue quality as well on top of everything else. In terms of the work we do with foundation model builders, we're seeing tons of traction, not just in our largest customer, but in others as well. We're very much aligned with what they're looking to accomplish and things like long context reasoning improvements. We have innovations that are contributing to that. So we're tremendously excited about where we are right now.

Your next question comes from Hamed Khorsand of BWS Financial.

So just the first question, you were talking earlier about scaling your operations as the revenue ramps. Do you have enough employees now? Do you see the need to add more employees? What's your time line as far as expecting gross margin to move up from here?

Sure. Thanks, Hamed. So I think it really depends on what we're seeing. I think if we begin to project internally growth rates that are very significant, we're going to be making investments in order to ensure that we capture those growth rates. I do think that as a result of digesting some of those people investments that we're making in COGS as a result of the innovations that we're discussing, different things like that, I do think that we're going to be seeing movement back toward our target gross margins over time.

Okay. And then is there a timing as far as this pipeline of deals that you're talking about with other customers other than your largest customer?

So there are pipelines, but we're -- the deals that I'm referring to are largely deals that we're closing or have closed. So we're not depending on -- we're not speculating about what will be happening. These are things that are actively underway.

Your next question comes from Allen Klee of Maxim Group.

For 2025, I think your adjusted EBITDA margin was around 23%. And I know it's important for you to reinvest back into the business for the health of the company. My question is, is there any reason to think that you would target a higher or lower adjusted EBITDA margin than what you did in 2025?

So we're very much focused on seizing opportunity right now. We believe that we can do that and stay profitable. But we also believe that it's more important to seize opportunity and to do some of the things that we are describing and prove out those innovations than it is to track adjusted gross margin percentages and try to maintain a certain percentage. So we're going to be actively reinvesting in the business. The more opportunities we see to some extent, the more we'll be reinvesting. We do believe, though, that maintaining profitability is something that we can do while we drive very aggressive growth and while we become more progressively more critical to a larger and widening set of customers.

Okay. One of the bullet points you had on the innovation was the structural foundation for margin expansion through automation, synthetic data generation and valuation platforms. Can you explain a little what you mean of which margin expansion are you referring to?

Yes. So we're referring to, over time, gross margin expansion. So a lot of the innovations that we're working on now and that we're bringing into the market are hybridizations of software and human teams. And I think that over time, we're going to be seeing the gross margins associated with those capabilities to be perhaps well in excess of the gross margins that we target today.

Got it. That makes a lot of sense. And the last question I had was just for first quarter '26, is there anything you'd want to point out in terms of -- that might stand out just in terms of, I don't know, revenues or expense spend.

Well, I'm not going to say it's next quarter necessarily, but I think very soon we're going to be seeing quarters that from a revenue perspective are beating what our revenue was for an entire year 3 years ago. So that's pretty good news right there. As we move through the year, I think you're going to be seeing more proof points and more evidence and more engagement that we have with some very interesting companies around the innovations that we're describing. I think that we'll start to demonstrate that we're somewhat migrating from a vendor to like a foundational layer within AI ecosystems, becoming someone that is able to unlock the promise of AI within enterprise engagements, a company that's able to help enterprises embrace complex agents that plan, call tools, execute complex workflows and create a lot of value. So I think this -- I think we'll be seeing that. I think we'll see evidence of that in the first quarter. I think we'll continue to see evidence of that through the year.

Maybe one last quick one. When you were talking about your largest customer, I don't know if I fully understand, you mentioned something about $20 million that maybe is going to be replaced with more than that? Or could you just explain a bit?

Yes. I think the point that we were making there is how important innovation is to our company today and how it's becoming increasingly important. There are things that we complete and we're starting new things. And by following the path of innovation by what did Wayne Gretzky used to say by skating to where the puck is going, we're able to deprecate things that the companies no longer require, but be there for them for the things that they're -- that are the emerging requirements. Again, we're seeing the emerging requirements to be more interesting from a business perspective and a revenue quality perspective and a differentiation perspective than the things that came before.

So the investments are proving out, they're enabling us to scale and increase the breadth of engagements. They're enabling us to win new engagements and new customers that -- some of which we think are going to be very substantial. They're going to really flower this year. That's going to address the diversification issue. So we're -- when we look at 2026, we see a huge growth year. We believe that we're going to be increasing likely our guidance from what we're starting the year at. We think that the solutions and how we're embedded in workflows is going to be progressively more interesting and margin and revenue enhancing. And it promises to be a tremendous year on all of those fronts.

There are no further questions at this time. I will now turn the call back over to Jack Abuhoff. Please continue.

Thank you, operator. So yes, to wrap up, 2025 was a great year and 2026 holds the promise of being even better. In 2025, we delivered strong top line growth. We exceeded expectations across major financial metrics. We expanded margins. We strengthened our balance sheet. We invested successfully ahead of demand, and those investments proved wildly successful and set us up well for 2026. I believe that 2026 is likely to be an incredible year. We've guided to approximately 35% growth based on visibility today, but I believe there may be very considerable upside to that. We'll update you through the course of the year, much like we have done in the last couple of years.

I also want to underscore our belief that this year, we will potentially diversify our revenue stream significantly. And we believe expertly engineered data ecosystems are going to be every bit as important as bigger models and new architectures will be in terms of advancing language models, media models, autonomous agents, robots, world models and other kinds of AI that hasn't even been conceived of yet. So we're very excited about what lies ahead. We're very confident in our positioning. We're very committed to building one of the most important and we think most capable AI enablement companies in the industry. It's going to be an exciting year. So thank you all for being on the journey with us. Look forward to next time.

Ladies and gentlemen, that concludes today's conference call. Thank you for your participation. You may now disconnect.

Good afternoon, ladies and gentlemen, and welcome to the Innodata Reports Third Quarter 2025 Results Conference Call. [Operator Instructions] This call is being recorded on November 6, 2025. I would now like to turn the conference over to Amy Agress. Please go ahead.

Thank you, Michael. Good afternoon, everyone. Thank you for joining us today. Our speakers today are Jack Abuhoff, CEO of Innodata; Rahul Singhal, President and Chief Revenue Officer; and Marissa Espineli, Interim CFO. Also on the call today is Aneesh Pendharkar, Senior Vice President, Finance and Corporate Development. We'll hear from Jack first, who will provide perspective about the business, followed by remarks from Rahul, and then Marissa will provide a review of our results for the third quarter. We'll then take questions from analysts.

Before we get started, I'd like to remind everyone that 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, assumptions and estimates and are subject to risks and uncertainties. Actual results could differ materially from those contemplated by these forward-looking statements.

Factors that could cause these results to differ materially are set forth in today's earnings press release 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 undertake no obligation to update forward-looking information.

In addition, during this call, we may discuss certain non-GAAP financial measures. In our earnings release filed with the SEC today as well as in our other SEC filings, which are posted on our website, you will find additional disclosures regarding these non-GAAP financial measures, including reconciliation of these measures with comparable GAAP measures. Thank you. I will now turn the call over to Jack.

Thank you, Amy, and good afternoon, everyone. Our third quarter was another record quarter for Innodata. We delivered record revenue of $62.6 million, representing a 20% year-over-year organic growth and a 7% sequential quarterly growth. Adjusted EBITDA was $16.2 million or 26% of revenue, up 23% sequentially, showing margin expansion even after factoring in growth investments I'll be talking about later in this call.

Cash rose to $73.9 million, up by $27 million since year-end and $14.1 million since last quarter. Our results exceeded analyst expectation across key metrics. As a result of strong business momentum, we reiterate prior guidance of 45% or more year-over-year growth in 2025, and we anticipate potentially transformative growth in 2026.

This afternoon, I'll share the basis of our confidence, including the significant growth we are anticipating from existing strategic vectors and the strong early returns from new investment areas. I'll then share how we are preparing the organization to reach the next level.

I'll start with our existing strategic vectors. Since we last reported, we have continued to make substantial progress deepening relationships of trust with high dollar value big tech customers. Our deal momentum continues to accelerate with meaningful expansion across a diverse set of foundation model builders, both existing and new customers.

Of the 8 big tech customers we talked about recently on these calls, we are currently forecasting 6 of them to grow next year several quite substantially. For example, we just received verbal confirmation for additional expansions with our largest customer and verbal confirmations of the deal we expect to potentially result in $6.5 million of revenue with another big tech.

Beyond that, our expectations are grounded in the assessment of these customers' 2026 training data and evaluation budgets and the accelerating trust we believe we're earning with them through proofs of concept, pilot and scale deployments.

Now in addition to these 8 customers, we landed in Q3 or expect to finalize shortly 5 additional big techs. We believe all 5 of these new big techs are poised to contribute meaningfully to our 2026 growth. Three of these new 5, we believe, are positioned to allocate up to hundreds of millions of dollars annually to generative AI data and evaluation, and we believe we're well positioned to capture a share of that spend.

It is worth noting that 2 of these are global leaders in commerce, cloud and AI. Now let's turn to our new 2025 initiatives, 6 in total, several of which I'm sharing with you for the first time today, all of which are already bearing significant fruit and all of which we believe will contribute significantly to 2026 growth.

The first initiative has been creating pretraining data at scale. Now pretraining data teaches the model language skills and knowledge. Up until now, our business has been primarily focused on post-training data, which teaches models how to reason, follow instructions and perform tasks. But earlier this year, we observed researchers drawing increasingly strong correlations between LLM benchmark performance and the quality of pretraining data. Models that trained on higher-quality pretraining corpora consistently did a better job understanding nuance, context and intent across languages and domains.

And when we saw this research, we concluded that our customers would increasingly be seeking sources for higher-quality pretraining data. So we invested about $1.3 million to build new capabilities to create high-quality pretraining corpora. This has proven to be a great investment. We've since signed contracts we believe could result in approximately $42 million of revenue, and we expect to soon sign contracts, which we believe could result in approximately $26 million of additional revenue on top of that. So that's $68 million of potential revenue from these programs that are either signed or likely to be signed soon. These programs span 5 customers. There are only a few months in motion and are just ramping up.

We believe the majority of the anticipated revenue would flow through 2026. but we've already fully recaptured our investment. As pretraining data gains recognition as a strategic differentiator for next-generation LLMs, we believe we are well positioned to capitalize on this early trend.

Today, we announced the launch of Innodata Federal, a dedicated government-focused business unit designed to deliver mission-critical AI solutions to U.S. defense, intelligence and civilian agencies. We expect this business unit to be a material revenue generator for us in 2026 and beyond. Today, we're also announcing that the business unit has won an initial project with a new high-profile customer. We anticipate this initial project to result in approximately $25 million of revenue mostly in 2026.

We have additional projects under the discussion with this customer, and we expect them to be large. This new relationship is strategically significant, not only for its potential size, but also for the visibility and market leadership we believe it will convey.

We expect to issue a joint press release about the relationship prior to year-end. We view it as a potential game changer for our next phase of growth. Additional early market validation includes the company's first direct government award from a major defense agency, potential engagements with other prominent defense technology companies and submitted proposals spanning the DoD, intelligence community and civilian agencies.

What sets Innodata Federal apart is our ability to deliver the complete AI life cycle, not just data annotation or point solutions, but true end-to-end capabilities from data collection through model deployment and operational support. Our platforms and expertise already serve the world's leading technology companies and Fortune 1000 enterprises. We are now bringing that same proven excellence to federal missions with the security, compliance and speed that government operations demand.

We believe the timing could not be better. Federal agencies are moving decisively to adopt AI. In July, the administration released America's AI action plan and signed 3 executive orders to streamline procurement and accelerate deployment. The General Services Administration, or GSA, is now revamping its acquisition processes to make AI services easier for agencies to procure.

Historically, federal procurement has been slow and complex, but that's changing rapidly, and we intend to meet that demand and that opportunity head on. As we announced today, General retired Richard D. Clarke, a retired four-star Army General and former Commander of U.S. Special Operations Command has joined the Innodata Board. We're excited about his expertise and relationships in helping guide the trajectory of Innodata Federal.

Another key focus this year has been on advancing our participation in the emerging sovereign AI market. Initiatives by governments around the world aimed at independently developing, deploying and governing AI systems as a matter of national interest. These efforts seek to ensure national control across the entire AI technology stack from the semiconductors on which models are trained to the data that gives them intelligence. We believe this is one of the most significant structural shifts in the global technology landscape.

The drive for sovereign capability has already triggered large-scale state-directed investment programs, effectively creating government-backed demand guarantees for the entire AI ecosystem from chip makers and cloud platforms to data engineering providers like us. As we have toured several countries in the Far and Middle East, we've been struck by the level of interest in our services. These countries, in most cases, do not have a homegrown enterprise like Innodata with a proven track record of helping enable generative AI and LLM initiatives. We were rapidly engaging in advanced discussions with sovereign AI entities across several regions, and we expect to announce one or more strategic partnerships over the next few months. Their economic capabilities and desire to move quickly is truly impressive, and we could not be more excited about this newer area of growth for the company.

Meanwhile, our enterprise AI practice is also gaining traction and holds promise for 2026. It provides full stack support to help enterprises integrate generative AI into products and operations. For example, the practice is helping a major social media platform automate its content monitoring and monetization workflows using generative AI and assisting a hyperscaler to integrate generative AI into their data center operations for real-time analytics.

We expect these projects to typically start in the $1 million to $2 million range and offer strong expansion potential and repeatability. We are also in discussions about strategic relationships that could help propel our enterprise AI practice forward in 2026. The next initiative I'll talk about is Agentic AI. As I've said before on these calls, we believe Agentic AI will unlock the usefulness of generative AI in the enterprise and that autonomous agents will soon be as ubiquitous as human employees performing many of their tasks. It's still very early days for Agentic AI.

We're working with big tech model builders to evaluate and refine autonomous agents across many real-world use cases, creating evaluation models and human-in-the-loop systems designed to measure, interpret and guide agent behavior. We start by judging task success, did the agent achieve the goal? And then we analyze why the agent behaved the way it did and profile how it generally behaves to inform further fine-tuning.

These capabilities, diagnostic judge, task success judge and profiling judge are increasingly used in RLHF and RLHA frameworks for Agentic systems, where agents act autonomously across multistep real-world workflows. We've also been building agents within our agility platform as a way of enhancing the product and consulting with a number of enterprise customers about incorporating agents within their environments.

This brings me to our sixth area of 2025 investment, model safety. As agents gain autonomy, companies must learn how to monitor and continuously improve them. Our goal is to become a trusted partner to software companies and other enterprises, helping them benchmark for safety, reliability and ethical behavior. Here's one example of the work we are now doing.

Recently, we began engaging with a leading chip company to stress test its multimodal AI products, simulating real-world risks like data exfiltration, privilege escalation, instruction manipulation and multimodal injection attacks. And once we identify vulnerabilities, we generate targeted mitigation data, fine-tune the model and prove the results with repeatable benchmarks. Our objective is to increase model safety with no degradation in model capabilities from the retraining.

We believe the area of model safety holds enormous potential, so much so that we've engaged one of the world's top consultancies to help us refine our product and go-to-market strategy around model safety. That's a quick recap of the 6 investment areas that we've driven in 2025, several of which we're announcing publicly for the first time today. In every case, our investments have been modest, but our returns have been outsized and product market fit has come quickly. We believe that there are start-ups that have raised tens of millions of ambitious valuations to chase some of these same opportunities. Yet we're getting more done faster and with far less capital investment at risk.

This year, we anticipate incurring approximately $9.5 million of capability building investments in these and other similar initiatives. This includes $8.2 million of SG&A and direct operating costs and $1.3 million of CapEx. We are also absorbing costs for substantial excess capacity within the organization in anticipation of likely soon-to-be captured business. While we could have elected not to incur these costs and instead present higher adjusted EBITDA, we believe these investments represent compelling short-cycle investments that position us for accelerated growth in markets. We believe we're prepared to serve, and we believe will yield considerable benefits in 2026 and beyond.

We've also strengthened our leadership bench and operational foundation for the scale we're anticipating. I'm pleased to announce the appointment of Rahul Singhal as President and Chief Revenue Officer. Rahul joined Innodata in 2019 and has been instrumental in helping shape our strategy and building deep relationships with our largest customers. We're also welcoming 2 outstanding new Board members, Don Callahan, who brings deep digital transformation expertise from Citigroup and Time and close relationships with Silicon Valley and Enterprise CEOs through Bridge Growth Partners; and General retired Rich Clarke, who retired four-star Army General and former Commander of U.S. Special Operations Command, who brings outstanding defense insight and strong federal relationships. Their expertise aligns with our focus on big tech, defense and enterprise markets, and I'm confident they'll help guide us through our next stage of transformative growth.

Finally, I want to thank Nick for 5 years of Board service. Nick has been tremendously helpful to me and to the company. He is stepping away to devote his time to a new opportunity outside of our markets, and we wish him very well. With that, I'll turn the call over to Rahul.

Thank you, Jack. I'm honored to step into this expanded role. Many of you may have seen Time Magazine recently ranked Innodata #24 on the inaugural list of America's Top 500 Growth Leaders for 2026, recognizing companies that "Capture trends and stay ahead of time." That mindset, seeing what's next and acting fast is core to who we are now.

You are seeing the results of that today. We are deepening relationships with both existing and new Silicon Valley customers while delivering quick successes across the 6 investment areas Jack just outlined and increasing number of world's largest technology companies and enterprises are seeing the value we bring today.

Looking past 2026, over the medium and long term, we believe the work we do with frontier model builders will expand and will become more complex. The next generation of models won't just need more data. They'll need more smarter data, data from simulation labs, large-scale synthetic generation and [ RL ] gems that capture human judgment, context and values.

On top of this, the AI enterprise services market, which we are now successfully aligning to, will likely grow to be 10 or more times larger than the model builder market. We believe Innodata is purpose-built for this broad enterprise transition. Our work alongside frontier model builders give unique insights into how large models are trained, tuned, scaled and evaluated. And we are succeeding at packaging these insights into solutions that bring value to enterprises.

For example, we have just begun -- recently begun providing model safety and remediation solutions that leverage the working we have done hand in glove over the past year or so with engineering teams from leading AI hyperscalers.

Today, we are bringing those capabilities to one of the world's leading SaaS software companies and one of the world's leading generative AI chip designers. In short, I believe we are at the very beginning of the generational technology shift that Innodata is at the center of and poised to capitalize on. When I look at the competitive landscape, there are not even a handful of companies that have the capability to service $50 million, $100 million or larger order sizes in our space.

And that's the need for hyperscalers today and sovereign entities. Plus they don't have the proven ability to scale the organization, provide flawless data accuracy and be highly nimble to addressing the changing client needs in a very dynamic environment. What an amazing time to be alive when the world is going through a seismic change driven by AI and to be in such a privileged position to help lead a company that is a critical part of catalyzing the change.

I'll now turn the call over to Marissa. And after her remarks, we'll be available to take your questions.

Thank you, Rahul and Jack, and good afternoon, everyone. Revenue for Q3 2025 reached $62.6 million, up 20% year-over-year. Sequentially, revenue increased 7% from Q2's $58.4 million. Adjusted gross profit for Q3 2025 was $27.7 million, an increase of $4.8 million or 21% year-over-year with an adjusted gross margin of 44%. Adjusted EBITDA was $16.2 million or 26% of revenue, up 23% quarter-over-quarter, reflecting the strong operating leverage in our business.

Net income for Q3 2025 was $8.3 million compared to $17.4 million a year ago. The decrease was mainly due to the tax benefit arising from the utilization of net operating loss carryforward in Q3 2024. We ended the quarter with $73.9 million in cash, up from $60 million at the end of the prior quarter and $46.9 million at year-end 2024 and did not draw down on our $30 million Wells Fargo credit facility. As Jack mentioned, based on our current momentum, we reiterate our prior guidance of 45% or more year-over-year growth in 2025, and we anticipate potentially transformative growth in 2026. Thank you, everyone, for joining us today. Operator or Michael, please open the line for questions.

Now for Q&A. Our first question comes from Allen Klee with Maxim Group.

Great job on the quarter. Just I was adding up the -- you mentioned a bunch of potential contract wins and what they could represent. And if I -- the ones that you put dollars amount on added up to close to $100 million. But what I wasn't sure about is these -- some of these could be contracts over multiple years. Is there a sense of what amount of that could potentially be in 2026?

Allen, it's a great question. I think the contracts that we -- when we talk about annualized recurring revenue, those are generally the contracts that we think will kind of roll at the number that we state is a year's value from that. Other contracts that we talk about, we're going to try to do some ramping up of some of them in this quarter, but then that revenue would primarily be falling into next quarter -- excuse me, next year.

Okay. And then in terms of -- you mentioned that you're going to spend an extra -- I think you said $8.2 million in incremental SG&A. Could you just explain what -- that's over what time period? And the way to think of that is over what type of base?

So that would be year-over-year, and that would be incremental in 2025 versus 2024.

Got it. And then with your largest customer, I think you've mentioned now more than once of potential to expand the relationship, which could be very large. But any commentary on just the existing business of them? Is that -- should that be considered kind of stable?

So the relationship is strong and the business is stable. I think as you'll see, the business went up sequentially in the quarters. And as we discussed just a few minutes ago, we got a verbal on what's potentially a very large new program that would come into -- with that customer. We haven't really baked that into anything yet because we're not sure of what the ramp-up would be, but it's certainly very significant relative to next year.

Our next question comes from George Sutton with Craig-Hallum.

Quite an update, and congrats both Jack and Rahul for your expanded roles. Relative to the verbal comment, Jack, with your largest customer, I assume that would just run through an existing statement of work, so you could take that business on relatively quickly?

That's correct. I mean, mechanically, it would run through the existing master services agreement and probably be a new statement of work. But your point is correct that it will be very easy and seamless in order to onboard that new requirement.

So I was thrilled to hear about your federal market win. And it begs the question, and I think you addressed it with your GSA comment. But typically, you need to be part of a FedRAMP program to take on material business like this. Can you just walk through how you're doing this under this GSA process? Or what's different than a normal FedRAMP process?

Yes. So I think the point that we were making is that the timing for us starting this practice is ideal. The federal government has clearly communicated the strategic emphasis that they're putting on AI and AI enablement, both in the DoD, the IC and even civilian agencies.

So you have that -- on top of that, they're recognizing that the procurement and acquisition programs and processes are cumbersome, and they will impede the AI progress that they're intending to make. And therefore, they've issued executive orders. I think there may even be some new pronouncements expected to come out tomorrow on that subject. So when you take these 2 things in combination, the prioritization that the government is placing on AI, again, spanning the entirety of federal on the one hand and then on the liberalizations that they're making in terms of acquisition and procurement, it really couldn't be a better time for us to be in that market.

Got you. And then finally, Rahul, you made a very interesting comment that the services market could be 10x the model builder market. I wondered if you could just put a little bit more meat on that. And how much of that do you think you've started to see thus far?

Yes, George. So if you think about the enterprise market today and the frontier models, these models are now getting integrated into workflows that are transforming either for cost reduction, predominantly today for cost reduction. And soon, we're going to see transformative workflows that will drive new business models and revenue generating.

As we talked about, we are seeing for one large social media company, we were able to dramatically save them over $24 million worth of cost. So it's early stages. We are starting to get into the stage where we are starting to deploy Gen AI solutions into our customer base, and we hope to expand this service in the future.

Thank you very much. That appears to be our last question. I will now turn the conference over to Jack Abuhoff for any additional remarks.

Thank you. Yes, I guess Innodata is executing really from a position of strength. We had another record-breaking quarter. Revenue is at an all-time high. We see profitability growing and the results exceeded our analyst expectations. Looking out ahead to 2026, we see the potential for continued transformative growth powered by deepening relationships among the Mag 7 and other Silicon Valley leaders.

And we see that growth coming from 2 sources. First, the continued expansion we're driving with existing and new customers. And then secondly, the strong returns we're beginning to see from our recent investments. Today, I talked about 6 specific investment areas. And across each of them, across the board, we're showing what happens when we do exactly what Time Magazine recognizes for, seeing what's next and acting fast.

So to recap quickly some of these early wins. First, $68 million in new pretraining data wins, $42 million that signed, $26 million that we believe gets signed very soon. The $25 million win with a new strategic federal customer that we expect to name soon, and we believe this is potentially the first of many projects with them, an additional expansion with our largest customer based on verbal confirmation, a $6.5 million verbal confirmation of the deal win with another big tech customer and new partnerships emerging with key AI and sovereign AI players, which we expect to be announcing in 2026. So thank you all for joining us today. We couldn't be more excited about what lies ahead. Thank you.

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