Innodata Inc. (INOD) Earnings Call Transcript & Summary

August 6, 2026

NASDAQ US Industrials Professional Services earnings 37 min

Earnings Call Speaker Segments

Amy Agress

executive
#1

Thanks so much. Hello, everyone, and good afternoon. We appreciate you being with us today. Joining us on this call are Jack Abuhoff, who serves as Chairman and CEO of Innodata, along with Rahul Singhal, our President and Chief Revenue Officer, and Jayant Chauhan, our Chief Financial Officer. Also participating today are Marissa Espineli, Chief Accounting Officer, and Aneesh Pendharkar, Senior Vice President, Finance and Corporate Development. Jack and Rahul will speak first and share their views on the business, after which Jayant will walk through a review of our Q2 results. Following that, we'll open things up for analyst questions. Before we begin, I want to remind all of you that on this call we will be making forward-looking statements, meaning predictions, projections, or other statements concerning future events. Such statements rest on current expectations, assumptions, and estimates, and they carry risks and uncertainties. Actual outcomes may differ materially from what those forward-looking statements contemplate. The factors that might cause results to diverge materially are described in today's earnings press release, in the Risk Factors section of our Form 10-K, our Forms 10-Q, and other reports and filings we make with the Securities and Exchange Commission. We assume no duty to update forward-looking information; additionally, on this call we may talk about certain non-GAAP financial measures. In our earnings release filed with the SEC today, and in our other SEC filings, all of which are available on our website, you'll find additional disclosures about these non-GAAP financial measures, including reconciliations of them to comparable GAAP measures. Thank you. I'll now hand the call over to Jack.

Jack Abuhoff

executive
#2

Thanks, Amy. Hello, everyone. Innodata had yet another record-breaking quarter in Q2. We hit new peaks in revenue, adjusted gross profit, adjusted EBITDA, and cash, and we beat analyst consensus across every key metric. Revenue came in at $92.1 million, which is 58% higher than the same period last year, surpassing analyst consensus by roughly $5.8 million, or 7%, and marking our 12th straight quarter of year-over-year growth. To give you a sense of scale, in Q2, just like Q1, the revenue we generated in a single quarter was greater than what we brought in during an entire year only three years ago. On the adjusted gross margin side, we landed at 49%, which is a two-point improvement from the prior quarter and nine points better than the 40% target we've publicly stated. Adjusted EBITDA reached $25.4 million, a 92% increase year-over-year, beating analyst consensus by about $8.5 million, or 50%. On a fully diluted basis, earnings came to $0.41 per share — that's almost twice what analysts had expected, which was $0.21 per share. Once again this quarter, we achieved growth, expanded margins, and generated cash all at the same time, while still putting money into innovation that turns into revenue in quarters rather than years. That's our business model doing exactly what it was built to do. During the previous quarter, we advised you that our biggest customer would account for a reduced share of overall revenue. In Q2, that top customer made up 37% of revenue, compared with 56% of revenue in Q1, while the major tech company we mentioned last quarter grew from 17% of revenue to 34% of revenue, making it our number two customer. During Q2, revenue from our biggest client came in lower than it did in Q1, which reflected a shift in the quarter related to program structure and service mix. Even so, for the entire year we still anticipate year-over-year growth from that customer. In addition, we brought on a significant new client this quarter — one of the frontier labs that is scaling up most rapidly. What this means is that our foundation keeps widening, both in terms of clients and the programs we run for them. Ahead of discussing guidance, I'd like to make a key announcement regarding leadership at Innodata. Starting September 30, Rahul Singhal takes on the roles of President and Chief Executive Officer at Innodata and will also sit on our board. At that point, I'll move into the Executive Chairman position. This transition was planned well in advance and comes at a moment when the company is strong — and for me, there's a personal dimension to it as well. A lot of you have gotten to know Rahul through these earnings calls, at investor conferences, and via the initiatives he's spearheaded across the last few years, during which he served as a principal architect of Innodata's shift into a strategic partner for the top AI developers globally. Our customers are familiar to him, our technology is familiar to him, and our team is familiar to him. When it comes to the strategy driving the results you've witnessed quarter after quarter, Rahul has played a central part in every piece of it. When it came to finding the right person to lead, neither the board nor I needed to search far. Through years of shouldering ever-growing duties and coming through each time, Rahul has demonstrated he deserves this position. That's the way we operate here: we sharpen our skills and advance talent from within. In my capacity as Executive Chairman, I'll stay closely involved, concentrating on working alongside Rahul to develop capabilities powered by our research team. Bringing those capabilities to the federal government and the enterprise market is, from my perspective, where I can add the most toward creating meaningful shareholder value. Given that I'm one of the company's largest shareholders, that's exactly where I want my attention to be. Our engagement with the Mag-7 along with top AI labs is firmly on a trajectory toward bigger achievements and broader diversification. The strategies we've built for enterprise AI and the federal market, resting on the differentiated technology we created for the frontier labs, open up chances to potentially generate high-quality recurring revenue that translates into meaningful value creation. Our ambition is for Innodata to become a generational company. Holding that same aspiration, we were happy to recently announce that Jayant Chauhan has come aboard Innodata as Chief Financial Officer. Jayant's strengths complete an already capable finance team, with Marissa Espineli moving into the Chief Accounting Officer position. Beyond the standard CFO duties, Jayant is going to handle strategy around how we deploy capital and tap capital markets, relationships with customers, acquisitions that could accelerate what we're building, and talking with investors — and at the same time he'll be building out the finance operation for the larger company we're turning into. Before I turn things over to Rahul, I want to address guidance: our outlook stays the same, calling for revenue to grow at least 40% compared with last year. There are a number of sizable opportunities in our pipeline, some with existing customers and some with new ones, that we believe we'll close. Until now, none of those have been included in what we projected. Being cautious, we won't include those deals in our projections until we're certain they're fully secured and can predict when the revenue will be recorded. At this point, I'll pass things over to Rahul, who will cover the market, what our approach looks like, and the execution milestones that, in our view, show this strategy is succeeding.

Rahul Singhal

executive
#3

Thanks, Jack. Good afternoon to all of you. Let me start with something personal. The trust that Jack and the board have shown in me means a great deal, and my plan is to pay that back through results. Innodata has incredible momentum, an incredible team, and an incredible opportunity ahead of it. My intention is to build on all three. One of the biggest things to happen over the last 18 months is how much more central research and innovation have become here at Innodata; saying that research has turned into a growth engine would not be an exaggeration, nor would saying it is how we increasingly set ourselves apart, grow existing partnerships, and create new customer relationships. Increasingly, our expansion is driven by R&D and innovation that touches every stage of the model training process — from pre-training through post-training, plus evaluation and benchmarking. The breakthroughs we create become intellectual property and set us apart, which in turn fuels demand. Several quarters ago, we explained our approach to measuring frontier model performance, identifying where those models fell short, building remediation datasets to address those gaps, and then proving those datasets worked by training smaller models whose architectures mirrored the large ones. Our current efforts extend far past that. Let me go over a handful of illustrations from what we're doing right now, because the projects themselves are really interesting and they also reveal the direction we intend to steer Innodata in across the next few years. Thanks to our research investments, we got in early on agentic reinforcement learning, which counts as one of the biggest frontiers in AI development today. Together with a major lab, we're operating a sizable new initiative focused on personalizing long-horizon agents, and it's expanding as we speak; on top of that, we've joined a second initiative that deals with reinforcement learning settings for desktop computer use and associated tasks. With enterprise customers, what we're seeing is this: they're fast to create AI agents, but getting those agents into production with real confidence is where things get hard. Our thinking is that if we combine a dependable observability platform with reinforcement learning gyms built on our novel architecture, we can claim the role of AI deployment assurance layer. That's a massive opportunity we see emerging, and Jack touched on exactly this a few minutes back. Through the quarter, we pushed further on shipping these capabilities with a major tech client, and kicked off delivery with a second one. On top of that, this innovation has sparked live conversations in insurance and banking that we anticipate converting into pilots. People creating cutting-edge models are now paying very close attention to agentic evaluation that spans long horizons and changes dynamically. This quarter we released two benchmarks available to the public; one measures model performance across interactions involving multiple turns, lengthy context, and several modalities. Expert-written prompts, rubric-based requirements, and LLM judges come together in a benchmark, all set up to evaluate frontier models. Each was designed to reveal the kinds of breakdowns that ordinary leaderboards miss. Examples include losing track of grounding and forgetting instructions. Those are precisely the failure modes that frontier labs are working hard to fix. Every benchmark engagement produces a data strategy recommendation and positions us to provide scaled data generation that enhances the model; during the quarter, we additionally broadened what we can do in generating training data that extends reasoning abilities of state-of-the-art models, delivering across 5 frontier labs and 5 domains. As AI shifts from digital tasks toward embodied intelligence, we are constructing the necessary data and measurement layer. In the quarter we just finished, we entered into two research agreements with a major university, and we also committed to a motion capture lab. That facility is expected to be up and running in a few months, and it will be capable of collecting data at sub-millimeter precision, which we'll use to train robots as well as physical AI foundation models. On the data side, our collection efforts have shifted away from standalone pilots toward defining multimodal programs at enterprise scale. Those include a multilingual speech initiative spanning 7 languages, plus an egocentric program totaling approximately 2 million hours, which we're hoping to win on the strength of how the pilots performed. Improving AI — and ensuring it's safe and trustworthy — depends centrally on data, data engineering, and data science. Because those sit at the core, our research is able to produce capabilities that reach into a wide range of areas. Advances in data engineering are capable of solving major AI problems, even within areas you wouldn't expect to see us in. On our Q4 call, we mentioned an example of this: an AI model we built to detect drones and other small objects, which beats earlier state-of-the-art benchmarks by 6.45%. In a discipline where gains are usually counted in tiny fractions of a percentage point, 6.45% counts as a meaningful step forward. Right now we're focused on showing that government customers can do this. One more illustration: earlier this week, we told you about the initial phase of the AI Cyber Training Suite, as we're naming it. We've got twelve datasets plus evaluation systems designed to teach AI coding agents how to produce secure code and fix vulnerabilities inside a company's current software. In our testing of top open weight models on repairing verified flaws, after just one round of fine-tuning using only a subset of our data, the repair rate increased by over twofold. Since a rising portion of the world's code is now authored by AI, not being able to trust that code unless a security team reviews every output is a genuine obstacle for enterprise adoption. Our suite, we think, could potentially eliminate that obstacle. Whether it's Frontier Labs, federal, or the enterprise, we see the same pattern. Innovation and R&D keep generating distinctive strengths that help us secure programs and build lasting ties with customers. The possibilities in front of us have us extremely enthusiastic. Over to you, Jack.

Jack Abuhoff

executive
#4

Thank you, Rahul. I'd like to spend a few minutes as well tying what happened this quarter to the structural economics of our business, and then use some time to discuss the wider market dynamics. Let's start with operating leverage. Our revenue was up 58% compared to a year ago, while adjusted EBITDA rose 92% — about 1.6 times quicker. Building that capability from scratch costs far more than what the next program costs at the margin. Now, margin quality. At 49%, adjusted gross margin sits nine points higher than the target we have publicly stated. Mix is what drives this expansion, and it gets further support from high-value pre-training programs plus off-the-shelf datasets, in which we keep the IP and earn money on one asset across many customers. This is precisely the software leveraged economics we have intentionally been working toward. Now let me move to the wider market dynamics. People are debating whether AI CapEx has peaked, whether competition will turn models into commodities, and what the industry should take from the recent security incidents. All of this debating is happening alongside truly extraordinary numbers. This year, hyperscaler capital spending is guided to approximately $700 billion, almost twice what it was last year, and estimates were revised higher all through the year. Our view is that every one of these debates ends up favoring the layer we provide: data, evaluation, and assurance. Let me walk through why. Should CapEx face pressure, monetization pressure grows, and monetization depends on deployment, fine-tuning, and assurance — which is our business. Should inference become commoditized, two consequences follow. Labs will engineer for differentiation specific to use cases, and that demands specialized data; meanwhile, AI reaches the enterprise more easily, and that demands greater assurance, not reduced assurance. Once more, that is our business. Should security incidents keep multiplying, they demonstrate the necessity of precisely the engineering we unveiled this week. And once again, that is our business. No matter which way the market goes, we believe it heads toward the work we perform. At this point I'll hand the call to Jayant, our new Chief Financial Officer, who will take you through the financials.

Jayant Chauhan

executive
#5

Thanks, Jack, and good afternoon to all of you. This is Jayant Chauhan, Innodata's Chief Financial Officer — and as you're aware, this marks my first earnings call after joining the company in July. Meredith has moved into the Chief Accounting Officer position, and I appreciate her collaboration in helping me ramp up quickly. Over the last several weeks, I've been learning about the business and getting to meet our teams across the U.S. and internationally. What I've discovered has me energized. I'm eager to become acquainted with many of you during this call and afterward. With that, let me take you through our Q2 2026 results. Revenue in Q2 2026 came in at $92.1 million, which is 58% higher year-over-year and 2% higher sequentially. That makes 12 straight quarters of year-over-year growth. It beat analyst consensus by $5.8 million, or 7%. Adjusted gross profit reached $45.4 million, which works out to an adjusted gross margin of 49%; that figure was two percentage points above Q1 and nine percentage points beyond the 40% target we've communicated externally. The gain came from a mix shift toward programs with higher margins. Adjusted EBITDA landed at $25.4 million, or 27.5% of revenue, which is up 92% year-over-year; that surpassed the analyst consensus figure of $16.8 million by roughly 50%. Net income for the quarter totaled $14.4 million — twice the $7.2 million reported in Q2 of last year. Fully diluted earnings per share came to $0.41, beating the consensus estimate of $0.21 by about 95%. Our effective tax rate this quarter was roughly 18%, versus our long-term target range of 23%-25%. That lower rate resulted from tax benefits recognized during the quarter. Moving to the balance sheet, we closed the quarter with $250.4 million in cash and short-term investments. If you exclude customer prepayments, which are pass-through in nature, our cash and short-term investments stood at approximately $134 million, an increase of $37 million sequentially. We continue to have nothing drawn on our Wells Fargo credit facility. Finally, after the market closes today, we'll file a prospectus supplement to establish an at-the-market equity program, with Goldman Sachs serving as lead agent alongside a wider syndicate. This program gives us an efficient supplemental capital markets tool to deploy selectively and opportunistically. Our balance sheet remains strong, with cash and short-term investments of approximately $134 million net of customer prepayments and no debt outstanding as of the end of Q2. The program maintains optionality to back future growth initiatives, possible strategic opportunities, and ongoing balance sheet strength as we scale. With that, let me wrap up. This was a good quarter for me to step into, and I'm eager to build on the growth and financial discipline this team has already put in place; with that, I'll hand things back to the operator. Operator, we're ready for questions.

George Sutton

analyst
#6

Thanks. To start, congratulations to Rahul Singhal, and welcome aboard to Jayant Chauhan. Jack Abuhoff, I still plan to keep pestering you with questions on a regular basis. I'd like to explore the items that fall outside your guidance. You brought up certain opportunities that aren't fully booked at this point, which means they're excluded from guidance. Could you paint a broader picture of what a few of those opportunities might look like? And will you provide updates on a more frequent basis, maybe beyond the quarterly announcements?

Jack Abuhoff

executive
#7

Of course, George Sutton. Thanks. It goes without saying, I'm happy to take your questions whenever you want to send them my way; the quarter left us really thrilled. I believe plenty of proof points got established during the quarter, and a few of the lessons we're picking up as we move ahead matter to us just as much as the financial signal you're seeing right now. The innovation we're achieving and generating is charting our course. It's demonstrating that reliability in agentic enterprise AI is something that can be engineered; it's demonstrating the types of innovations we're able to build and the impact we can have by working at the data engineering layer in very random things, drone detection, cybersecurity. Those are only two illustrations. When I consider the collection of opportunities before us, they span everything, and now I'm answering your question regarding the ones that are significant, a few quite large ones that aren't in our guidance today. They stretch across our capabilities. Some are on the government side and the enterprise side. Some are on the frontier model side. Many of the capabilities we're now showing in agentic AI, both deployment and training, figure prominently in our pipeline. We're excited about it, from a methodological standpoint, we keep the discipline of counting our chickens only once they're hatched; we look forward to sharing more as we move through the second half of the year; we think it's going to be exciting.

George Sutton

analyst
#8

The safety breaches we're beginning to observe involving AI are clearly troubling, and they appear to have opened up a really promising new avenue for you. I was wondering whether you could just talk us through that, and of course if you could tie in the press release from a couple of days ago along with some of your capabilities. What does that translate into when it comes to the opportunity available to you?

Jack Abuhoff

executive
#9

Yeah, that's a good question. I think when we examine the challenges the enterprise is facing, they're eager to adopt agentic AI, but the question is whether they can place their trust in it. And what might be the factors that prevent them from being able to trust it? Of course—when people come across stories about models breaking out of their sandboxes or acquiring top-tier cyber skills once they get past containment, that kind of thing, it turns into genuine worry. Part of why that worry is there: many of the leading-edge models able to cause these kinds of cybersecurity problems were trained on data that included code which hadn’t been patched yet. It’s really interesting. If you’re able to figure out what got mixed into their training data, and then create an agent capable of spotting those code anomalies—spotting the code that got added in, even if fixes came later— If you can get that AI to apply what it’s learned to unfamiliar, previously unseen dangers and also spot risks already present in software, then you end up with a strong technology stack that lets the enterprise bring in AI with less risk. We’ve been having some intriguing conversations along those lines. To us, this is one more illustration of the sort of innovation we keep getting better at producing.

George Sutton

analyst
#10

Understood. One more item comes to mind. Clearly, we're observing increased federal government scrutiny of models prior to launch, much of that happening via red teaming. Could you provide some insight into how your company fits within the wider federal landscape?

Jack Abuhoff

executive
#11

Certainly. A few points come to mind here. I'd say we're engaged in numerous fascinating conversations with government participants regarding potential partnerships and areas of collaboration. Conversations with agencies are also underway. Being able to appear in the Tradewinds marketplace as an approved solution for various purposes represents an enormous chance and an enormous edge that we currently possess. Regarding what the federal government's connection to AI will ultimately look like, two factors stand out. To begin with, they are significantly speeding up their capacity to acquire AI solutions and secure the finest options. Secondly, we're noticing that frontier model firms are proactively extending invitations to the government to assist in regulating the agency. Considering what the ultimate demand will be for items such as benchmarks and evaluations and red teaming, we put out two benchmarks during this quarter that we believe are quite innovative and quite valuable. Providing those sorts of items and evaluation services on the government's behalf is a prospect that we are monitoring.

George Sutton

analyst
#12

Super. Thanks for answering the questions.

Jack Abuhoff

executive
#13

Thank you.

Allen Klee

analyst
#14

Yes, hi there. One of the bright spots you brought up for this quarter involved a greater proportion of engagements with better margins. My question is whether we ought to view that as a directional shift, or whether it was simply how things landed in this period and could go back to the levels seen historically.

Jack Abuhoff

executive
#15

Yeah. That's a great question. Here's how I'll address it. My view is that it's both. Let me clarify what I mean. There is work we pursue where gross margin is lower than what you're observing right now. A number of those engagements may be sizable. We would plan to accept them. Should we secure them, the cash flow they generate, in our expectation, would be pretty attractive. Would that cause gross margin, on a weighted basis, to come down to some degree? Yes, it would. Conversely, looking at it strategically, the initiatives we're pursuing, the areas where we're innovating, are probably going to carry better revenue quality; when we assess revenue quality, or when we think about it, we treat it as depending on both gross margin and how recurring that revenue is. Strategically, I believe the direction over time is up. On a quarterly basis, though, product mix will be the determining factor.

Allen Klee

analyst
#16

That's useful, thank you. You also mentioned that you're increasingly relying on pre-existing datasets for training purposes. Could you elaborate a bit — is the data owned by you, or do you have permission to use it repeatedly? If that's not the case, how are you obtaining the data?

Jack Abuhoff

executive
#17

Of course. Up to this point, the ready-made datasets — and I'll explain shortly why I phrased it that way — are mostly ones that we create ourselves. We build them in response to weaknesses we uncover through our benchmarking process. Whenever we spot a shortcoming, or when we recognize a capability that the frontier models are aiming to develop, we're able to construct a dataset that assists them in reaching that goal. Instead of sitting around waiting for them to ask us for it, we go ahead and create that dataset. We hold onto or keep the IT linked to that dataset, and we let them leverage those datasets to train their models. It's a win-win situation, correct? It benefits our customers, and it benefits us too — that's among the factors driving our stronger margin profiles. There are also occasions when we act on behalf of another party who possesses a dataset, and we represent them. We might perform some engineering work on that data. We'll set it up so it's prepared for models to undergo training on it; we'll also extend an invitation to our customer partners to make use of that data. What you're mostly observing at present is data that we've determined how to put together to address specific model requirements and frontier model capabilities.

Allen Klee

analyst
#18

Thanks. For my final question: under the scenario deemed most probable, is there any basis to expect that revenue could fall sequentially during Q3 or Q4?

Jack Abuhoff

executive
#19

Given the limits inherent in how our business operates, that's absolutely a possibility, and should it happen, I'm not sure it would bother me much. My main concern is the direction we're steering this company and its future trajectory, rather than how any single quarter performs. The types of breakthroughs we're bringing to market at present, the performance history we're building, the fresh clients we're bringing aboard — I believe these things will keep paying off for us as time goes on, and I believe we'll keep expanding this business quite dramatically across the coming several years. Suppose we landed one enormous one-off engagement delivered across 2 quarters, leaving a gap in demand following Q3 — would I view that as a setback? Absolutely not. What I would view as a setback is failing to preserve the relevance we currently have with our clients, or failing to spot enormous market opportunities that I'm confident we can pursue across the coming several years.

Allen Klee

analyst
#20

Thanks a lot. I appreciate that.

Jack Abuhoff

executive
#21

Thank you as well. In closing, the second quarter of 2026 marked yet another record period for Innodata, and we beat expectations on every measure. Revenue rose 58%, adjusted gross margin came in at 49%, adjusted EBITDA climbed 92%, and we produced substantial cash. That also made it our 12th straight quarter of growth compared with the same period a year earlier. Diversification is clearly taking shape in real terms. Our biggest customer's share of revenue fell to 37% even as the whole business expanded. A customer that brought in virtually nothing 12 months ago now ranks as our number two customer. We announced a planned leadership transition; Rahul will step into the roles of President and CEO on September 30th. I will take on the role of Executive Chairman. My focus will be on developing long-term differentiating capabilities spanning our enterprise and federal markets. In the meantime, Jayant Chauhan has come aboard as CFO, which makes our financial leadership even stronger and frees me to pursue some of the things I want to do. As someone who holds one of the largest stakes in the company, I see this as an outstanding route toward creating very substantial shareholder value. As we have discussed, our expansion is increasingly driven by research and innovation. That ranges from novel benchmarks and reinforcement learning environments to capabilities in agentic deployment assurance and physical AI. I believe we are only at the very beginning of much of this; we are thrilled about what lies ahead; we have strong confidence that 2026 can be a tremendous year for Innodata, and I thank you all for staying on this journey with us.

Operator

operator
#22

Everyone, that wraps up our call for today. Thanks to each of you for taking part. You're free to leave the line now.

Read the full transcript via the API

You're viewing the first half of this call. Get the complete Innodata Inc. transcript — plus 131,000+ transcripts from 7,000+ companies, speaker segments, AI summaries and full-text search — through the EarningsCalls.dev API.

Get the API View API docs →

This call discussed

For developers and AI pipelines

Programmatic access to Innodata Inc. earnings transcripts and 131,000+ others is available through the EarningsCalls.dev REST API. Plans from $39.99/month — full transcripts, speaker segments, full-text search, and the recently-added /api/v1/transcripts/recent polling endpoint for ETL pipelines.

More tools for stock research

OptionsBell: an unusual options activity scanner that tracks options flow across all US stocks and emails you when large trades appear in Innodata Inc. or any ticker on your watchlist.

StockMarketScan: a stock screener with alerts: 24 pre-built screens refreshed every 5 minutes, watchlists and email notifications. Both offer API and MCP access and fair monthly pricing.