AI data growth now has a second anchor
- Innodata is a picks-and-shovels supplier for generative AI, making complex data used to train and improve large language models.
- Q1 2026 revenue was $90.1 million, up 54% year over year, with adjusted gross margin of 47%.
- A new Big Tech engagement could bring $51 million of 2026 revenue and become the second-largest customer.
- Customer concentration is still high: the largest customer was 56% of Q1 2026 revenue, and another customer was 17%.
- The stock now asks investors to pay up for fast growth, so any slowdown in AI spending could hurt the case.
Diversification finally shows up
The bull case got stronger in Q1 2026. Innodata posted $90.1 million of revenue, up 54% year over year, and management raised its full-year growth view to approximately 40% or more. Adjusted gross margin was 47%, above the company’s 40% public target.
The bigger change is customer mix. Management announced a new set of engagements with a leading Big Tech company that could generate $51 million of revenue in 2026. Twelve months earlier, revenue from that customer was zero. If the ramp holds, it becomes a real second anchor, not just a promise of future diversification.
The bear case is smaller, but not gone. The largest customer still made up 56% of Q1 2026 revenue, and another customer made up 17%. That means a few large AI buyers still have major power over Innodata’s results.
The price also matters. Growth and margins look strong, but the market already expects a lot. For the thesis to keep working, Q2 and Q3 filings need to show that the new customer is ramping, the top customer share keeps falling, and margins stay above the company target.
Selling data to the AI buildout
Innodata makes money by creating high-quality data sets and services that help companies train, test, and tune AI models. A large language model, or LLM, learns patterns from data so it can write, reason, code, or answer questions. Better data can make the model safer and more useful.
The core customer base is large technology companies building foundation models. These customers need custom data for supervised fine-tuning, reasoning, pretraining, and special use cases. Innodata’s pitch is simple: if AI labs are racing to build better models, they need suppliers that can deliver accurate data at scale.
The second layer is enterprise AI work. This includes fine-tuning models and building RAG applications, which means retrieval augmented generation. In plain English, RAG lets an AI system look up trusted company information before answering.
The top layer is Innodata’s own software platforms. Agility serves public relations teams, and Synodex extracts and processes medical records. The problem for investors is that Innodata now reports as one segment, so the public filings no longer show how these platforms perform on their own.
What Innodata sells
AI Data Services
This is the core business. Innodata builds custom data sets for generative AI models, mainly for large technology companies.
Advanced LLM Training Data
The company engineers specialized data for long-context reasoning and other hard model-training tasks. This work is tied to the race to make AI models reason better.
Agentic AI Evaluation Data
Innodata is building data and testing tools for autonomous AI agents. These tools help check whether agents can handle real-world tasks and resist bad prompts or edge cases.
Physical AI Data
This work supports robotics and machines that act in the physical world. It includes data about first-person views and what actions objects allow.
Agility Platform
Agility is software for public relations teams. Its PR CoPilot feature adds generative AI to media monitoring and PR workflows.
Synodex Platform
Synodex extracts and structures medical record data. It has been used in life insurance underwriting and is expanding toward clinical use cases for hospitals and doctors.
One segment, big customer exposure
As of Q1 2026, Innodata reports one business segment, so former DDS, Synodex, and Agility financial splits are no longer disclosed. Finvest shows Q1 2026 revenue exposure by customer group instead: largest customer 56%, another customer 17%, and all others 27%.
What could break the story
AI buyer concentration
High impact · Medium oddsThe top customer was still 56% of Q1 2026 revenue. Another customer was 17%. If one major AI lab cuts spend, delays projects, or changes vendors, Innodata could lose a large slice of revenue fast.
New customer ramp misses
High impact · Medium oddsThe new Big Tech engagement is expected to generate $51 million of 2026 revenue. That is a major part of the diversification story. If the work is delayed, smaller than expected, or does not renew, the bull case weakens.
Margins fall back
Medium impact · Medium oddsQ1 adjusted gross margin was 47%, above the company’s 40% public target. That level may not hold if the mix shifts toward lower-margin services or if hiring and delivery costs rise. It is also unclear how much margin came from less recurring dataset sales.
Less reporting detail
Medium impact · High oddsIn Q1 2026, Innodata moved to one reportable segment. That matches how management says it runs the business, but it lowers outside visibility. Investors can no longer see standalone growth or margins for Agility and Synodex.
Legal and regulatory overhang
Medium impact · Medium oddsThe company remains subject to a putative securities class action filed in February 2024. It has also previously disclosed SEC and DOJ investigations. An adverse result could cost cash, distract management, or hurt investor trust.
Future dilution
Medium impact · Low oddsInnodata maintains a $50 million universal shelf registration. That gives it flexibility to raise equity or debt. If the company uses equity while the share price is weak, existing shareholders could be diluted.