INOD (Innodata) prepares training materials for artificial intelligence models, evaluates their answers, and helps correct their mistakes. In simple terms, it develops textbooks and exercises for AI—and then grades its work. The real barrier is not basic data labeling, but the ability to organize specialized professionals and consistently deliver complex projects on time and to uniform quality standards.
In the second quarter of 2026, Innodata generated approximately $92.14 million in revenue, up 58% year over year, and about $14.41 million in net income, representing a net margin of nearly 16%. These results provide initial evidence that customer demand is real and that the business can be profitable. As AI models take on more complex tasks, customers may increase their spending on high-quality training data and evaluation services, allowing Innodata to expand individual projects into longer-term relationships.
The risks are equally clear. Its two largest customers account for approximately 71% of revenue, creating significant customer concentration. Customer advances also inflate reported cash and should not be treated entirely as funds freely available to shareholders.
As of September 13, 2026, the latest available closing price was $53.21. I estimate fair value at approximately $51.60, meaning the shares were trading about 3% above that level. Innodata is therefore an AI services company worth following, but at its current price, it appears close to fairly valued and does not yet offer a meaningful margin of safety. The key metrics to monitor are customer diversification, profit margins, and underlying free cash flow.