Innodata Inc. — AI supply-chain exposure

Innodata is a data-engineering company that prepares the material models learn from — curating, labelling and structuring the supervised fine-tuning and evaluation datasets that hyperscalers and AI labs use to shape and test their systems. It sits in the chain as a supplier into the training and evaluation layer, and its revenue has scaled as those customers moved from raw web scraping toward carefully supervised, human-checked data.

The structural point is that a model is only as good as the data curated for it, and as labs push into fine-tuning, safety and evaluation, high-quality human-in-the-loop data becomes a bottleneck rather than an afterthought. That gives a specialist supplier real leverage, since expert annotation is hard to fully automate. The exposure cuts the other way too: Innodata's fortunes track a concentrated set of large customers, so its trajectory is tied closely to how much those labs and hyperscalers spend on data.