2026.09.22 | Substack | Data | INOD Up ~20%: What Happened

INOD is ripping today.

Hunterbrook published an investigation identifying Meta as Innodata’s largest customer. Former employee interviews and hiring records support that identification. Hunterbrook also argues that an Innodata agent program could be connected to Muse, Meta’s personal AI agent that the market is super hyped on right now.

Inod disclosed the agent program on August 6. In its second-quarter results, management said it had won a significant new program with its largest customer covering the “personalization of long-horizon agents,” and that the work was scaling. It also announced a separate program building reinforcement-learning environments for computer-use agents. Hunterbrook’s contribution today is evidence pointing to Meta as that largest customer, plus the possible connection to Muse. The new Big Tech customer Innodata announced earlier this year is a separate account.

The Muse link makes sense to me at a product level. Meta describes Muse as an agent that remembers individual preferences, works across apps and a browser, and keeps going on longer tasks. An agent doing those things needs training examples, human judgment, and repeated testing when it gets details wrong. Those are areas where Innodata already gets paid. Still they obviosuly have yet to confirm Innodata’s involvement in Muse, and Hunterbrook’s affiliated fund disclosed a long INOD position when it published the story. So, take all that into consideration.

When I first wrote about INOD in May, the appeal was that customers were already paying for data engineering, model evaluation, expert feedback, and agent workflows. In my August portfolio update, I focused on the concentration risk where the largest customer fell from 56% of revenue in Q1 to 37% in Q2. In dollars, that was roughly $50.5 million down to $34.1 million. Management still expected full-year growth from that customer, but we needed evidence the relationship would keep expanding, especially after Meta invested in Scale AI.

Today’s reporting gives me more confidence in the agent side of the thesis. If Inod is helping Meta improve a flagship consumer agent, there could be repeated demand for training data and evaluation as Muse evolves. The size, duration, and profitability of this specific program remain undisclosed. The robotics data opportunity I discussed in the original INOD piece remains a separate, earlier-stage part of the thesis.

So let’s quickly visit some of my current financial modeling for INOD and what my 12-month cases could look like.

Q2 delivered $92.1 million in revenue and $14.4 million in GAAP net income, or $0.41 per diluted share. Management’s 40%+ growth guide implies at least $352 million of 2026 revenue. I am still using roughly $370 million as an illustrative 2026 starting point.

• Bear, ~$45: I have revenue reaching roughly $420 million in 2027 and $470 million in 2028 as growth falls into the low teens. At a 14% net margin and roughly 38 million diluted shares, that gets me to around $1.70 of 2028 EPS. At roughly 26× earnings, I get to about $45.
• Base, ~$110: I have revenue around $480 million in 2027 and $600 million in 2028, representing roughly 30% growth next year followed by another 25% in 2028. At an 18.5% net margin and roughly 38 million diluted shares, that gets me close to $2.90 of 2028 EPS. At roughly 37× earnings, I get to around $110.
• Bull, ~$160: I have revenue reaching roughly $560 million in 2027 and $730 million in 2028 as the newer Big Tech customer and agent programs scale together. At a 21% net margin and roughly 39 million diluted shares, that gets me close to $3.90 of 2028 EPS. At roughly 40× earnings, I get to around $160.

I have assigned zero incremental revenue specifically to the possible Muse connection because the value of the program remains undisclosed.

I will not be adding to my position at this time. The way the market has been treating it the last couple months has been rough. While encouraging today, I don’t know how durable this move will be. I will most likely hold my position and then decide next earnings whether to maintain or not. There are so many exciting Physical AI stocks out there that I am uncovering that I might want to own more. We will see. But for now, holding.

Financial Disclaimer: This article is for informational and educational purposes only and should not be considered investment advice, a recommendation, or a solicitation to buy or sell any security. The views expressed are my own and are based on information I believe to be reliable at the time of publication, but accuracy or completeness cannot be guaranteed.

Any financial models, valuation ranges, price targets, revenue estimates, earnings estimates, or forward-looking scenarios discussed are based on assumptions that may prove incorrect. Actual results can differ materially. References to bear, base, or bull cases represent illustrative scenarios rather than predictions of future stock prices or business results.

I may hold positions in securities discussed and may buy or sell those securities at any time without providing an update. My personal portfolio decisions, risk tolerance, time horizon, and cost basis may differ materially from yours.

Investing involves risk, including the potential loss of principal. Conduct your own research and consider your individual financial circumstances, objectives, and risk tolerance before making any investment decision. Nothing published here should be relied upon as a substitute for advice from a qualified financial, tax, or legal professional.