So Innodata just reported solid numbers for Q4 2025, but the market basically yawned and sold off instead. Kind of interesting how that plays out sometimes in tech stocks.

Let's break down what actually happened. Revenue jumped 22% year-over-year to $72.4 million in the quarter, and full-year 2025 hit $251.7 million—that's a 48% jump annually. Adjusted EBITDA came in at $15.7 million for Q4 and $57.9 million for the whole year. On paper, this looks pretty solid, especially considering the AI boom they're riding. But after earnings dropped on Feb 26, shares tanked about 3.6% in after-hours trading and kept sliding the next day. By early March, the stock had wiped out nearly $117 million in market cap. Over the past month leading up to that report, INOD was down 7.2% while the broader Engineering-R&D Services industry actually climbed 4.2%. The S&P 500 was down about 2% in the same stretch.

Here's the thing though—the fundamentals driving this company are actually pretty compelling. Innodata's basically positioned at the center of the generative AI wave. They provide the data infrastructure that trains these large language models, handles evaluation work, and optimizes performance. Hyperscalers and enterprise AI teams are demanding this stuff like never before. Management's guiding for around 35% or higher revenue growth in 2026, which suggests they see a real new opportunity emerging across their customer base.

What's caught my attention is how they're expanding beyond just data annotation. They're now working on agentic AI systems, robotics datasets, and adversarial testing frameworks for AI safety. They recently showed a 6.45% improvement in drone detection models compared to prior benchmarks. That's the kind of thing that signals a company moving up the value chain rather than just commoditizing their services.

But here's where the market got nervous. Net income actually declined to $8.8 million from $10.3 million year-over-year, which spooked some investors. Gross margins are expected to compress into the 35-40% range early in 2026 as they ramp new programs, though management expects recovery to 40% or higher as volume scales. That margin volatility is real and worth tracking.

There's also the customer concentration angle. Historically, Innodata's been pretty dependent on a handful of large AI clients. They're pushing hard on diversification—adding more hyperscalers, sovereign AI initiatives, and enterprise accounts. They expect their largest customer to spend more in 2026 while the rest of the customer base grows even faster. If that actually happens, it could be a meaningful new opportunity to reduce concentration risk.

Valuation-wise, the stock's trading around 35.7x forward earnings versus the industry average of 26.7x, which is definitely a premium. Analyst estimates have drifted slightly—2026 EPS consensus came down from $1.12 to $1.01 over the past week. That said, long-term expectations remain strong with 36% projected revenue growth in 2026 and nearly 72% EPS growth expected in 2027.

Where does this leave us? The stock's around $44.43 as of early March, well below its 52-week high of $93.85 but above the low of $26.41. Technically, the company's sitting on legitimate new opportunity quotes in the AI services space—the data infrastructure play is only getting more critical as these AI systems scale. But the market's right to be cautious about near-term margin dynamics and customer concentration. You've got real growth drivers here, but also real execution risks in the next 12 months.

For long-term investors, there's probably something worth considering in Innodata's positioning. But the recent sell-off suggests the market wants to see more proof on profitability stability and customer diversification before getting excited again. Watching how they execute on margin recovery and new customer wins over the next couple quarters should tell you a lot about whether this pullback was justified or if it created a genuine new opportunity for entry.