BigBear.ai revenue fell nearly nineteen percent to one hundred twenty-seven point seven million dollars in the fiscal year ended December 2025 while Innodata revenue jumped forty-seven point six percent to two hundred fifty-one point seven million dollars over the same period, leaving the two small-cap AI names on opposite sides of the profitability line.
Government contracts versus private-sector scale
BigBear.ai builds predictive analytics and decision-support software for defense and homeland security agencies. That customer base delivered fifty-one percent of 2025 revenue from clients each exceeding ten percent of sales, a concentration the company flags as a risk. The business posted a net loss of two hundred ninety-three point nine million dollars and a net margin of negative two hundred thirty point two percent. Free cash flow came in at negative forty-six point three million dollars. The balance sheet shows a current ratio of one point eight times and zero debt-to-equity, but the firm is also contending with a class action lawsuit and the restatement of multiple prior years that has delayed regulatory filings.
Innodata rides the generative AI training wave
Innodata supplies data engineering and human annotation services to large language model builders. One customer in its Digital Data Solutions segment accounted for fifty-eight percent of 2025 revenue, a similar concentration risk from the other side of the AI supply chain. The pivot produced thirty-two point two million dollars of net income and a twelve point eight percent net margin. Free cash flow reached thirty-five point six million dollars, though stock-based compensation represented twenty-three point eight percent of operating cash flow, inflating the cash generation figure. The current ratio sits at two point seven times with zero debt-to-equity.
What the divergence tells you about the AI trade
The split reflects a broader market dynamic: government procurement cycles are long, lumpy, and subject to unilateral termination, while private-sector demand for labeled data has scaled rapidly alongside model releases. BigBear.ai’s path to growth now hinges on acquisitions that have yet to close and on reversing a revenue decline that predates the current AI boom. Innodata’s profitability is real but leans heavily on a single relationship and on a non-cash compensation add-back that flatters cash flow.
Watch the filings and the customer lists
BigBear.ai’s restated financials and overdue filings are the near-term catalyst, any further delay or adverse finding in the class action will pressure the shares. For Innodata, the fifty-eight percent customer concentration means a single contract renewal or volume shift could swing the entire P&L. Both companies carry zero debt, which buys time, but time is the only thing either can afford to waste.