Innodata raises 2026 outlook after record $95.6 million quarter

By Marcus Bennett

Innodata generated record third-quarter revenue of $95.6 million, up 53% year over year, despite a program ending that is expected to cut fourth-quarter revenue by about $12 million. Management nevertheless raised 2026 revenue-growth guidance to 42% or more from 40%-42%. The result beat the $83.3 million analyst consensus and marked Innodata’s 13th consecutive period of year-over-year revenue growth.

What changed in the revenue outlook?

The quarter also marked Rahul Singhal’s first earnings call as chief executive officer. Singhal said the guidance increase could have been larger if the historically largest customer had not ended its program in mid-October. New programs at that customer are expected to replace the lost revenue by the first quarter of 2027, although several opportunities remained in late-stage discussions.

How did customer concentration shift?

Customer concentration shifted during the quarter. Singhal said a customer that was Innodata’s second-largest in the prior quarter became its largest in the third quarter.

Innodata signed a $29 million pre-training data program with that customer during the third quarter and an approximately $18 million program in the fourth quarter so far. That customer contributed less than 5% of revenue a year earlier; Innodata now has two customers of comparable scale, and management identified neither.

Innodata won five programs with another large hyperscaler that generated approximately $3.1 million for the company over the trailing 12 months. The company expects that customer to contribute more than $20 million in 2027, more than six times the trailing-12-month revenue.

The quarter also brought five new customers, including model builders and an enterprise data platform. Workers from the ended program can sometimes be redeployed, depending on skills and the timing of other customers’ training cycles. Singhal described the expected first-quarter 2027 work as “different programs, rather than a continuation” of the engagement that ended.

How strong were earnings and cash generation?

Profit rose alongside sales. Net income increased 24% to $10.3 million from $8.3 million a year earlier, while diluted earnings per share reached $0.29 against the $0.17 analyst consensus. Adjusted EBITDA rose 9% year over year to $17.7 million, equal to 19% of revenue and above the $15.4 million consensus cited by management.

Adjusted gross profit reached $38.3 million, producing a 40% adjusted gross margin that matched Innodata’s target. Singhal said margins change with program mix and that the company sometimes accepts lower-margin work to build capabilities or support future growth.

CFO Jayant Chauhan put the year-to-date effective tax rate at approximately 12%, below the 23% to 25% long-term target, because tax benefits reduced the quarter’s charge. Cash and equivalents reached approximately $164 million, or $153.7 million excluding customer prepayments.

On that basis, cash increased $19.3 million from the second quarter. Innodata had no debt, and its $50 million credit facility remained undrawn.

Where can growth come from?

Singhal identified six areas for growth over the next 12 months. The disclosures range from delivered robotics work to opportunities awaiting signed contracts.

• Physical AI: Innodata collects and annotates training data for robotics models. It has delivered more than 620,000 robot-executable labeling jobs so far this year and launched a motion-capture laboratory in September. Singhal said continuous collection and annotation could create recurring demand as robotics models take on specific tasks.
• Reinforcement-learning environments: Sandbox environments, known as reinforcement-learning gyms, let AI agents practice tasks without acting on live systems. Some safety-focused environments test whether agents cut corners or misrepresent their work.
• Trust and safety: Innodata was finalizing a statement of work (a document defining services and scope) representing approximately $6 million in potential revenue for risk evaluation and agent safety. A separate verbal customer commitment for agent safety could generate approximately $2.5 million next year; neither opportunity was a completed contract.
• End-to-end model training: Innodata included the category in its six-area plan.
• Enterprise AI: The company added approximately $5 million of annualized revenue run rate (the latest revenue pace expressed as a 12-month total) from implementation work for AI agents at a large customer in the prior quarter.
• Federal programs: The federal practice reported a verbal win on a contract worth $5 million annually for five years, with signing expected by the end of the fourth quarter.

Several disclosed growth amounts remain potential or verbal rather than completed contracts, including the $6 million opportunity, the $2.5 million commitment and the federal award. Contract conversion will therefore determine how much prospective revenue enters guidance for 2027.

How will management allocate capital?

Chauhan said “organic investment remains the company’s capital-allocation priority,” while selective acquisitions also remain under consideration. With no debt and an undrawn $50 million facility, Innodata retains capacity to pursue both.

Management includes revenue from active programs, new awards and smaller opportunities with clear visibility, but excludes large pipeline prospects that lack awards. Innodata plans to provide 2027 guidance alongside fourth-quarter results.

Innodata is a global data-engineering and technology company that helps organizations prepare, manage and use information for digital transformation and artificial-intelligence initiatives. Its services support AI and machine-learning development, training, evaluation and deployment through data annotation and labeling, collection and preparation, content transformation, document processing, natural-language services and generative-AI support.