Innodata reports 3Q earnings, raises outlook; eyes growth in AI agents and robots

The management of Innodata, which supplies data and tools for training AI models, has raised its full-year guidance despite the recent loss of a major contract. The loss had triggered mass layoffs at Innodata and a selloff in its stock as investors feared slowing growth.

The management has identified training AI agents and robots as new growth opportunities. Following the third-quarter earnings release, the stock jumped more than 9% in premarket trading on Friday before dropping around 7% in the regular session, tracking the broader market.

Innodata’s third-quarter revenue surged 53% year over year to a record $95.6 million, the company reported. That beat the Wall Street consensus by 15%. The company added five new customers in the third quarter and won additional work from an existing one, which made it Innodata’s largest, CEO Rahul Singhal said. He did not identify the companies.

Net income rose 24% to $10.3 million in the quarter. Diluted earnings per share increased 21% to $0.29, beating expectations by 70%, according to the earnings release.

The results allowed the management to raise its full-year revenue growth guidance from 40-42% to 42% or more. The guidance could have been stronger had the company’s historically largest customer not terminated one of its contracts in October, Singhal said on the earnings call. The contract loss led to mass layoffs in India and the Philippines and a sharp drop in Innodata’s stock. Investors feared it could threaten the company’s ability to meet growth expectations for 2027.

Singhal said the move would reduce fourth-quarter revenue by $12 million. “The good news is that we are expecting this revenue to be replaced by new programs at this customer by the first quarter,” he said.

Several factors will support the company’s growth over the next 12 months, including physical AI, or AI robotics, agent safety, and enterprise AI, Singhal said on the call.

Venture investment in physical AI reached $47.4 billion in the first half of 2026, exceeding the total for 2022-2024 combined, Singhal noted. As robots are deployed in factories, warehouses, and homes, demand for data to train them for specific tasks is expected to grow, Singhal reckons. Innodata launched a motion capture lab in September to train humanoid robots. The company is preparing for a pilot project that, if scaled up in 2027, could generate $7 million in annualized revenue.

Singhal also identified agent training as a growth driver. An agent learns by making mistakes, he explained, but companies cannot allow that on live systems because of safety and privacy concerns and because many actions cannot be undone. Agents therefore train in a sandbox environment where they can perform tasks repeatedly and learn from their mistakes, an opportunity Innodata plans to monetize.

In the enterprise sector, models are now being asked to perform expert work. “We are building sophisticated RL gyms to train these models, and we are building what goes inside them as well,” Singhal told analysts.

Testing the safety of AI agents is another source of growth for Innodata. As agents are deployed in healthcare and finance, the cost of mistakes rises, Singhal said. In several pilot projects, Innodata is testing whether an agent will circumvent rules or fabricate results to complete a task.

Following the third-quarter earnings, Innodata shares jumped 9.14% in premarket trading on Friday before turning lower after the opening bell. At the time of this writing, the stock was trading at around $36 per share, down around 30% year to date.

Wall Street, however, remains upbeat on the stock: all four analysts tracked by MarketWatch recommend buying. The average target price of $123.70 per share implies more than 220% upside.