Artificial intelligence that operates in the real world — robots, autonomous vehicles, machines that see and move — needs training data that is far messier than plain text. That is the space iMerit has built its reputation in, and now the company belongs to enterprise services firm ExlService Holdings (NASDAQ: EXLS).
The deal was announced on June 24, 2026 and completed on August 3, 2026. It carries a total value of $310 million, split into $170 million paid upfront and up to $140 million in incentives and earnouts spread over two years, contingent on hitting agreed milestones. The structure tells you plenty: EXL is paying a solid sum now and dangling a larger reward for sustained performance.
Why does data labeling command that kind of money? Because “physical AI” — the term for systems that perceive and act in physical environments — lives or dies on the quality of its training inputs. iMerit specializes in exactly this, handling AI model training, evaluation and reinforcement learning across a broad spread of data types.
Its expertise covers:
- Text and image annotation
- Video labeling for motion and scene understanding
- Voice data for speech systems
- LiDAR point-cloud data — the depth-sensing backbone of self-driving perception
That combination is the foundation for AI powering robotics, autonomous vehicles and what the companies describe as intelligent real-world environments. LiDAR annotation in particular is notoriously labor-intensive; every frame of a driving scene can contain thousands of points that a machine must learn to identify as pedestrian, curb, cyclist or shadow. Reliable physical AI is only as trustworthy as the humans and pipelines that teach it.
For EXL, the acquisition slots neatly into a strategy of moving up the AI value chain. The firm already sells analytics and digital operations to large enterprises; folding in iMerit’s model training, evaluation and reinforcement-learning capabilities gives it an end-to-end story — from raw sensor data to a deployed model that behaves in the field. Reinforcement learning matters here too, since real-world agents improve through feedback loops rather than a single static training pass.
iMerit, for its part, gains the scale and enterprise reach of a publicly traded parent. The company is recognized as a leader in AI model training, evaluation and reinforcement learning, and the executives from both sides have framed the tie-up around a shared goal: helping developers ship physical AI they can actually depend on.
This is a business-to-business play rather than a shiny gadget you can order, but it is a telling one. As robots and autonomous systems edge further into everyday settings, the unglamorous work of labeling and validating sensor data is becoming one of the most valuable commodities in the AI economy — and, at $310 million, one of the more expensive.
