Meta Platforms has closed its acquisition of Assured Robot Intelligence, a stealth startup building foundation models for general-purpose robots. The deal landed on Friday, May 1, 2026, and folds the team into Meta Superintelligence Labs as part of CEO Mark Zuckerberg's bid to lead the next wave of humanoid hardware.
Why Meta Wanted ARI
Meta described Assured Robot Intelligence as operating "at the frontier of robotic intelligence designed to enable robots to understand, predict and adapt to human behaviors in complex and dynamic environments." Founders Lerrel Pinto, a New York University assistant professor, and Xiaolong Wang, a former NVIDIA researcher and UC San Diego associate professor, will work alongside Meta Robotics Studio, the humanoid team Meta stood up in 2025.
Inside the Acquisition
Financial terms were not disclosed. The startup had remained largely under the radar, focusing on training large behavior models and physical AI policies that can transfer across embodiments. Wang's MLSys 2024 best-paper work on AI model optimization is expected to feed directly into Meta's effort to compress humanoid policies into low-latency on-robot inference stacks.
A Bigger Humanoid Bet
The acquisition follows a string of robotics moves by Meta, including poaching engineers from Figure, Apptronik and Boston Dynamics. Meta is positioning its humanoid stack as an "Android for robots" — a software platform that hardware partners can license — a strategy that would directly compete with NVIDIA's Isaac and GR00T ecosystem.
Race for Robotic Foundation Models
The deal underscores how robotic foundation models have become the most contested layer in the humanoid stack. Google DeepMind, Physical Intelligence, FieldAI and Skild AI all command multibillion-dollar valuations, and Meta has signaled it will pay aggressively to close the gap.
Reporting based on coverage from Bloomberg, Engadget, TechCrunch and The Next Web.