Meta acquires Assured Robot Intelligence to accelerate humanoid robot‑learning models

Meta has acquired humanoid robotics startup Assured Robot Intelligence (ARI) and folded the company’s team, including co‑founders Xiaolong Wang and Lerrel Pinto, into its Superintelligence Labs research unit, a Meta spokesperson told TechCrunch. The purchase price was not disclosed. Meta characterized ARI as working at the frontier of robotic intelligence with the goal of enabling robots to understand, predict and adapt to human behaviors in complex, dynamic settings.

ARI had been developing foundation models aimed specifically at whole‑body humanoid control, with an emphasis on practical physical tasks such as household chores. According to public reporting, the startup pursued model architectures and training regimes meant to help robots generalize across diverse, contact‑rich interactions rather than simply replaying scripted behaviors. ARI had raised an undisclosed seed round from AI seed firm AIX Ventures prior to the acquisition.

The deal brings into Meta two co‑founders with academic and industry research backgrounds. Xiaolong Wang previously worked as a researcher at NVIDIA and served as an associate professor at UC San Diego, and Lerrel Pinto came from an academic role at NYU and earlier co‑founded the kid‑size humanoid company Fauna Robotics, which was acquired by Amazon. Meta and public materials frame the incoming team as adding deep expertise in robot control and self‑learning applied to whole‑body humanoid control.

The acquisition follows years of internal work at Meta on humanoid robotics. A leaked Meta memo from about a year ago publicly outlined ambitions to combine AI models and custom hardware toward consumer humanoid systems, and the ARI buy fits into that longer‑running effort. Observers outside Meta have also signaled a surge of interest and investment in embodied AI; industry forecasts cited in reporting range widely, for example from Goldman Sachs’ projection of $38 billion by 2035 to Morgan Stanley’s estimate of $5 trillion by 2050, a spread that underscores both potential and uncertainty around the technology.

While the move is positioned as a technical acceleration — tightening feedback loops between simulated model training and physically grounded control and advancing sample‑efficient algorithms that can learn from real interactions — reporting leaves key operational details unspecified. TechCrunch’s coverage and related materials do not disclose the acquisition price, specific product road maps, or measurable technical results from ARI’s models, and Meta has not announced a consumer product timeline. The consolidation also highlights broader questions about competition, safety and governance as large labs integrate startup talent to push robots that will operate directly around people.