Meta has acquired humanoid robotics startup Assured Robot Intelligence for an undisclosed sum. The team, including co-founders Xiaolong Wang and Lerrel Pinto, will join Meta's Superintelligence Labs research division. The deal signals Meta's growing seriousness about building AI-powered robots — a pursuit that many researchers believe is essential for achieving artificial general intelligence.
Who ARI Is
ARI was building foundation models for humanoid robots designed to perform physical labor such as household chores. The models enable robots to understand, predict, and adapt to human behaviors in complex environments.
Co-founder Xiaolong Wang was previously a researcher at Nvidia and an associate professor at UC San Diego with a string of prestigious awards in robotics and computer vision. Co-founder Lerrel Pinto taught at NYU and co-founded Fauna Robotics, a kid-sized humanoid startup that Amazon acquired last month. Pinto also holds multiple awards in AI research.
Both founders bring expertise that Meta has been acquiring aggressively. A Meta spokesperson said the team will bring deep knowledge in designing models for robot control and self-learning applied to whole-body humanoid control.
Why Meta Wants Robots
Meta's humanoid ambitions have been building for years. A leaked internal memo from early 2025 revealed plans to build AI models and hardware for a consumer robot. Meta's AI research division has published extensively on robotics. And the company's Reality Labs — which has lost over $65 billion since 2020 — has been exploring physical AI interfaces alongside its AR/VR products.
The strategic logic connects to a broader thesis in the AI industry. Many researchers now believe the path to AGI requires training AI in the physical world — where robots learn through direct interaction rather than data alone. Software-only AI models trained on internet text and images may plateau. Physical embodiment — robots that learn by doing — could be the next breakthrough.
Meta CEO Mark Zuckerberg has been positioning the company for this future. The company is spending $115 to $135 billion on AI infrastructure in 2026. It is buying millions of custom chips from Amazon. And it is fighting to acquire AI startups against geopolitical headwinds. Adding ARI's robotics expertise to that portfolio makes Meta one of the few companies pursuing AI across software, hardware, wearables, and now physical robots.
The Humanoid Robot Race
The ARI acquisition joins a wave of activity in humanoid robotics. China recently showcased robots that beat humans at a half-marathon in Beijing. Amazon acquired Fauna Robotics. Tesla continues developing its Optimus humanoid. And startups across the US, China, and Europe are racing to build robots capable of useful physical work.
The market projections vary wildly. Goldman Sachs forecasts the global robot market reaching $38 billion by 2035. Morgan Stanley estimates $5 trillion by 2050. The spread reflects both the enormous potential and the uncertainty around technology that is still far from commercial readiness.
For Meta, even if a consumer humanoid product never ships, the research has value. Training AI models through physical interaction produces capabilities that transfer back to software — better spatial reasoning, improved understanding of cause and effect, and more robust real-world knowledge.
What It Means
Meta's ARI acquisition positions the company at the intersection of AI software and physical robotics. Combined with its talent war with Thinking Machines Lab, its smart glasses strategy, and its massive infrastructure spending, Meta is building one of the broadest AI portfolios in the industry.
Whether that breadth becomes a competitive advantage or a distraction remains the central question of Zuckerberg's AI strategy. But the ARI deal makes one thing clear: Meta is not content to build AI that lives only on screens. It wants AI that can walk, reach, and interact with the physical world. The race to build that future just gained another well-funded competitor.
