Meta has bought Assured Robot Intelligence (ARI), a startup that creates artificial intelligence models for humanoid robots, the company said on Friday. This deal puts the company in a fast-growing industry where major tech companies, car makers, and startups are all working to deploy physical AI at scale. The companies did not share the financial details.
Meta described the startup as “at the frontier of robotic intelligence designed to enable robots to understand, predict, and adapt to human behaviors in complex and dynamic environments,” according to a company spokesperson quoted by Bloomberg. The acquisition happened only two days after the company raised its 2026 capital spending forecast from $125 billion to $145 billion. This increase points to higher component prices and more spending on AI data centers.
Why Meta Acquired Assured Robot Intelligence, and who is joining
ARI was developing core models for humanoid robots capable of handling physical tasks, such as household chores. The startup’s co-founders have solid backgrounds in AI and robotics. Xiaolong Wang, one of the co-founders, worked as a researcher at Nvidia and is now an associate professor at UC San Diego. Lerrel Pinto, another co-founder, taught at New York University and also co-founded Fauna Robotics, a humanoid robotics startup that Amazon acquired in March. ARI raised a seed round from AIX Ventures, an AI-focused investment firm, but did not share the amount.
Pinto shared on X that ARI’s stack “is built on human experience, condensed into actionable tokens that can be rapidly adapted to” guide robot behavior. He also said the team could “transform AI that can think and talk to AI that can do, assisting humans safely and reliably in the physical world”. Wang said the team aims for “physical AGI” and explained that scaling “will come from learning directly from human experience, not teleoperation alone.”
Both co-founders and their team are joining Meta Superintelligence Labs’ research division. They will work closely with Meta Robotics Studio, which started last year to develop key technology for humanoid robots. The group, according to a spokesperson, “will bring a deep expertise in how we can design our models and frontier capabilities for robot control and self-learning to whole-body humanoid control.” The company researchers have been working on humanoid robotics technology for years. A leaked 2025 memo revealed the company’s plans to build a robot, including AI models and consumer-focused hardware.
What Meta plans to do with Humanoid Robotics Technology
The company’s plans for robotics go beyond a single acquisition. The company is developing its own humanoid robots and the AI that powers them. This includes sensors, software, and other key systems that Meta plans to share with the wider industry. According to Bloomberg’s reporting from last year, Meta’s long-term goal is to become the foundation of the humanoid robot industry, much as Android and Qualcomm’s chips became the foundation of the smartphone market.
Meta faces stiff competition in the humanoid robotics space. Major technology companies, including Tesla, Google, and Amazon, are all making significant investments in the field. Market forecasts for this field vary widely. Goldman Sachs predicts it could reach 38 billion by 2035, while Morgan Stanley estimates it could reach as much as 5 trillion by 2050. This wide range shows both the technology’s huge potential and the uncertainty as it develops.
The acquisition comes at a time when Meta is increasing its AI spending and cutting costs elsewhere. The company’s higher capital expenditure forecast for 2026 followed the company’s announcement of plans to cut about 8,000 jobs, or roughly 10% of its workforce, as it reallocated resources toward AI projects. The layoffs are set to start on May 20.
Even if Meta never launches a humanoid robot for consumers, many AI experts think that reaching artificial general intelligence, which means AI that matches or surpasses human abilities in every area, will need training in the real world. Robots must learn by interacting directly, not just from data. By making this acquisition, Meta is taking a real step toward making that idea possible.
