Meta moves into physical AI with acquisition of robotics AI startup
Meta Platforms has acquired Assured Robot Intelligence (ARI), a robotic artificial intelligence startup focused on humanoid systems. That’s because the company is expanding its AI efforts beyond software to models that can help robots operate in physical environments. Financial terms were not disclosed.
Meta confirmed the acquisition. wall street journalARI said it is working on robot intelligence, which aims to help robots understand, predict and adapt to human behavior in complex environments.
ARI co-founder Xiaolong Wang announced the acquisition in a post on X, which read in part:
“When we started ARI a year ago, our mission was clear: to achieve physical AGI. Through deep customer engagement and real-world deployments, it became clear that we needed truly versatile physical agent training to address the great opportunities ahead.
“We believe that this agent will be humanoid, and that scaling will come not just by remote control but by learning directly from human experience. The Meta ecosystem brings together the key components needed to make this vision a reality. We join the Meta Superintelligence Lab (MSL) to help bring personal superintelligence into the physical world.”
ARI’s team, including co-founders Wang and Leler Pinto, will join Meta’s AI research organization, Meta Superintelligence Lab. The group is expected to work on model capabilities for robot control, self-learning, and whole-body humanoid control.
The acquisition appears to be aimed at adding robotics AI talent and technical expertise rather than purchasing a finished robotics product. ARI was developing basic models of humanoid robots, including systems that could support physical tasks such as housework.
The founders bring academic and industry experience in robotics and machine learning. Mr. Wang previously worked as a researcher at Nvidia and is an associate professor at the University of California, San Diego. Meanwhile, Pinto previously taught at New York University and co-founded Fauna Robotics, a small humanoid robot startup that was later acquired by Amazon.
The deal comes as Meta continues to increase spending on AI infrastructure and related research. wall street journal Meta recently reported raising its 2026 capital expenditure forecast by $10 billion to a range of $125 billion to $145 billion, citing rising component costs and additional spending on AI data centers.
Meta is also shifting resources toward AI after years of significant investments in augmented reality and virtual reality. The ARI acquisition suggests that Meta sees robotics as a potential extension of its AI strategy, but the company has not announced any consumer-facing humanoid robot products or a timeline for doing so.
The acquisition fits into a broader trend in the industry, often referred to as “embodied AI” or “physical AI.” Big tech companies and startups are increasingly turning to models that can interact with the real world through robots, in addition to generating text, images, code, and video. Challenging issues in this model include perception, dexterity, manipulation, navigation, safety, and adaptation to the unpredictable human environment.
The deal also reflects growing interest in the software layer of robotics. Rather than focusing solely on building robot bodies, companies are developing “brains” that enable robots to learn tasks, manipulate objects, and operate across different hardware platforms. This is still an early market, and it remains unclear whether general-purpose humanoid robots will find commercial use in homes, warehouses, factories, and medical settings.
At Meta, while competitors are also investing in physical systems, ARI has a small specialized robotics team in its AI organization. Amazon acquired Fauna Robotics earlier this year, but other startups are also raising money to build robot learning systems and humanoid platforms.
The commercial case remains uncertain. Humanoid robots face high cost, reliability issues, safety requirements, regulatory issues, battery limitations, and difficulty consistently performing common physical tasks. Even advanced AI models can struggle when moving from digital benchmarks to the home, office, or industrial setting.
About the author
John K. Waters is the editor-in-chief of many Converge360.com sites focused on high-end development, AI, and future technologies. He has been writing about Silicon Valley’s cutting-edge technology and culture for more than 20 years and is the author of more than a dozen books. Co-wrote the script for the documentary film Silicon Valley: 100 years of renaissanceaired on PBS. You can contact him at: [email protected].
