TL;DR
- ARI Acquisition: Meta closed its purchase of humanoid-AI startup Assured Robot Intelligence on May 1, folding co-founders Pinto and Wang into Superintelligence Labs.
- Platform Bet: Rather than ship its own humanoid, Meta plans to license robotics software in an Android-style platform play.
- Market Stakes: Goldman Sachs projects a $38 billion robot market by 2035; Morgan Stanley pegs the humanoid segment at $5 trillion by 2050.
Meta has acquired Assured Robot Intelligence, calling it a step in its robotics push, with the startup’s team joining Meta’s humanoid robotics push inside Superintelligence Labs alongside Meta Robotics Studio. The deal closed Friday, May 1, on undisclosed terms, and slots into the company’s earlier humanoid-AI ambitions.
The market backdrop is bullish: Goldman Sachs projects the total addressable market for humanoid robots at $38 billion by 2035. Morgan Stanley pegs the humanoid segment at $5 trillion by 2050. Meta appears to be betting that, while Tesla, Amazon, and several Chinese vendors chase humanoids with vertically integrated hardware, the more durable position is the underlying intelligence layer, monetized as software.
What Meta Is Buying
ARI’s team, including co-founders Lerrel Pinto and Xiaolong Wang, will work with Meta Robotics Studio, the in-house unit Meta launched last year for humanoid systems. Concentrated in San Diego and New York, the acquired group was building foundation models for humanoid robots aimed at household-grade physical tasks, and had previously raised an undisclosed seed round from AIX Ventures; co-founder Xuxin Cheng rounds out the team. By Meta’s own framing, the new hires bring expertise across model architecture, robot control systems, and self-learning techniques applied to full-body humanoid platforms.
“This team, led by Lerrel Pinto and Xiaolong Wang, 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.”
Meta spokesperson
The Founders And Their Track Record
Xiaolong Wang was previously a researcher at Nvidia and an associate professor at UC San Diego, and has demonstrated a humanoid control system that streams a robot’s stereo camera feed to a VR headset, with the robot turning its head as the wearer does. The demo is a useful tell: it points to a research thesis grounded in human teleoperation as a data source for humanoid policy training, rather than purely simulated environments.
Pinto previously taught at NYU and co-founded the kid-size humanoid startup Fauna Robotics before departing in 2025. Amazon acquired Fauna in March 2026, putting Pinto on the unusual side of two humanoid-AI acquisitions inside roughly twelve months. Between them, Wang and Pinto bridge the academic, hyperscaler, and venture-funded humanoid stacks that Meta now wants to consolidate inside one team.
Excited to share that Assured Robot Intelligence (ARI) has joined @Meta to help build the future of humanoid intelligence!
When we started ARI one year ago, our mission was clear: achieve physical AGI. Through deep customer engagements and real-world deployments, it became clear… https://t.co/4hGmcgcJzt
— Xiaolong Wang (@xiaolonw) May 1, 2026
Platform Versus Vertical
Tesla, Google, and Amazon are all pursuing humanoid robot projects of their own. Tesla earlier in 2026 converted its Fremont line from Model S and X production to Optimus humanoid manufacture. Amazon’s March 2026 pickup of Fauna Robotics was a vertical play of the same kind, attaching humanoid hardware to a logistics roadmap.
Meta’s stated path is different. Rather than ship a Meta-branded humanoid, the company envisions a platform role comparable to Android, with an explicit aim to license robotics software to other companies. Meta CTO Andrew Bosworth set the doctrine in 2025, stating “Software is the bottleneck.”
Google DeepMind, the closest peer on that path, is mounting its own Android-for-Robots push, a sign the platform-versus-vertical split is itself becoming contested. Backing the bet, Meta is committing hundreds of billions of dollars to AI, and a software-licensing path lets that spend earn a return across many robot brands rather than just one.
Prior Coverage And What To Watch
Earlier moves point in the same direction. Meta released its V-JEPA 2 world model with explicit robotics framing last year, and has spent a reported $2 billion on Manus to deepen its agentic-AI bench – although that acquisition appears to be cancelled now after Chinese regulatory intervention.
ARI fits as the model and control layer connecting those pieces to physical hardware. Meta’s Superintelligence Labs will price its platform thesis when an outside humanoid maker (like Figure, Apptronik, 1X, or a Chinese OEM) ships hardware running its models on a signed, disclosed license, rather than another internal demo from the Robotics Studio bench.
