Meta buys Assured Robot Intelligence to double down on AI for humanoid robots
Meta is doubling down on humanoid robots: it has bought Assured Robot Intelligence and will move the team into Superintelligence Labs. ARI built models that help machines understand human behavior, adapt to complex household environments and learn to control the whole body. Key co-founders Lerrel Pinto and Xiaolong Wang are joining Meta along with the startup. *Meta is designated an extremist organization and banned in Russia.
AI-processed from 3DNews AI; edited by Hamidun News
Meta Platforms closed its acquisition of the startup Assured Robot Intelligence on May 1, 2026. The deal strengthens the company's robotics effort: the ARI team will move into Superintelligence Labs and work on AI for humanoid machines that need to operate safely around people.
What Meta bought
Assured Robot Intelligence developed models that help robots understand, predict and adjust to human behavior in complex, constantly changing environments. This is not about an assembly line where everything is known in advance, but about apartments, offices and other spaces built for people. In those conditions a machine has to do more than carry out a command: it must keep its balance, correct its movements on the fly and account for what the person next to it is doing.
"This is a company at the forefront of robotics AI that helps robots
understand, predict and adapt to human behavior."
Meta did not disclose the financial terms. ARI employees in New York and San Diego will join Meta together with co-founders Lerrel Pinto and Xiaolong Wang. For Meta this is not just the purchase of a small team, but a way to quickly gain applied expertise in controlling humanoid robots, self-learning and tactile interaction with a real environment, where mistakes happen and pre-scripted scenarios almost never apply.
Why Meta wants it
Meta already has Robotics Studio, created in 2025 to develop core technologies for humanoid robots. The ARI acquisition fills an important layer of that strategy: the company needs not only sensors and hardware, but also models that turn a set of motors and cameras into a machine able to act in the real world. In effect, Meta is building not a single showcase robot but a technology platform for a whole class of devices.
After the deal Meta gets several important assets that strengthen both the software and hardware sides of its robotics strategy:
- control and self-learning models for humanoid robots
- expertise in whole-body control, where a robot coordinates its entire body rather than individual limbs
- a team with experience in household robotics and applied machine learning
- a chance to offer sensors, software and AI components not only for its own use but to other manufacturers
This approach resembles the role of Google and Qualcomm in the smartphone market: you don't have to build every device yourself if you can become the supplier of a key technology layer. That is what the logic of the deal hints at. Meta wants to be not just a player in the robot market but one of those who define its base stack, integration standards and the set of components on which future products for the industry as a whole will be built.
Who strengthened the effort
The makeup of the team is telling too. Lerrel Pinto previously helped create Fauna Robotics, a startup that Amazon bought in March 2026, although he had left it back in 2025. Xiaolong Wang did research at Nvidia before ARI and also worked as an academic researcher. For Meta this is a convenient way to bring in specialists who combine theory, robotics experience and an understanding of how to get models running on real hardware.
The deal also shows how quickly competition in physical AI is intensifying. Tesla, Alphabet and Amazon have openly shown interest in humanoid systems, and Meta itself has been actively expanding Superintelligence Labs in recent months. Against that backdrop the ARI purchase looks less like an experiment for the future and more like part of a broader race: big companies want to be first to build general-purpose software for robots that will handle household and service tasks.
What it means
The AI market is shifting steadily from chatbots and content generation to systems that can act in the physical world. The ARI purchase shows that Meta is betting not only on models for text and media, but also on intelligence for machines with hands, sensors and tasks in real environments. If the strategy works, the fight for the humanoid market will be waged not only among hardware makers, but also among the companies that supply them with the "brain."
*Meta is designated an extremist organization and banned in Russia.
- Nov 1, 2026MoneyGate safety layer blocked all 42 fraudulent requests to local AI models
- Oct 31, 2026Open-source LLM gateway KTMS 1.0 plugs your own GPUs into the cloud and survives outages
- Oct 31, 202622B-Parameter LTX-2.5 Runs on a 10 GB RTX 3080, but Takes 5 Minutes per Shot
- Oct 29, 2026T-Tech Introduces LIMe, a Method That Trains Language Models With 15% Less Compute
Key Questions Answered
What Meta bought?
Assured Robot Intelligence developed models that help robots understand, predict and adjust to human behavior in complex, constantly changing environments. This is not about an assembly line where everything is known in advance, but about apartments, offices and other spaces built for people. In those conditions a machine has to do more than carry out a command: it must keep its balance, correct its movements on the fly and account for what the person next to it is doing. "This is a company at the forefront of robotics AI that helps robots understand, predict and adapt to h
Why Meta wants it?
Meta already has Robotics Studio, created in 2025 to develop core technologies for humanoid robots. The ARI acquisition fills an important layer of that strategy: the company needs not only sensors and hardware, but also models that turn a set of motors and cameras into a machine able to act in the real world. In effect, Meta is building not a single showcase robot but a technology platform for a whole class of devices. After the deal Meta gets several important assets that strengthen both the software and hardware sides of its robotics strategy: control and self-learning m
Who strengthened the effort?
The makeup of the team is telling too. Lerrel Pinto previously helped create Fauna Robotics, a startup that Amazon bought in March 2026, although he had left it back in 2025. Xiaolong Wang did research at Nvidia before ARI and also worked as an academic researcher. For Meta this is a convenient way to bring in specialists who combine theory, robotics experience and an understanding of how to get models running on real hardware. The deal also shows how quickly competition in physical AI is intensifying. Tesla, Alphabet and Amazon have openly shown interest in humanoid system
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.