Meta purchases Assured Robot Intelligence to develop the Android for humanoid robots.

      Meta has acquired Assured Robot Intelligence, a startup co-founded by former Fauna Robotics co-founder Lerrel Pinto and ex-Nvidia researcher Xiaolong Wang, to enhance its humanoid robotics platform strategy. This acquisition, which brings full-body robot control models and tactile sensor technology into Meta Superintelligence Labs, indicates Meta’s ambition to become the Android equivalent for humanoids: providing the intelligence infrastructure while allowing others to build the physical robots.

      Lerrel Pinto previously co-founded Fauna Robotics, a startup known for creating an approachable bipedal robot named Sprout, which he left in 2025. Amazon purchased Fauna in March, acquiring its 50 employees and its $50,000, three-and-a-half-foot-tall dancing robot, to enter the consumer robotics space. Following that, Pinto partnered with Xiaolong Wang, who is a former Nvidia researcher and associate professor at UC San Diego, to launch Assured Robot Intelligence. Wang won the MLSys 2024 Best Paper Award for his contributions to AI model optimization. On Friday, Meta finalized the acquisition of ARI, with both founders joining Meta Superintelligence Labs the same day. The financial details of the deal were not disclosed. The key question isn't what Meta invested in the startup, but rather how it plans to utilize the technology and what that implies about its vision for the development of the humanoid market.

      Meta aims to replicate the success of Google’s Android operating system and Qualcomm’s chips in the smartphone sector by establishing a foundational platform for robotics. The company initiated Meta Robotics Studio last year, appointed former Cruise CEO Marc Whitten to lead the initiative, and began hiring about 100 engineers to create in-house humanoid hardware alongside the AI models needed to operate it. CTO Andrew Bosworth has characterized humanoid robots as a significant focus for Meta, akin to augmented reality, a field in which Meta has already invested tens of billions of dollars through its Reality Labs division. The ARI acquisition specifically adds capabilities to this endeavor by providing robot control models that enable humanoids to interpret, anticipate, and adjust to human behavior in unpredictable environments.

      Meta's platform strategy is clear: it plans to develop sensors, software, and AI models for robots and make these resources accessible to the wider industry, allowing manufacturers not owned or controlled by Meta to utilize the technology. This mirrors the Android model applied to physical devices; in the smartphone industry, Google offered its operating system for free and generated revenue through search, advertising, and the Play Store ecosystem. Similarly, in robotics, Meta would distribute the intelligence layer at no cost while monetizing the data, model ecosystem, and integration with its existing platforms, which already reach 3.3 billion daily users. Meta has been aggressively acquiring AI talent, hiring notable figures such as five founding members of Thinking Machines Lab and one researcher with a reported six-year compensation of $1.5 billion. The ARI acquisition aligns with this pattern, adding a small team with pioneering capabilities that can be immediately integrated into the Superintelligence Labs research division.

      The core technological contribution from ARI revolves around "robotic intelligence" that allows robots to comprehend, anticipate, and adapt to human behaviors in complex and dynamic settings. This translates to AI models for full-body humanoid control, enabling a robot to coordinate its limbs, maintain balance, and adjust movements based on real-time sensory data from an unpredictable physical environment. Wang's award-winning research on activation-aware weight quantization, which was instrumental in Nebius’s recent acquisition of Eigen AI, is also significant here, as it focuses on optimizing AI models for efficiency on the limited computing power available within a robot, rather than relying on a connection to a distant data center.

      Additionally, the company has created e-Flesh, a tactile sensor that detects deformations in 3D-printable microstructures using magnets and magnetometers. Tactile sensing is a major unsolved issue in humanoid robotics; a robot equipped with cameras and lidar still struggles to distinguish between grasping an egg and a tennis ball without tactile feedback. The existing divide between robots' learning in simulations and their performance in the real world remains a major hurdle to large-scale deployment. ARI’s innovations in self-learning for robot control along with its sensor technology aim to bridge both sides of this divide: by improving models and enhancing sensory input.

      The humanoid robotics market has rapidly transitioned from speculative to competitive within just 18 months. Tesla is set to begin large-scale production of its Optimus V3 humanoid between July and August, targeting an annual production capacity of one million units by late 2026, with pricing anticipated between $20,000 and $30,000. 1X Technologies has launched a factory in Hayward, California, with a goal of producing 10,000 NEO humanoid robots in its first year, selling out its initial production run within five days of opening preorders. Apptronik has secured $520 million, valuing the company at $5 billion,