Andrew Bosworth says Meta is working on humanoid robots, internally called “Metabot”, as its next “AR size bet”, envisions licensing software to robot makers
During a recent trip to Meta HQ, I took the opportunity to ask one of the company's top execs about its new robotics effort.
Context & Ripple Effects
Meta’s humanoid-robot effort extends a robotics foundation that included Habitat 3.0 simulation tools and the HomeRobot platform, while placing the initiative at a scale Bosworth compares with AR.
The stated licensing ambition matters because it positions Meta less as a single-device vendor and more as a prospective software supplier to robot manufacturers. Related coverage later adds Meta’s acquisition of a startup building AI models for robots, reinforcing that the company is assembling capabilities around robot intelligence.
First-order effects
- Meta is formally signaling humanoid robotics as a major internal investment area, alongside its existing AR and AI priorities.
- Robot makers become the intended external customers for Meta’s software, rather than merely buyers of a Meta-branded machine.
Second-order effects
- A licensing strategy makes software compatibility, simulation, and robot AI the near-term competitive surfaces; hardware partners would gain an alternative to building every layer in-house.
- Meta’s robotics work can reuse and concentrate investment around its existing simulation and home-robot efforts, while increasing the importance of integrating AI models with physical-device software.
Third-order effects
- If Meta can establish a licensable robotics stack, humanoid robotics could develop around platform providers and hardware assemblers rather than fully vertically integrated vendors.
- As humanlike robots move from research platforms toward broader deployment, questions captured by robot-focused AI model development—including behavior control and accountability—are likely to become more central to product and policy choices.
The trend: Humanoid robotics is increasingly becoming a contest to own the AI-and-software layer that can scale across multiple hardware makers.