Meta is placing a massive bet on humanoid robots, with CTO Andrew Bosworth comparing it to the company's multi-billion dollar augmented reality investment. In an exclusive interview with The Verge's Alex Heath, Bosworth revealed that CEO Mark Zuckerberg personally directed the creation of a new robotics research team earlier this year, signaling Meta's intent to dominate the next frontier of AI-powered hardware.
Meta just threw down the gauntlet in robotics. The social media giant is preparing what CTO Andrew Bosworth calls an "AR-size bet" on humanoid robots - a comparison that suggests the company's ready to spend billions chasing its next moonshot. During a candid conversation at Meta's headquarters, Bosworth pulled back the curtain on a robotics initiative that could reshape how we think about AI-powered hardware. The timing isn't accidental. As competitors like Tesla showcase their Optimus robots and startups race to crack the humanoid code, Meta is betting it can win through software supremacy rather than hardware prowess. "I don't think the hardware is the hard part," Bosworth told The Verge's Alex Heath. "I'm not saying the hardware isn't also hard, but it's not the bottleneck. The bottleneck is the software." To illustrate his point, Bosworth picked up a glass of water during the interview. While robots can run and do backflips on stable ground, he explained, the delicate task of grasping that glass without crushing it or spilling water remains the holy grail of robotics. It's a problem Meta thinks it can solve through its unique combination of AI research and massive data capabilities. The company's approach centers on "Metabot," their internal humanoid prototype, but the real strategy mirrors what made Google dominant in mobile. Rather than compete as a hardware manufacturer, Meta plans to license its robotics software platform to other companies. "I don't care about us being the hardware manufacturers," Bosworth explained, envisioning a future where any robotics company can use Meta's software "as long as your robot meets these specs." This software-first strategy gets a boost from Meta's newly formed Superintelligence AI lab, led by former Scale CEO Alexandr Wang. The team is working with the robotics group to build what Bosworth calls a "world model" - essentially an AI system that can simulate the complex physics required for dexterous manipulation. It's the same concept that Google DeepMind's Demis Hassabis has championed, bringing spatial awareness to AI systems. The talent acquisition tells its own story about Meta's ambitions. Former Cruise CEO Marc Whitten now leads the robotics team, while the company poached Sangbae Kim - whom Bosworth calls "the greatest tactical roboticist in the game right now" - from MIT earlier this year. Even more telling, Meta convinced Jinsong Yu, the architect behind their Orion AR glasses prototype, not to retire and join the robotics effort instead. But Meta's approach deliberately differs from Tesla's Optimus strategy. Where Tesla relies purely on vision-based AI - the same approach powering their self-driving cars - Bosworth questions how they'll gather the massive datasets needed for robotics. "I can see how Tesla cars are getting enough data. I can't figure out how they're going to get robotic data," he said. Meta's answer involves building that dataset from scratch through simulation and their world model approach. The company isn't chasing perfection either. While Tesla's Optimus boasts 23 degrees of freedom in each hand, Bosworth takes a pragmatic view: "I don't think you need 23 degrees of freedom in your hand. Two thumbs would be nice. I only need two thumbs." This philosophy of "good enough" robotics could prove decisive in a market where companies often get bogged down pursuing perfect human-like dexterity. The timing couldn't be better for Meta. As the robotics industry grapples with hardware costs and manufacturing challenges, a software-licensing model offers instant scalability. Companies like Nvidia and Qualcomm are competing to provide the silicon backbone for humanoids, but Meta could own the software layer that makes these machines actually useful. Industry experts see this as Meta's attempt to avoid the mistakes that plagued its VR push. Rather than betting everything on expensive hardware that takes years to reach consumers, the company's positioning itself to benefit regardless of which robotics manufacturers succeed. It's a hedge that could pay off massively if the humanoid robot market explodes as predicted.
Meta's robotics bet represents more than just another R&D project - it's a fundamental shift toward software-powered hardware dominance. By focusing on the AI and simulation challenges that have stumped the industry, rather than getting bogged down in hardware manufacturing, the company's positioning itself to win regardless of which robotics startup or tech giant builds the best physical robots. If Bosworth's vision pans out, we could see Meta become to robotics what Google became to smartphones: the invisible software layer that powers an entire industry. For consumers, that could mean more capable, affordable robots hitting the market sooner than expected.
