Meta is building an Android for humanoid robots

Meta just found its next multi-billion-dollar obsession: humanoid robots.

Meta confirmed internally that it has assembled a dedicated robotics team inside its Reality Labs division. The unit is led by Marc Whitten, who stepped down as CEO of GM's self-driving division Cruise earlier this month. Reporting directly to Meta CTO Andrew Bosworth, Whitten has been cleared to hire around 100 engineers this year alone.

The Android playbook for physical bodies

Meta plans to build its own hardware prototypes to test capabilities, but do not expect a Meta-branded robot on store shelves anytime soon. The company isn't trying to build a direct retail rival to Tesla's Optimus just yet.

Instead, Meta wants to supply the brains, sensors, and software operating system. Sources familiar with the project say the ambition is to mirror what Google's Android and Qualcomm's chips did for smartphones: build the foundational platform the rest of the robotics industry runs on. Meta is already pitching its open-source Llama model as the default brain for robotics researchers globally.

Early outreach has already started. Meta has been in discussions with several hardware makers, including Unitree Robotics and Figure AI, which recently cut ties with OpenAI after claiming large language models are becoming a commodity.

Meta believes it has a massive head start thanks to years of heavy spending on augmented and virtual reality. In an internal memo, Bosworth noted that core technologies developed for the Quest and Ray-Ban smart glasses—like hand tracking, low-bandwidth edge computing, and always-on sensors—map directly to humanoid robotics. More importantly, Meta has been quietly gathering first-person physical world data through efforts like Project Aria since 2022.

Moving from city streets to the living room

Current humanoids are still remarkably clumsy. None can reliably fold your laundry, carry a glass of water without spilling, or pack a dishwasher. Meta wants to tackle those domestic chores first.

While companies like Boston Dynamics focus on industrial warehouses and Tesla aims for manufacturing plants, Meta is aiming directly at the consumer home. That comes with unique headaches. An open road for an autonomous vehicle has standard lanes and signs; a family living room has stray toys, varied floor plans, and moving pets. While the safety stakes of a slow-moving bot are lower than a speeding robotaxi, the environmental unpredictability is arguably much higher.

Meta is budgeting roughly $65 billion this year across AI infrastructure and related hardware bets, giving Whitten plenty of runway to experiment.

Why it matters

For everyday consumers, don't hold your breath for a robot butler by Christmas. Commercial hardware costs remain astronomical, and reliable physical dexterity inside cluttered homes is still largely unsolved. This is firmly an R&D phase.

For the robotics industry, Meta's move could be a seismic shift. Building humanoid hardware is an expensive, capital-intensive manufacturing nightmare. If Meta successfully turns Llama and its sensor stack into a plug-and-play platform, it could dramatically lower the barrier of entry for dozens of struggling hardware makers who lack frontier AI talent.

In our view, this is one of Mark Zuckerberg's smartest portfolio plays yet. Reality Labs has long faced criticism for burning tens of billions on the metaverse. Re-channeling its spatial computing pipelines, sensor tech, and egocentric training data into physical AI gives those massive sunk costs a tangible second life. The catch? Software-only platforms often struggle when precision hardware integration is required. If generic models fail to handle delicate physical tasks across messy living rooms, Meta may eventually be forced to build the whole machine itself.