OpenAI Deepens Humanoid Robotics Bet

Bay Area Expansion

OpenAI expanded its robotics footprint in the Bay Area during the past week, reinforcing its push toward a humanoid platform at a time when the company is broadening physical-world AI work beyond software. The most recent reporting indicates the robotics effort has moved from a quiet internal project into a larger operational setup with a second site in Richmond, California, tied to development and testing for human-like robots.

Recent Reporting

The first clear public account of the expanded robotics lab appeared in January 2026, when Business Insider reported that OpenAI had built a humanoid robotics lab in San Francisco and was scaling it with about 100 data collectors working in shifts. That reporting also said the lab had more than quadrupled in size since its February 2025 start, making the current expansion a continuation of a program that began well before this week’s attention.

Humanoid Ambition

The company’s robotics strategy appears aimed at training systems that can handle household tasks, not just industrial motions. According to the reporting, workers are teaching a robotic arm to perform tasks such as moving objects and folding laundry, with cameras capturing both the operator and the robot to generate training data for physical manipulation. That approach suggests OpenAI is trying to build the data engine needed for a broader humanoid platform, where software models learn from repeated real-world interaction.

Richmond Facility

A major sign of momentum came from coverage on April 3, 2026, which said OpenAI had leased a 202,000-square-foot facility in Richmond on the east side of the San Francisco Bay. That report, citing local media, said the site would support robotics-related development and testing and referenced earlier indications that OpenAI intended to open a robotics lab there. The size and power needs of the building point to a hardware-heavy effort rather than a typical office expansion.

Production Demands

The lab’s operating model is unusually labor intensive. Business Insider reported that the San Francisco site runs three shifts, uses a few dozen workstations around the clock, and rates workers on how many “good hours” of training data they produce. The setup relies on teleoperated robots and data collection to improve robotic control, and the work is being done in the same building as OpenAI’s finance team.

Leadership Shift

The robotics push has also moved through leadership turbulence. A March 2026 analysis reported that robotics head Caitlin Kalinowski resigned on March 7, 2026, after objecting to the company’s Pentagon-related work. Even with that departure, the robotics effort appears to have continued, with reporting after the resignation still describing active lab operations and ongoing plans for Richmond expansion.

Industry Positioning

OpenAI’s renewed robotics effort arrives as rivals race to turn AI models into embodied systems. Reports in the past week also noted that OpenAI issued a request for proposals seeking U.S.-based manufacturers for consumer devices, robotics parts such as motors and actuators, and cloud infrastructure, signaling a broader supply-chain strategy around physical products. That would give OpenAI more control over hardware sourcing if its humanoid platform progresses toward commercial deployment.

Sector Signal

The company’s move matters because it reflects a shift in how frontier AI firms think about growth. Instead of treating robotics as a side project, OpenAI is building the infrastructure, workforce, and facilities needed to collect physical-world data at scale. That is a strong indicator that humanoid robotics has become a central strategic bet rather than a speculative experiment.