TLDRs;
- OpenAI is hiring humanoid robotics experts as it reboots its push into physical AI systems.
- New hires, including Stanford’s Chengshu Li, signal a focus on household-task-capable humanoid designs.
- Job listings show interest in teleoperation, simulation, and even mass production of robotic systems.
- Competition is fierce, with Tesla, Google, and humanoid startups already shaping a $5T future industry.
OpenAI, is intensifying its focus on robotics. The San Francisco-based company is quietly assembling a team of experts skilled in humanoid systems and advanced AI algorithms, signaling that the firm sees physical-world intelligence as critical to its long-term mission of achieving artificial general intelligence (AGI).
Job postings and recent hires reveal that OpenAI is actively recruiting specialists in areas such as teleoperation, simulation, and mechanical design.
These fields are essential for training robots to navigate real-world environments, learn from human demonstrations, and ultimately carry out complex physical tasks.
A Wave of Key Hires
Among the most notable additions is Chengshu Li, a robotics researcher who joined OpenAI in June 2025 after completing his Ph.D. at Stanford University.
Li’s work focused on benchmarking humanoid robots for household tasks, with designs that blended humanlike upper bodies with wheeled mobility systems. His expertise underscores OpenAI’s growing interest in humanoid robotics as a test bed for developing more adaptable and general-purpose AI.
Other researchers from prominent robotics labs have also recently signed on. Although OpenAI has declined to comment on its recruitment efforts, the cluster of new hires indicates a renewed commitment to robotics research, an area the company had scaled back in 2021 to focus on large-scale language models.
Robotics as a Path to AGI
The strategic pivot suggests that OpenAI believes achieving AGI will require more than mastering digital language and image tasks. Robots provide an avenue for AI systems to perceive, interpret, and act within the physical world, a leap that many researchers argue is necessary for true intelligence.
Recent job listings point to projects involving teleoperation, where humans guide robots through tasks while algorithms learn to replicate their motions. The use of simulation platforms such as Nvidia Isaac is also mentioned, enabling AI models to train in realistic virtual environments before being transferred to physical machines.
Notably, one opening called for a mechanical engineer with mass-production expertise, hinting that OpenAI may eventually design its own hardware rather than relying exclusively on third-party robotics companies.
A Crowded and Competitive Field
OpenAI is entering a space already buzzing with activity. Startups like Figure, Agility, and Apptronik are advancing humanoid designs, while tech giants such as Tesla and Google have launched their own robotics initiatives.
Billions of dollars have flowed into the humanoid sector since 2024, with Morgan Stanley predicting that humanoids could represent a $5 trillion market by 2050.
Still, humanoid robotics remains a frontier filled with hurdles. Robots can now perform carefully choreographed tasks such as walking, lifting, or even dancing, but they still lack the intelligence to handle unpredictable environments. To succeed, AI must not only generate text or images but also control limbs, grasp objects, and respond to dynamic physical challenges.
