OpenAI Moves Into Humanoid Robots

OpenAI is preparing to move beyond software and into the physical world. CEO Sam Altman has confirmed that the company intends to build humanoid robots, as well as machines in other form factors, signalling a major expansion of OpenAI’s ambitions beyond generative AI models and digital agents.

For OpenAI, the bigger challenge is not necessarily building the robotic body, but developing the AI “brain” capable of controlling it. Altman argues that intelligence—enabling robots to understand environments, reason, learn, manipulate objects and adapt to unfamiliar situations—will ultimately determine whether general-purpose robots become commercially useful.

The strategic significance is enormous. Today’s AI largely operates inside computers, answering questions, generating content and executing digital workflows. Humanoid robots could allow advanced AI to perform physical tasks across manufacturing, warehouses, healthcare, retail, data centres and eventually homes. This represents the convergence of generative AI, computer vision, robotics and autonomous agents.

Building its own robots could also give OpenAI access to valuable real-world training data. Instead of depending entirely on external robotics companies, OpenAI could create a feedback loop in which robots continuously generate data about movement, manipulation and interaction with physical environments—potentially accelerating the development of physical intelligence.

However, robotics introduces challenges far beyond software: actuators, batteries, sensors, manufacturing, reliability, safety and complex supply chains. OpenAI will also enter an increasingly competitive market involving companies developing both humanoid hardware and robotic intelligence. The larger shift is unmistakable: AI is evolving from generating answers and operating software to perceiving, deciding and acting in the physical world. That could make humanoid robotics one of AI’s next major technology battlegrounds.