OpenAI’s Humanoid Robot Plan: What Sam Altman Confirmed—and What Remains Unknown
Sam Altman has publicly committed OpenAI to building humanoid robots. The company is recruiting across robotics hardware and manufacturing, but has not named a robot-making partner, unveiled a prototype, or confirmed that ChatGPT itself will serve as the robot’s operating system.
Sam Altman’s clearest statement so far came in a September interview with Sources. Asked whether OpenAI was pursuing a humanoid or a robotic data-center machine, he answered: “Both. We will definitely do a humanoid. We will do other form factors as well.” He argued that a human-shaped machine makes sense in spaces built around people, from doors and keyboards to tools and household work.1
That is a firm statement of intent, not a product announcement. OpenAI has not shown its own humanoid, announced a launch date, or identified a manufacturing partner. The distinction matters because the headline shorthand “ChatGPT-powered robot” can suggest a technical design that the company has not actually described.
A commitment to build, not a disclosed alliance
OpenAI’s public materials support the view that the company is building more than a software demo. Its robotics team says it works across hardware and software, integrating them to explore several robotic forms. A current manufacturing-engineering vacancy describes developing custom motors and actuators, building prototypes, and working with supplier teams to turn designs into repeatable hardware.2
That is evidence of real engineering and manufacturing work. It does not confirm an exclusive deal with a named robot manufacturer. Altman’s public comments confirm OpenAI intends to make humanoids; the company’s hiring points to work with suppliers and production specialists, but no factory partner or production plan has been publicly identified.1
There is an important historical footnote. In 2024, OpenAI partnered with Figure to develop AI for humanoid robots. Figure later ended that collaboration as it moved toward building its AI systems in-house. That earlier agreement is not proof of a new alliance for OpenAI’s present robot effort.7 8
Why build the body as well as the AI?
Altman’s explanation starts with the environment. Many everyday objects and workspaces are designed for people. A robot with a human-like body could, in principle, operate doors, handle tools, and move through spaces without every building being redesigned for a specialized machine.1
The harder question is what the robot can reliably do. A useful humanoid must interpret camera and sensor data, understand a task, plan a sequence of movements, coordinate motors and joints, and adapt when an object shifts or a grip fails. OpenAI has described its robotics effort as spanning AI and physical hardware, while its recruitment includes control, actuator, and manufacturing roles.2
That makes “ChatGPT in a robot” an incomplete description. A language model may help interpret requests or communicate with a person, but a working robot also needs systems for real-time perception, motion, force, and safety. OpenAI has not released the architecture for its own robot or said which models would run on it. Its stated commitment is to humanoid robots and other forms, not a confirmed consumer product built around the ChatGPT app.1
Developing its own platform could let OpenAI test intelligence and machinery together and collect examples of physical tasks. That may help the company improve how its systems act in the real world, but it remains a potential advantage, not a proven result. The company has not reported a deployed robot fleet or publicly demonstrated what its robot can do.5
What to watch next
Some recent coverage reports that Altman expects OpenAI to have a “cool demo” in 2027. The demonstration’s capabilities, audience, and hardware remain undisclosed, and a demo would not by itself establish when a reliable product could be manufactured at scale.4
The next meaningful evidence will be concrete: a public demonstration of tasks completed without constant human correction, details about the robot’s safety controls, and an explanation of how the system handles cameras, microphones, and data. Those questions become particularly significant if the company eventually pursues personal robots for homes, an ambition Altman has expressed but not turned into a consumer roadmap.3
Closing Thoughts
The news is not that ChatGPT has already been placed inside a household robot. The notable change is that OpenAI’s CEO has made humanoid development an explicit company goal, while the company is staffing up in robotics hardware and manufacturing. The direction is clear; the machine, partners, and timeline are not.
How does this play out in reality?
In the near term, it likely looks less like a robot arriving in every home and more like engineering teams testing motors, actuators, control systems, and AI on physical machines. Industrial and infrastructure tasks could offer a narrower place to learn before attempting unpredictable household work. That is a reasonable interpretation of the hiring and public remarks, not a confirmed OpenAI deployment plan.2
Why is this noteworthy?
OpenAI has become known for software, yet a humanoid program would bring the company into a field where intelligence has to function through sensors, moving parts, power limits, and the risks of sharing space with people. Its progress will be judged not by a fluent conversation, but by whether a machine can carry out useful tasks safely, consistently, and at a cost that makes sense. For now, Altman has confirmed the ambition; the public evidence of a working OpenAI humanoid is still ahead.
Video references
Sources Podcast, September 1, 2026 — the segment on humanoid robots begins at 58:22. Watch the full interview
Bloomberg Originals, June 12, 2025 — an earlier discussion of humanoid robots and AI’s effect on work. Watch the interview
References
Comments