OpenAI CEO Sam Altman has confirmed that the company intends to develop a humanoid robot alongside machines designed for other specialized roles. (Shutterstock)OpenAI intends to develop a humanoid robot as well as purpose-built machines for specialized environments, CEO Sam Altman confirmed during a wide-ranging Sources Podcast interview. His response to a question about humanoids and machines designed for data centers indicated that the company’s robotics plans encompass both categories, along with additional physical configurations. The exchange offers the clearest public indication so far that OpenAI aims to develop robot bodies alongside the artificial intelligence systems that control them.
Altman’s remarks establish the company’s direction, not the arrival of a market-ready machine. The episode, published by Sources covered OpenAI’s broader strategy as well as robotics.
Altman’s comments presented the human form as a practical choice because homes, workplaces, tools and infrastructure are largely designed around human movement. A robot with a familiar body shape may be better suited to using doors, keyboards, vehicles and kitchens, while specialized environments such as data centers could require different designs. The interview also placed greater emphasis on the challenge of developing intelligence capable of understanding physical surroundings, making decisions and acting reliably.
The company’s recruitment activity adds substance to Altman’s comments. OpenAI describes its Robotics team as working across hardware and software to develop general-purpose robotic systems in a range of physical configurations. Current openings cover electrical engineering, firmware, thermal simulation, actuator development, safety, data acquisition, field engineering and robot-learning infrastructure. Those positions point to a program extending well beyond the development of a model that could be licensed to outside manufacturers.
Several vacancies focus on components that would sit inside a robot’s body. An actuator gear design position involves custom gearing, motors, sensing, thermal architecture and the transition from prototypes to production-ready assemblies. OpenAI is also seeking an electrical engineer for robotics to design compact electronics, power distribution, communications systems and actuator interfaces. The descriptions do not identify a humanoid platform, but they show that OpenAI is building capabilities needed to create integrated electromechanical systems.
Developing the robot’s intelligence presents a separate set of problems. A physical system must interpret visual information, understand an instruction, plan a sequence of movements and adjust when the environment changes. OpenAI has previously described video-generation models as a path toward systems that understand and simulate the physical world. Its research accompanying the original Sora model examined how large-scale video training could help models represent objects, motion and interactions over time. More recently, OpenAI said such world-simulation capabilities could be important for AI systems required to operate in real environments, although video prediction alone does not provide the reliability or physical control needed by an autonomous robot.
OpenAI would enter a field already attracting substantial investment from AI laboratories, automakers and robotics startups. Figure AI, one of the prominent U.S. humanoid developers, announced in September 2025 that it had secured more than $1 billion at a $39 billion post-money valuation. Figure is developing its own robot bodies and a vision-language-action system called Helix, which connects perception and language processing with physical movement. Its decision to integrate the intelligence and hardware illustrates how control of both layers has become a strategic priority for some robotics companies.
Commercial progress across the humanoid robotics industry remains uneven. Completing a rehearsed task in a controlled setting is different from operating reliably in a workplace where objects move, layouts change and unexpected interruptions occur. Commercial viability will depend on whether these machines can perform routine jobs repeatedly and safely without constant human assistance. Precision, adaptability and dependable autonomous operation are therefore likely to matter more than isolated demonstrations of speed or movement.
Moving into robotics could give OpenAI a direct source of physical interaction data while allowing its engineers to develop software and hardware together.
It would also expose the company to costs and operational challenges that differ sharply from running cloud-based AI services. Actuators, sensors, batteries, supply chains, safety testing and manufacturing quality all become part of the equation once an AI company builds machines that move in the physical world.
Altman has now confirmed the intended destination, while OpenAI’s job postings show that the necessary technical foundation is being assembled. What remains unknown is when those efforts will produce a working humanoid and whether OpenAI ultimately plans to manufacture one at scale, deploy it internally or work with outside partners.
