OpenAI is developing a humanoid robot, according to comments by CEO Sam Altman, who believes that designing the body is not the biggest challenge compared with building the intelligent system capable of operating it. The company has not yet announced an actual model or a commercial launch date, but it is targeting a demonstration in 2027.
Why did the company choose a humanoid form?
Altman believes that a humanoid form makes sense because the world around us was designed for humans; doors, keyboards, and tools assume specific dimensions and ways of moving. Replicating the human form could therefore help the robot work in the same spaces without requiring the surrounding environment to be redesigned.
However, Altman focuses on what he calls the robot’s “mind.” Having two arms and two legs is not enough unless the system can understand what the cameras and sensors capture, interpret instructions, choose how to carry them out, coordinate movement, and then modify its behavior when things do not go as expected.
Embodied intelligence is more difficult than software control
Simple tasks such as picking up a cup illustrate the scale of the problem. The robot must identify the cup and its handle, estimate where to grasp it and the appropriate gripping force, and then lift it without dropping it. This directly connects artificial intelligence models to the constraints of physics, where it is not enough for the system to learn the theoretically expected result.
OpenAI has previous experience in this area through the Dactyl project, which trained a robotic hand to handle real objects using training in simulation. The company also applied related techniques to a Rubik’s Cube, in an example that reflects the gap between what a model can learn in software and what movement, friction, and weight permit in the real world.
The body itself remains a major engineering project
Altman’s focus on intelligence does not mean that OpenAI treats hardware as secondary. Job listings reveal work on custom electromechanical actuators, including motors, transmissions, sensing, and thermal systems, along with positions for manufacturing and the supply chains needed for large-scale production.
The company currently has 27 robotics-related positions, covering motion control, state estimation, actuator control, and full-robot behavior, as well as infrastructure for training physical artificial intelligence systems.
What is changing in practical terms?
The published positions indicate that the near-term goal is to develop general-purpose robots that help skilled workers build infrastructure, while the longer-term ambition is a personal robot capable of carrying out a broader range of tasks. OpenAI is competing in a field in which Tesla is working on the Optimus robot, while XPENG has demonstrated the IRON robot walking autonomously after leaving the production line.
Nevertheless, the project is still at an early stage. OpenAI has not revealed the robot’s appearance or what the 2027 model will demonstrate, nor has it announced a price, subscription model, release date, or a dedicated privacy framework for a robot equipped with cameras and microphones inside homes. Therefore, the current announcement concerns a research and hiring direction more than the launch of a ready-to-use product.