OpenAI CEO Sam Altman says the company is targeting a 2027 demonstration of its humanoid robot work. However, this date does not constitute a release schedule for ready-made robots operating on the streets or sold to consumers. As the growth in the humanoid robot market continues, Altman states that widespread use will take a few more years. The company has not yet announced the robot’s appearance, price and sales date.
Altman shared this assessment in an interview in September. Answering positively to the question about whether they are developing humanoid robots, the manager explains his 2027 expectations through a demo. For this reason, the statement should not be read as the official introduction of a new robot model. There is no product page or verified feature list from which the consumer can order.
The robot’s body and the software that manages it also constitute separate development topics. In Altman’s narrative, the company’s interest in systems that can operate in the physical world continues. However, the interview does not reveal the engines that will be used, the mobility or the payload that can be carried. Without such technical details, it is not possible to make a direct performance comparison with another robot.
OpenAI’s robotics work predates the humanoid body plan. The company shared its work on solving the Rubik’s cube with a robot hand on October 15, 2019. In that research, two neural networks guided a robot hardware that resembled a human hand. The researchers aimed to train the networks entirely in simulation and transfer what they learned to the physical robot.
OpenAI explores transition from simulation to physical world in robotics history
At the center of the study in 2019 was reinforcement learning and a method that gradually diversifies the educational environment. The team aimed to make the system adapt to different situations by changing conditions such as the size of the cube and the friction of the fingers. In addition, the company clearly explained at that time that the method did not produce perfect results in every attempt. This old experiment is not a test of today’s humanoid robot’s hardware or success rate.
The common research area between the two studies is software’s management of objects in the physical world. However, it does not follow from an experiment in the hands of a robot that a humanoid device will complete household chores or factory tasks. It is not yet clear which tasks the new system will perform. For this reason, it is necessary not to expand the 2027 target with the promise of a specific use.
The distinction between representation and common use in Altman’s explanation draws the most concrete boundary of the calendar. A demo can show a capability working under certain conditions, while sales and daily use require separate phases. On the other hand, OpenAI does not give regional access or Türkiye plans for the robot in this interview. The company’s next statement will reveal what it wants to show in 2027 and what stage the project has reached.