OpenAI works on humanoid robot, Sam Altman targets a 2027 demo

OpenAI is building a humanoid robot meant to operate in spaces designed for people, according to remarks by chief executive Sam Altman reported on October 18, 2026. Altman said the company plans to show an impressive demo in 2027, while cautioning that robots moving freely through everyday life will take considerably longer.

Altman argued that a human-like body is a practical choice. Doors, vehicles and keyboards are all built around human proportions, so a machine with a similar shape should fit more naturally into that world. Even so, the company says the hardware form is secondary to the software that controls it.

The brain comes first

In Altman's view, attaching arms and legs to a machine achieves little if the machine has no idea what to do with them. The robot has to read its surroundings through cameras and sensors, interpret instructions and coordinate its movements. Something as ordinary as picking up a coffee mug requires recognising the object, planning a grip and applying the right amount of force.

To tackle these problems, OpenAI is recruiting specialists in motion control, state estimation and actuator control, with 27 robotics roles currently open on its careers page. The listings describe teams working on custom electromechanical actuators, motors and sensor systems, signalling that the company is investing in hardware as well as software.

Robotics is not new territory for OpenAI. Its earlier Dactyl project trained a robotic hand to manipulate objects with dexterity. The physical world, however, is far messier than the rules software learns in simulation, and OpenAI says it is building machine learning infrastructure capable of handling that complexity.

Goals and competition

The near-term aim is general-purpose robots that help skilled workers build infrastructure, while the longer-term vision is a personal robot able to handle a range of daily tasks. OpenAI has not officially disclosed what the 2027 demo will include, the robot's technical specifications, a consumer price, a release date or a privacy framework.

The company is entering a crowded field. Tesla's Optimus project and XPENG's IRON model show competition in humanoid robotics is intensifying. For a firm best known for large language models, the move underlines how AI developers are pushing beyond software into machines that act in the physical world.