OpenAI is moving beyond computer screens and digital interfaces into physical space. CEO Sam Altman has confirmed that the artificial intelligence (AI) pioneer is developing its own humanoid robots, aiming to showcase an early working prototype as soon as 2027.

The initiative marks a major strategic expansion for OpenAI, transitioning the firm from software models into hardware design, mechanical engineering, and manufacturing. Speaking on the Sources podcast and in a separate interview with Fortune, Altman outlined the firm’s ambition to take AI out of the server rack and into real-world environments.

“I think we’ll have a cool demo in 2027,” Altman said. “I don’t think there will be robots walking the streets for a few years longer, but we’ll have something impressive.”

OpenAI’s initial deployment strategy focuses on industrial applications, particularly facilities construction and data center management. However, the company’s long-term vision encompasses multipurpose machines capable of assisting with household chores and operating within human-centric environments.

Designing humanoids allows machines to navigate spaces built for human anatomy such as doorways, stairs, and narrow corridors without requiring structural modifications to existing infrastructure.

Unlike conversational AI models that merely process text inputs, physical robotics demand real-time spatial navigation, object recognition, situational decision-making, and safety protocols to ensure safe co-existence with humans.

This is not OpenAI’s first foray into physical systems.

In 2018, the firm debuted Dactyl, an AI system that controlled a robotic hand capable of complex manipulation. By 2019, OpenAI demonstrated a robotic hand solving a Rubik’s Cube using simulated AI training models. The specialized robotics research unit was subsequently disbanded, but momentum returned in early 2025 as the company quietly resumed its hardware endeavors.

Recent job listings on OpenAI’s careers platform confirm an aggressive recruitment drive across mechanical engineering, control systems, electronics, and hardware safety. The firm is actively seeking specialists to engineer motors and actuators, the core components that convert digital commands into mechanical movement, indicating plans to develop hardware and software concurrently rather than relying on third-party manufacturers.

By manufacturing its own hardware, OpenAI enters an increasingly crowded and well-funded sector. Competition includes Tesla Inc., which continues development on its Optimus humanoid, as well as robotics startup Figure AI Inc., which secured over $1 billion in funding at a $39 billion valuation in September 2025.

Notably, OpenAI previously backed Figure and Norwegian startup 1X through investments and software partnerships. Its decision to build proprietary hardware signals a shift toward direct competition with former partners.

While the 2027 target represents a prototype demonstration rather than a consumer product launch, the move marks a significant step toward autonomous physical labor. If successful, OpenAI’s prototype will provide the tech industry with its first clear view of how advanced AI models operate when embodied in the physical world.