Signal

OpenAI disbands its standalone robotics team, moving its staff to research and Applied units, saying models like GPT-6 Astra can be directly applied to robotics

First reported by Fortune ·

The signal ●●●○ Compiled by AI from Fortune and Techmeme
Why you might care

Large AI models can now control robots, removing the need for specialized robotics AI and changing how robot brains are built.

What happened

OpenAI has disbanded its dedicated robotics team, reassigning its members to its core research and applied AI divisions. The company stated this move is intended to accelerate progress in robotics by leveraging its general-purpose AI models, such as GPT-6 Astra, to directly control robots. This shift follows the end of a partnership with Figure AI in February 2025 and OpenAI's ambition to demonstrate its own humanoid robot by 2027. The decision reflects a broader industry trend of applying large language and vision models to "embodied AI," potentially negating the need for specialized robotics AI. OpenAI's CEO, Sam Altman, has been pushing the company to focus on core products and reduce costs, with this reorganization aligning with that strategy. Despite the disbandment, OpenAI maintains that robotics is central to its long-term vision.

What it means

This reorganization signals a significant bet that foundational AI models, already adept at language and vision, can inherently possess the capabilities needed for physical world interaction, thereby democratizing robotics development. Companies previously focused on specialized robotics AI may need to pivot or risk becoming obsolete as general-purpose models become more versatile and cost-effective. The success of this approach could drastically lower the barrier to entry for robotics innovation, shifting focus from bespoke AI engineering to the integration and refinement of existing powerful models.

The move also suggests a potential consolidation in the AI hardware and software sectors, favoring entities capable of developing and deploying large-scale foundational models. For robotics startups, the challenge will be to differentiate by showcasing real-world deployment advantages and data acquisition capabilities, as general AI models may lack the nuanced, on-site learning that specialized robots currently possess. The industry will be watching closely to see if OpenAI's integrated approach can truly deliver on the promise of adaptable, general-purpose robots without extensive task-specific training.

AI-written summary. May contain errors.

AI Robotics