OpenAI's Robotics Team Disbanded Amid Shift to General AI Models

OpenAI has recently disbanded its dedicated robotics team, reallocating its researchers to broader projects within the company. This move signals a strategic shift in how OpenAI approaches robotics, emphasizing the integration of general-purpose AI models into robotic applications rather than maintaining a separate robotics division. The company maintains that robotics remains a core aspect of its long-term vision, suggesting this reorganization is not a retreat but a recalibration of focus.

The reorganization comes at a critical juncture for OpenAI, which faces mounting pressure to streamline operations and reduce costs. In the last quarter, the company reported significant operating losses of $10.9 billion. As OpenAI prepares for a potential IPO in early 2027, CEO Sam Altman has urged the company to concentrate on its core products, leading to the discontinuation of several projects, including the Sora video generation model. By merging the robotics team into its main research and applied divisions, OpenAI aims to enhance efficiency and leverage its advancements in AI technology.

Historically, OpenAI's foray into robotics has been fraught with challenges. The company initially disbanded its robotics team in 2020, citing insufficient data and a strategic pivot towards developing artificial general intelligence (AGI) through large language models. A brief revival occurred in 2024 with a partnership aimed at humanoid robotics, but that collaboration ended in February 2025, with Figure opting to develop its own AI solutions. OpenAI is now reportedly working on its humanoid robot, with a public demonstration slated for 2027.

The shift in strategy reflects a growing belief within the AI industry that general-purpose models can effectively manage robotic tasks, potentially eliminating the need for specialized robotics teams. OpenAI's recent experiments with its GPT-6 Astra model have shown promising results, with the model successfully controlling robotic functions in simulated environments. This approach challenges the traditional view that robotics requires tailored AI solutions, as many startups have raised substantial funding based on the premise that robots need dedicated intelligence.

Experts are divided on the implications of this shift. Guanya Shi from Carnegie Mellon University interprets OpenAI's reorganization as a recognition that robotics is no longer a distinct challenge but rather an application of general intelligence. Conversely, Russ Tedrake from MIT emphasizes that robotics companies possess a unique advantage through real-world deployment experience, which allows them to gather valuable data and adapt their robots to specific environments and tasks. The insights gained from these deployments could enhance the effectiveness of general-purpose models when applied in practical settings.

Despite the potential of general AI models, challenges remain. Large language models, while powerful, require constant internet connectivity and can be slow, which may necessitate the continued development of specialized models capable of operating independently on robotic hardware. The future of robotics may involve a hybrid approach, where general models are complemented by smaller, specialized systems that can function autonomously.

OpenAI's decision to integrate its robotics researchers into broader teams reflects a strategic pivot that could reshape the competitive landscape in robotics. As the capabilities of general-purpose models expand, the traditional barriers between AI and robotics may blur, leading to new opportunities and challenges. Companies in the robotics sector must adapt to this evolving paradigm, reconsidering their reliance on specialized models and exploring how they can leverage advancements in general AI to enhance their offerings. The implications of this shift could redefine the future of robotics, as firms that successfully integrate general intelligence into their operations may gain a significant competitive edge.