Source: Fortune exclusive by Wen Shao published November 13, 2026.
The dissolution ends OpenAI's second attempt at an autonomous robotics group: the company previously wound down its original robotics unit in 2020 before rebuilding one in 2024 around a short-lived partnership with Figure that terminated in February 2025. OpenAI told Fortune the latest shuffle represents deeper integration rather than retreat, pulling embodied experiments onto the same core training pipeline powering coding and reasoning products.
Economically, consolidating hardware and embodied researchers under Chen and Brockman arrives amid heightened scrutiny over compute allocation following recent operating losses reported at $10.9 billion in a single quarter. Over the past year CEO Sam Altman urged staff to cut side quests—culminating in shutting down Sora—so frontier compute stays concentrated on core revenue-generating intelligence.
Carnegie Mellon University Robotics Institute assistant professor Guanya Shi called the move a clear signal that OpenAI views physical manipulation as 'no longer a special problem' but a downstream application of general spatial reasoning. However, Shi emphasized that large models remain slow and tethered to cloud inference, arguing smaller specialized models will still be required onboard edge controllers without reliable network links.
Boston-based MIT professor and Walden Robotics CEO Russ Tedrake noted a complementary moat for industrial robotics firms: physical fleets gather deployment experience and tactile feedback inside live factories that foundation labs cannot scrape online, making better general models amplify rather than erase field data value. Generalist cofounder Pete Florence cautioned it remains unclear which starting paradigm will ultimately govern physical intelligence.