AI achieves breakthrough on long-standing fluid dynamics problem
OpenAI reports that its AI model has produced a solution to the Navier-Stokes equations, a problem that has resisted mathematicians for two centuries. The work involved deploying thousands of AI agents, with the final step taking only 11 hours. The company estimates the computational cost at $15 million.
The Navier-Stokes equations, formulated roughly two centuries ago, describe fluid motion. The Millennium Prize challenge asks whether these equations always hold or eventually produce nonsensical "blow-ups". OpenAI's approach involved 1,000 AI agents working 50 hours on the related Euler problem, followed by 10,000 agents extending those findings to Navier-Stokes in just 11 hours.
This achievement follows a string of AI-driven mathematical wins, including cracking an Erdős conjecture and formalizing Fermat's last theorem. However, the announcement sparked friction, as NYU's Tristan Buckmaster and Anthropic's Levent Alpöge had released related stepping-stone results hours earlier, and Buckmaster later criticized OpenAI's opaque handling of his clarification requests.
This breakthrough could reshape mathematical research, potentially shifting focus from human proof to AI-driven discovery. Engineers relying on fluid simulations for aircraft design, weather forecasting, and climate modeling may gain new tools to predict turbulent flows, though the $15 million cost and opaque methodology could limit immediate accessibility. The dispute between OpenAI and academic researchers may also prompt debates about credit, transparency, and the future role of human mathematicians in verifying AI-generated proofs.