OpenAI announced on September 8, 2026, that an artificial intelligence model solved the Navier-Stokes existence and smoothness problem, one of the Millennium Prize Problems. The proof required 88 hours of computation by 10,000 parallel AI agents, sparking an immediate academic dispute over credit and methodology.
The announcement represents what OpenAI researchers describe as a major milestone for advanced mathematical research. Formulated originally in the nineteenth century, the Navier-Stokes equations describe how fluids such as water and air move through space. They govern everything from atmospheric weather patterns to the flow of blood through the human vascular system and the aerodynamics of aircraft design. For generations, mathematicians have sought to determine whether these equations remain smooth indefinitely or whether an initially regular fluid can develop a singularity, meaning a point where its velocity grows without limit in a finite amount of time.
The 88-Hour Computation and the Lean Formalization
OpenAI researcher Sebastian Bubeck told reporters that the company initiated training on a new, unreleased AI model on August 28, 2026. After hearing unconfirmed reports that competing researchers were making headway on fluid-motion puzzles, the company decided to dedicate significantly more computing resources to the task.
To tackle the problem, the system deployed approximately 10,000 autonomous AI agents working in parallel over 88 hours, generating a 165-page document. According to technical descriptions provided by the company, an internal system produced an analytical demonstration and a formalization using Lean, a programming language designed for verifying mathematical proofs. The computation consumed millions of dollars in computing power and generated 2.7 million messages.
Ven Chandrasekaran, a computer scientist at OpenAI, explained the core finding during a press briefing.
Ven Chandrasekaran, computer scientist at OpenAI, explained that their proof demonstrates the existence of fluids that begin perfectly normal and ultimately reach infinite speed within a finite amount of time under the Navier-Stokes equations.
Because infinite velocity is physically impossible for real-world liquids or gases, Chandrasekaran noted that the result suggests the equations may fail to reflect physical reality under extreme circumstances. The solution relies on a contracting, stretching vortex subjected to a smooth external force, preserving a finite total energy throughout the collapse while allowing acceleration and viscosity terms to scale upward simultaneously.
Controversy Over Timing and Rival Research Teams
The timing of the announcement triggered a sharp dispute over precedence and collaboration. Just hours before OpenAI made its findings public, Tristan Buckmaster, a mathematician at New York University, and Levent Alpöge, a researcher at Anthropic, published independent work addressing related fluid-dynamics equations.
Buckmaster asserted in an accompanying statement that OpenAI pivoted its resources toward the Millennium Problem only after learning of his and Alpöge’s ongoing research, which utilized AI models from both Anthropic and OpenAI.
OpenAI executives denied having direct access to the rival team’s unreleased prompt logs or working papers prior to publication. Mark Chen, director of research at OpenAI, stated that the computing power required for their internal model cost millions of dollars. Company representatives also confirmed that they do not intend to claim the US$1 million prize associated with the solution.
The Clay Mathematics Institute and Future Scrutiny
Established in the year 2000, the Clay Mathematics Institute selected seven fundamental open questions in mathematics, offering US$1 million for each correct resolution. Only one of these challenges, the Poincaré conjecture solved by Russian mathematician Grigori Perelman, has been officially classified as resolved.
Martin Bridson, president of the Clay Mathematics Institute, acknowledged the significance of the moment, describing it as an exciting day while contemplating the announcement of major advances in the human understanding of mathematics. Nevertheless, the institute officially maintains Navier-Stokes on its active roster of unresolved problems pending formal independent verification by the academic community.
Independent experts have also raised broader questions regarding the role of automated reasoning in mathematics.
