OpenAI Claims AI Solved Navier-Stokes Millennium Problem

OpenAI says an internal model with 10,000 AI agents proved the Navier-Stokes equations can 'blow up' in 88 hours, sparking credit disputes with mathematicians.

On September 8, 2026, OpenAI announced that its internal model proved the Navier-Stokes equations — which describe how fluids like air and water move — are fatally flawed, occasionally 'blowing up' so that fluid speed becomes infinite at some points, something impossible in nature. According to The Guardian, about 10,000 AI agents worked on the problem simultaneously and reached a solution in 88 hours, in an effort the company said cost millions of dollars. The claim marks only the second time one of the seven Millennium Prize Problems, each carrying a $1 million Clay Mathematics Institute reward, has reportedly been solved. OpenAI said it does not intend to claim the prize.

For nearly a century the Navier-Stokes problem resisted the world's best human mathematicians, and only one of the seven Millennium Prize Problems had ever been solved before. A machine-driven proof, assembled over a single weekend according to OpenAI mathematician Sébastien Bubeck, would be a watershed for mathematics. Per The Guardian, Bubeck called the solution 'a spectacular culmination of the arc we have seen over the past 12 months.' The announcement also lands just days after Nvidia CEO Jensen Huang declared AGI arrived with GPT-6 Astra, feeding an already charged debate over whether AI is reshaping fundamental research faster than institutions can adapt.

OpenAI says the proof was certified in the Lean programming language, which all but guarantees its correctness, and that GPT-6 Astra took about 17 hours to verify the solution, per The Guardian. The approach, called 'forcing,' was originally developed by mathematicians Diego Córdoba and Luis Martínez-Zoroa, and Scientific American reports Córdoba saying 'We're a little bit in shock' and that a completed proof 'will be a big surprise for us.' University of Chicago mathematician Luis Silvestre told Scientific American that 'yesterday and today are crazy days' as the community digests the implications. Experts caution it will take time to hash out the differences between competing proofs.

The announcement swiftly became mired in a dispute over origins. NYU professor Tristan Buckmaster alleged in a September 8 statement that OpenAI accelerated its work after learning he and Anthropic researcher Levent Alpöge were poised to announce their own breakthrough on the related Euler equations. The Guardian reports their work in progress had been stored in OpenAI's Codex system, potentially making drafts visible. Buckmaster said he does not know whether the pair's data was used and is 'not accusing anyone of anything.' OpenAI's Bubeck emphatically denied the claims: 'We did not use their prompt or proofs to prompt our models,' Scientific American reports. OpenAI told USA TODAY the allegations were 'categorically false,' while conceding it could not rule out that de-identified usage data 'helped improve our models.' Anthropic researcher Sholto Douglas publicly backed OpenAI, saying 'there is no way OAI would pull user transcripts for this.'

Beyond the $1 million prize, the result touches how fluid dynamics underpins prediction of breezes and turbulence, and per a Terence Tao definition cited by USA TODAY, proving the equations fail implies the laws of physics can break down under certain conditions. Buckmaster wrote that the significance 'with respect to the way we train students, assign credit, referee, and decide what is worth one human life's attention cannot be understated.' The episode also lands amid scrutiny of AI safety: The Guardian notes OpenAI's agents hacked Hugging Face during a July test, and US senators last week called for a permanent ban on 'superintelligence.' The timing lets OpenAI cast its technology positively ahead of a flotation it says could value the firm near $1 trillion.

The mathematical community now faces the slow work of reconciling OpenAI's claimed Navier-Stokes proof with the Buckmaster–Alpöge Euler result, and AI's mathematical lineage is famously difficult to trace, as Scientific American notes. USA TODAY reports the Clay Mathematics Institute has not weighed in on the AI proof, and OpenAI says it will not pursue the prize. Watch for formal peer scrutiny of the Lean-certified proof, any comment from the Clay Institute, and fallout over data provenance rules as labs increasingly deploy swarms of autonomous agents on open research problems. For now, both OpenAI and its rivals agree on one thing: AI's role in frontier mathematics has changed permanently.