OpenAI claims it solved a 90-Year-Old maths problem, but not everyone is convinced
Key takeaways
- OpenAI claims to have solved the Navier-Stokes existence and smoothness problem, one of seven Millennium Prize Problems each worth one million dollars
- The result was announced via blog post before any peer review, drawing criticism from the formal mathematics community
- The Clay Mathematics Institute has not commented or indicated any prize will be awarded
- If independently verified, this would be one of the most significant scientific achievements of the 21st century
OpenAI says it has found a solution to the Navier-Stokes problem, one of the seven Millennium Prize Problems in mathematics and one that has resisted every serious attempt for roughly 90 years. The claim was first reported by The New York Times and Wired before OpenAI published a blog post on Tuesday. It has sent the maths and AI communities into a frenzy, and not entirely in the way OpenAI might have hoped.
According to OpenAI, its model, built on work from its internal research team, produced a proof that satisfies the existence and smoothness conditions in three-dimensional space. That is the full claim. It is enormous, and so far nobody outside the company has checked it.
What is the Navier-Stokes problem?
The Navier-Stokes equations describe how fluids move, whether that is water flowing through a pipe, air swirling around an aeroplane wing, or blood moving through arteries. Engineers use them every day to model weather, aircraft design and pipelines, usually with computer approximations rather than exact answers.
The open question is a different one. Nobody has proved that smooth, well-behaved solutions to these equations always exist in three dimensions. A solution could, in principle, "blow up", meaning a quantity such as velocity becomes infinite in finite time. The question has been open since 1934, and it is the reason the problem is called the existence and smoothness problem.
The Clay Mathematics Institute lists it as one of the Millennium Prize Problems, each carrying a $1 million reward for a verified solution. The institute announced the list in 2000, and only one of the seven, the Poincare conjecture, has been solved. Grigori Perelman, who solved it, declined the prize money.
At a glance
| Item | Detail |
|---|---|
| Problem | Navier-Stokes existence and smoothness, in three dimensions |
| Open since | 1934, roughly 90 years |
| Prize | $1 million from the Clay Mathematics Institute |
| Millennium Prize Problems | Seven in total, one solved so far |
| Claimed by | OpenAI, via a blog post on Tuesday |
| Peer review | None yet |
| Clay Mathematics Institute | Has not commented |
Why does smoothness matter?
A smooth solution is one with no sudden jumps or infinities, so the equations keep giving sensible answers as time passes. If a proof showed that smooth solutions always exist, it would confirm that the maths behind fluid flow never breaks down. If someone found a case where it does, that would be just as important, because it would show that the equations have limits as a model of nature.
Either outcome would be a landmark. That is why the problem has attracted so much effort, and why a claim of success needs such careful checking before anyone celebrates.
So why is there drama?
Mathematics is not a press release. The moment OpenAI made the claim public without first submitting it to peer review, the formal mathematics community pushed back hard. Several prominent mathematicians, on social media and in public statements, noted that extraordinary claims need verified, reproducible proofs checked by domain experts, not blog posts.
A few also pointed out that AI-generated proofs have a history of containing subtle errors. They look convincing, then collapse under scrutiny. A single flawed step in a proof of this length can invalidate everything built on top of it.
The Clay Mathematics Institute has not commented. Until an independent verification process is completed, the $1 million prize is not being handed to anyone.
Why does the way OpenAI announced it matter?
Announcing a result like this the way a tech company announces a product feature, with a blog post and press coverage seeded before the academic community could weigh in, is exactly the kind of thing that erodes trust between the AI industry and the scientific establishment. It is not how maths works, and OpenAI almost certainly knows that.
The normal route is slow by design. Authors post a preprint, specialists read it line by line, errors are found and fixed, and eventually a journal publishes it. For a Millennium Prize Problem the Clay Institute also requires publication and a waiting period of two years before it considers an award, according to its published rules. A blog post skips every one of those checks.
There is also a strategic read. OpenAI has been under heavy pressure in 2026, with Anthropic closing the gap on benchmarks, Cognition hitting a $48 billion valuation in the AI coding space, and the company navigating multiple legal challenges. A headline about solving an unsolvable maths problem is, to put it plainly, excellent for the share price and for recruiting. For more on how the two rivals compare commercially, see Anthropic vs OpenAI and the IPO race.
Could the proof still be real?
That does not mean the result is fake. OpenAI has serious researchers, and AI-assisted theorem proving has been advancing rapidly through tools such as Lean and other formal verification frameworks. A formal verifier like Lean checks every logical step by machine, which removes the risk of a human reader missing a flaw. If the proof can be written in that form, the argument about whether it is correct gets much shorter.
It is genuinely possible something real is here. OpenAI has already been in the news for AI maths results, including Astra solving ten open maths problems in a single day. Not every one of those claims has been welcomed, and the row with the profession is covered in the mathematicians' feud with OpenAI.
But the honest answer right now is that nobody outside the team has verified the Navier-Stokes claim. The drama swirling around the announcement is largely self-inflicted.
What happens next?
If OpenAI submits the proof to a formal journal and independent mathematicians confirm it, this becomes the story of the decade. If the proof contains errors, the reputational cost will be significant, and it will give critics of AI-generated research a fresh example.
Even getting close to a result like this using AI-assisted reasoning would represent a meaningful shift in how mathematics can be done. If the proof survives peer review, it would be one of the most significant scientific achievements of the century, and not just for AI, but for human knowledge.
For readers, the practical test is simple. Look for three things over the coming weeks: a public preprint with full details, comments from specialists in fluid dynamics and partial differential equations, and any statement from the Clay Mathematics Institute. Until at least some of those appear, treat the claim as a claim.
There is a related story on the agent side of the work in ten thousand AI agents and a fight over who got there first, which is worth reading alongside this one.
