Manifold traders just priced a near-certain bet that at least two of the seven Millennium Prize Problems will be solved by 2040. A market asking exactly that sits at 99% implied probability of YES as of September 2026, and trading was so intense that the equivalent of 106% of the market’s entire historical volume changed hands in a single day. A tougher companion market on the same platform, wagering that all seven problems will be solved by 2030, tells a far more cautious story: it had climbed to just 24% with 107 bettors weighing in as of September 2026. The spark behind the 99% reading is OpenAI’s September 8 claim that an internal model produced a proof touching the Navier-Stokes problem, one of the seven, though the result and the credit around it are already contested.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What the market is asking
The market in question, hosted on Manifold, poses a simple conditional: will at least two of the seven Millennium Prize Problems be solved by the year 2040? The Millennium Prize Problems are a set of seven famous open questions named by the Clay Mathematics Institute in 2000, each carrying a $1 million reward, and as of September 2026 Manifold Markets notes that every one of the six still-open problems has its $1 million prize sitting unclaimed. To date, only one has been formally resolved and recognized: the Poincare conjecture, proved by Grigori Perelman in work released between 2002 and 2003. Perelman later declined the prize in 2010.
That baseline is the whole point of the wager. With just one problem off the board after more than two decades, a bet that a second falls within the next 14 years is a bet on the pace of mathematical progress accelerating, whether by human insight, machine assistance, or both. For a primer on how these contracts convert crowd money into a probability, see our explainer on how prediction markets work.
The odds, and what they imply
As of September 09, 2026, the Manifold market reads 99% implied probability of YES. On this venue that number reflects play-money positions rather than regulated cash, so it is best read as a confidence signal from an engaged crowd rather than a dollar-weighted forecast. The reading is still striking: 99% leaves almost no room for the scenario where mathematics stalls and only Poincare stands solved through 2040. Traders are far less confident about the harder version of the bet, though: the sibling market asking whether all seven problems fall by 2030 sat at only 24% in September 2026, a gap that shows the crowd sees “two by 2040” as nearly locked in while “all seven within four years” remains a long shot. The repricing was not confined to Navier-Stokes, either: a separate Manifold market tracking the Birch and Swinnerton-Dyer conjecture jumped by 38 percentage points in the days after OpenAI’s claim, a sign that traders are repricing the whole slate of problems rather than just the one in the headlines.
An implied probability is simply the market price restated as a percentage chance. If you are new to the mechanic, our note on implied probability walks through the conversion. The gap between a play-money board like Manifold and a cash exchange matters here, which is why we keep a running comparison of Kalshi versus Polymarket for readers who want regulated venues.
Why the market suddenly moved
The catalyst is concrete and datable. On September 8, 2026, OpenAI published a page titled “On the Navier-Stokes Millennium Prize Problem” and said an internal AI system had produced a proof, formally checked in the Lean proof assistant, that the equations governing fluid motion can develop a singularity in finite time. The claim was covered the same day and the next by CNN, the Washington Post, CNBC, and Nature.
The volume signal on Manifold lines up with that timing. Roughly 106% of the market’s all-time volume traded in the 24 hours around the announcement, with about 75 unique bettors active and one of the highest single-day volumes recorded for the market in its source data. In other words, the money moved when the headline moved. Side bets on individual Millennium problems predate this news cycle, too: posts on X describe a former DeepMind engineer who reportedly staked $45,000 that Navier-Stokes would be solved by December 2025, and who reportedly paid out on those wagers in January 2026, months before OpenAI’s announcement; separately, another Manifold bettor staked $25,000 on the Hodge conjecture that same December 2025, a reminder of how much side money was already circulating around these problems before OpenAI’s claim.
What OpenAI actually claimed
According to OpenAI’s own writeup and the mainstream coverage, the effort ran for about 88 hours and used as many as 10,000 AI agents working roughly concurrently, with executives putting the compute cost in the millions of dollars. The mathematical content, as described, is a finite-time blowup: an initially smooth fluid at rest, with a smooth force applied and its energy staying bounded throughout, develops a singularity within a finite amount of time. OpenAI framed this as establishing statements labeled C and D in the official Millennium formulation.
There is a large caveat, and OpenAI stated it plainly. Its configuration has a smooth external force applied to the fluid the entire time. The version the Clay Institute uses to define the prize is the unforced case, where no external force is present. Because of that gap, OpenAI said it will not claim the $1 million prize, as reported by The Next Web. That decision is the clearest signal of how OpenAI itself rates the result against the formal criteria.
The credit dispute
The announcement arrived tangled in a priority fight. On September 7, 2026, NYU mathematician Tristan Buckmaster and Levent Alpoge, an Anthropic employee, posted their own results: finite-time blowup with smooth forcing for incompressible porous media, for the Boussinesq system, and for the three-dimensional incompressible Euler equations, along with a Lean formalization. Crucially, the variant they resolved is the unforced Euler case, a related but distinct target from OpenAI’s forced Navier-Stokes construction.
OpenAI said its own run began around September 1 after it heard a rumor it later connected to the two researchers, completed its proof and Lean check by September 6, and then reached out to propose a joint announcement. Buckmaster’s public account, summarized by Fortune and Axios, describes nearly a year of collaboration and accuses OpenAI of aggressive conduct around the reveal. OpenAI and the researchers dispute how much overlap exists between the two efforts.
Who is weighing in
Terence Tao, one of the most prominent living mathematicians, called the Buckmaster and Alpoge work a “remarkable achievement” and said he saw no fundamental obstacle to extending their approach toward the full Navier-Stokes problem. He was praising the human researchers, not endorsing OpenAI’s method. Tao also voiced a broader concern, captured by Quanta Magazine, that AI firms risk treating landmark problems as marketing proof points, producing answers without the insight that usually makes such results valuable.
The Clay Mathematics Institute, which defines and awards the prizes, has not recognized any new solution. As of September 09, 2026, Poincare remains the only officially solved Millennium problem, and Navier-Stokes remains formally open. The institute has kept the problems in public view regardless, having scheduled a series of seven lectures on the Millennium problems running from September 2025 through April 2026.
Status of the seven problems
| Millennium Prize Problem | Status as of Sept 9, 2026 |
|---|---|
| Poincare conjecture | Solved (Perelman, 2002-2003); prize declined 2010 |
| Navier-Stokes existence and smoothness | Officially open; contested AI-assisted claim on a forced variant |
| P versus NP | Open |
| Riemann hypothesis | Open |
| Yang-Mills existence and mass gap | Open |
| Hodge conjecture | Open |
| Birch and Swinnerton-Dyer conjecture | Open |
Why a tech and finance reader should care
Strip away the fluid dynamics and the story is about AI capability and how markets price it. A frontier lab spending millions of dollars and running thousands of agents for days to attack a landmark proof is a capability claim, and prediction markets are one of the faster instruments for absorbing that kind of news. A more systematic gauge tells a humbler story: the tracker MillenniumPrizeProblemBench, which scores frontier models directly against all seven problems, listed Claude Fable 5.1 with 7 failures out of 7 attempts and GPT-6 Astra at 0 solved out of 7 as of September 2026, a sign that one heavily resourced proof effort has not yet translated into broad model competence on the rest of the list. The same reflex showed up in other recent bets tracked by markets, including the speculative repricing we covered in the Polymarket odds around a $1 billion valuation event.
The distinction between a play-money crowd forecast and a regulated, cash-settled contract is not academic. On a US-regulated exchange, dollars discipline the price. On Manifold, reputation and play-money do the work, which can produce sharper swings on a hot narrative. Our Polymarket review covers how liquidity and settlement change what a number actually means.
What to watch next
Three checkpoints will decide whether the 99% reading ages well. First, independent verification: OpenAI has publicized a Lean formalization, but external mathematicians will want to inspect the full argument, and Lean checking a formal derivation is not the same as the community agreeing it resolves the Clay formulation. As of late September 2026 that verification had not arrived: reporting in the two weeks after the announcement continued to describe the proof as awaiting external review, and no institution had moved to confirm or reject it. Second, the forced-versus-unforced question: any move toward the unforced case, whether by OpenAI, by Buckmaster and Alpoge, or by others, is what would actually put a second official solution in reach. Third, the Clay Institute itself, which has issued no recognition. Until those clear, the market is pricing a direction of travel, not a settled fact.
FAQ
Does this mean two Millennium Problems are now solved? No. As of September 09, 2026, only the Poincare conjecture is officially recognized as solved. OpenAI’s claim addresses a forced variant of Navier-Stokes and has not been ratified by the Clay Mathematics Institute.
Why is OpenAI not claiming the $1 million prize? OpenAI’s construction applies a smooth external force to the fluid, while the Clay prize is defined on the unforced equations. OpenAI acknowledged that gap and said it will not claim the prize.
What does the 99% Manifold number mean? It is the crowd’s implied probability that at least two of the seven problems will be solved by 2040, expressed as a percentage. On Manifold it reflects play-money positions, so treat it as a confidence signal rather than a cash-weighted forecast.
Who are the human mathematicians in this story? Tristan Buckmaster of NYU and Levent Alpoge, an Anthropic employee, posted related finite-time blowup results on September 7, 2026. Terence Tao publicly praised their work.
The Bottom Line
Sources
- OpenAI — On the Navier-Stokes Millennium Prize Problem
- CNN Business — OpenAI says it has solved one of math’s Millennium Problems
- Washington Post — OpenAI claims it solved elusive math problem with $1 million prize
- CNBC — OpenAI claims to have solved the 90-year-old Navier-Stokes problem in 88 hours
- Nature — OpenAI claims huge maths breakthrough on a famed Millennium Problem
- Quanta Magazine — AI Has Solved One of Math’s $1 Million Millennium Prize Problems
- Fortune — OpenAI says it cracked Navier-Stokes amid Buckmaster accusations
- Axios — OpenAI’s historic math solution overshadowed by credit controversy
- The Next Web — OpenAI publishes its Navier-Stokes proof and says it will not claim the Millennium Prize
Prediction markets carry risk and are not investment or betting advice. Market availability is restricted by jurisdiction: Polymarket is not available to US persons, while Kalshi is a CFTC-regulated US exchange. Participants must be of legal age (18+ or 21+ as applicable). If gambling is a problem for you or someone you know, call 1-800-GAMBLER for confidential help.
Related Coverage
- Navier-Stokes 'Solved by 2027' Bet Surges to 79% on Manifold After a Claude Rumor
- ARC-AGI-3 'Above Human' Bet Hits 92% on Manifold After GPT-6 Astra
- GTA VI 2026 Release Odds Hit 96% on Manifold After Netflix Extended Look
- Trump National Emergency Odds Sit at 24% on Manifold as Traders Rush a Late-August Bet
- Jack Lowden's Next-Bond Odds Jump 12 Points on Kalshi After Gary Oldman's Comments
