How many more Millennium Prize Problems will AI solve in 2026?
I think there is a solid majority chance that no additional qualifying Millennium Prize Problem will be announced solved by an eligible AI lab in 2026. The recent Navier-Stokes claim raises attention, but under this market’s rules it does not count unless a qualifying official solution announcement occurs for one of the five listed problems.
Analysis
The key issue is that the market is not asking whether AI can make a dramatic math breakthrough in 2026, but whether an eligible AI lab will make a qualifying announcement that one of the five counted Millennium Prize Problems has been solved by year-end. That is a much stricter bar than a research claim or a media report. Right now, the only widely discussed 2026 AI math claim is OpenAI’s Navier-Stokes announcement, but the market rules explicitly exclude Navier-Stokes from qualification. So even if that claim later survives scrutiny, it would still not move this market. As of now, there is no confirmed qualifying announcement for any of the five eligible problems, which leaves “0 more” as the default state.
The strongest reason to lean toward Yes is that the field has clearly shifted from speculative progress to active claims of major theorem discovery, and one large lab has already said it had substantial progress on another Millennium problem. That means the probability of at least one more public announcement by an eligible lab before December 31 is no longer tiny. However, the distance between internal progress and a market-qualifying announcement remains very large. The lab must be eligible, the announcement must clearly state the problem is solved, and it must concern one of the five qualifying problems. Historically, this kind of announcement is rare even in ordinary mathematics, and for a Millennium Prize Problem it is especially likely to be delayed by internal review, hesitation, or the decision to wait until the proof is better vetted.
The market price is close to a coin flip, which suggests traders are assigning substantial weight either to the possibility that OpenAI’s pace continues or to the chance that another eligible lab makes a surprise announcement. I am somewhat more skeptical than the market because the timing is short and because the standard for a qualifying announcement is narrow. A lab can have a valid proof and still not announce it by year-end, and a claim can be made yet later fail to meet the standard of a solved problem. On balance, I think the most likely outcome is still that no additional qualifying Millennium Prize Problem gets announced in 2026, but the probability of a late-year surprise is real enough to keep this well below certainty.
Arguments
For
- Arguments for Yes: The recent pace of AI math claims shows that a qualifying announcement before year-end is plausible, not purely hypothetical.
- Arguments for Yes: The current market price implies traders see a meaningful chance that at least one eligible lab will make a counted solve announcement in 2026.
Against
- Arguments against Yes: The only prominent 2026 claim so far is Navier-Stokes, which is explicitly excluded from qualification.
- Arguments against Yes: A qualifying announcement requires a public, explicit solved claim for one of five problems, and that standard is difficult to satisfy on a short timeline.
Key drivers
- The market requires a clear qualifying announcement for one of only five specific problems, which is a much narrower condition than a general AI math breakthrough.
- The recent Navier-Stokes claim increases confidence that AI labs can generate major proof claims, but it does not count for this market and does not guarantee another qualifying announcement.
Risk factors
- A major lab could suddenly announce a proof for one of the eligible Millennium problems before year-end, especially if internal validation is already advanced.
- If OpenAI or another eligible lab has been quietly working on a qualifying problem and chooses to publish aggressively, the market could flip quickly.
Scenarios
Best case
An eligible AI lab announces a clear solution to one of the five qualifying problems before the end of 2026, and the announcement is unmistakably framed as a solved result by the lab itself.
Most likely
AI continues to produce notable theorem and proof claims, but none of them become a qualifying announcement for one of the five counted Millennium Prize Problems before December 31, 2026.
Worst case
No eligible lab makes any qualifying solved announcement in 2026, and the year ends with only partial results, internal progress, or non-qualifying claims such as Navier-Stokes.
More from this day
- PoliticsKalshi1y
2026: Trump's bad year?
AI44%MKT4%Edge+40Hidden GemThe bear case for Trump in 2026 looks materially more likely than the market implies, because multiple legal, regulatory, and political setbacks are already active and could easily compound into a broader negative year. I’d price Yes at 44%, not 4.2%, because the market seems to be treating this as a rare tail event when the underlying drivers are already in motion.
- pop culturePolymarket8d
Who will Trump publicly insult by September 30?
AI83%MKT50%Edge+33Hidden GemTrump is very likely to insult Tucker Carlson publicly again before September 30. The feud is active, recent reporting already documents multiple Trump insults, and Carlson’s continued attacks give Trump strong incentive to پاسخ in kind.
- CompaniesKalshi1y
Boeing commercial airplane deliveries in 2026
AI18%MKT4%Edge+14UndervaluedBoeing is improving, but the reported year-to-date pace and production constraints make more than 680 deliveries in 2026 a long shot. I estimate a low-probability yes, but still materially higher than the market’s 4% price.