How many more Millennium Prize Problems will AI solve in 2026?

The market question is narrowly defined around five qualifying Millennium Prize Problems, and it explicitly excludes Navier-Stokes. That matters a great deal because the most prominent 2026 AI math headline so far is the reported Navier-Stokes claim, which may be a genuine breakthrough in the broader public sense but does not move this specific market unless one of the five listed problems is announced as solved by an eligible AI lab. On that definition, the current evidence still points strongly toward zero qualifying solves by year-end, because there has not yet been a comparable public announcement for any of the five counted problems.

The strongest argument for a Yes outcome is that the year has already seen a major AI mathematics milestone, which suggests the ceiling on what a frontier lab can credibly claim may be rising faster than many expected. If one lab has demonstrated the ability to organize a successful large-scale proof effort, it becomes more plausible that another problem, especially one with active research momentum, could be announced before December 31. However, the transition from solving a single problem to solving any of the remaining Clay problems is extremely steep, and the remaining set includes some of the deepest open questions in mathematics. A credible announcement would also need to come from an eligible lab itself, not merely an outside collaboration, which further narrows the path to a qualifying event.

The case against a qualifying solve is still much stronger. There are only a few months left in 2026, and there is no public indication that any eligible lab is close to announcing a solution to Birch and Swinnerton-Dyer, Hodge, P versus NP, the Riemann Hypothesis, or Yang-Mills. These problems have resisted decades of elite human effort, and even with advanced AI assistance, the gap between promising research support and a full announcement is enormous. The market price on Yes is high, but that likely reflects the gravitational pull of the Navier-Stokes story and general enthusiasm around AI math rather than strong evidence that one of the five qualifying problems will be announced in time.

The right framing is that the market is not asking whether AI can solve a Millennium Prize Problem in 2026 in general, but whether it will solve one of the five counted problems. On that narrower reading, the base rate remains low. A late-year surprise is possible, especially from a lab already investing heavily in formal reasoning and proof systems, but absent concrete signals the most defensible assessment is that zero qualifying solves remains the likeliest outcome.