Which Millennium Prize Problem will AI solve next?

The core issue is that this market does not reward vague progress, promising preprints, or even a convincing informal claim; it requires an eligible AI lab to publicly announce that BSD has been solved. The recent context contains talk of AI systems making major advances in math, but the evidence described is still indirect and mostly unverified. That matters a lot because Millennium problems are unusually resistant to partial breakthroughs, and the gap between impressive theorem proving and a complete accepted solution remains very large.

BSD has a few features that keep it in the conversation. It is deeply structural, heavily connected to modern number theory, and a future AI system that can combine symbolic reasoning, experiment design, and proof synthesis might eventually find a path through it. If one of the large frontier labs is already pushing on advanced mathematical reasoning, BSD is the sort of problem that could plausibly emerge from a breakthrough wave rather than from slow incremental work. The rumor environment also suggests that some teams may be testing frontier systems on hard math targets, which increases the odds that a major public announcement could happen before 2027 ends.

Against that, BSD is not the most naturally accessible Millennium problem for AI. Even if an AI lab produces something mathematically serious, the announcement must be complete, clearly credited to the lab, and strong enough to survive scrutiny as a solution rather than a partial result. The recent references specifically emphasize that current claims have not risen to that standard. In practical terms, the market should discount heavily for the combination of stringent resolution language, the difficulty of full proof generation, and the fact that only one problem will count as the next solved one. That leaves BSD as possible, but far from the most probable outcome.

The current price implies a roughly one-in-five chance, which feels a bit rich given how little confirmed BSD-specific evidence exists. I would still assign meaningful probability because frontier AI progress can be abrupt and because the market horizon extends through the end of 2027, but the base rate for a verified AI-led Millennium Prize breakthrough remains low. My estimate is that BSD is more likely than a random guess, yet still clearly below the market-implied level.