- Levent Alpöge, a mathematician at Anthropic (NASDAQ:ANT), claims that working with Claude AI he has proved the Riemann Hypothesis and its stronger generalization, with a Lean formalization posted for checking.
- Independent validation is pending; the claim remains unverified, and the Lean proof must be audited for correctness and absence of unproved assumptions.
- If confirmed, it would be a historic breakthrough for both mathematics and AI-driven scientific discovery, but experts caution that formal verification is a multi-step process.
A Stunning Claim
In a potentially paradigm-shifting announcement, Anthropic mathematician Levent Alpöge says he and the company's Claude AI have discovered a proof of the Riemann Hypothesis, the 167-year-old conjecture about the distribution of prime numbers, and its more ambitious generalization, the Generalized Riemann Hypothesis. The proposed proof has been formalized in Lean, a mathematical verification system, according to a report by Scientific American (OTC:SPGNY). The formalization has been posted for checking, but independent validation remains pending.
Alpöge's wording is notably qualified: "I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis." The claim extends to the Generalized Riemann Hypothesis, though the precise scope of that generalization has not been detailed in available reporting.
Lean Formalization: A Promising but Not Final Step
The Lean component is significant because it allows for machine-checkable proofs, but it does not by itself constitute final acceptance. Reviewers must establish that the formal theorem expresses the intended mathematical claim, that the proof checks reproducibly, and that its dependencies contain no unproved placeholders or additional assumptions that effectively assume the desired result. A public Lean formalization project for the Millennium problems explicitly distinguishes formal statements from solutions and requires proofs without sorry placeholders, alongside an audit of their axioms.
"The formalization is a crucial step, but it's not the final verdict," said one mathematician familiar with the verification process, who spoke on condition of anonymity. "We need to see the full audit before we can say anything definitive." Attempts to reach Anthropic for comment were not immediately successful.
Background and Context
The Riemann Hypothesis, first proposed in 1859, asserts that all nontrivial zeros of the Riemann zeta function have a real part of 1/2. Its proof would have profound implications for number theory and cryptography. The Generalized Riemann Hypothesis extends this to a wider class of functions, making the exact formal statement particularly consequential.
This claim follows a series of AI-related mathematical developments in 2026. In July, a Claude-generated result concerning the Jacobian conjecture was reported, and in October, OpenAI reported progress on a Riemann-related problem. Those earlier claims' validation status remains unsettled. In July 2025, Google DeepMind (NASDAQ:GOOGL)'s Gemini Deep Think system solved five of six International Mathematical Olympiad problems, scoring 35 out of 42 under competition time limits. In February 2026, Google DeepMind described Aletheia, a mathematical research agent that generates, checks, and revises candidate solutions.
Financial and Industry Implications
The announcement comes at a critical time for Anthropic, which just completed its initial public offering. Shares priced on November 18 at $83, raising $81.3 billion through the public offering, at approximately $1.5 trillion fully diluted valuation. An accompanying private placement brought the total transaction to $107.3 billion. The company reported third-quarter 2026 revenue of $16.96 billion, up 47% from the preceding quarter, though it posted a $567 million operating loss and a $23.3 billion net loss including non-operating items associated with convertible securities. Annualized revenue run rate stood at $81 billion as of October 31.
A validated proof could strengthen the case that frontier models can contribute to original scientific work, extending their commercial narrative beyond coding and office automation. However, the immediate economic significance is more about confidence in AI research capabilities than any demonstrated new revenue stream. The wider economics remain demanding: Reuters (NASDAQ:TRI) reports nearly $9 billion in computing and infrastructure spending in the first half of 2026, and nearly $900 billion in computing commitments to infrastructure partners. Competition is also broadening, with a Capgemini (EURONEXT:CAP) study finding that 59% of executives at large companies planned to increase open-weight model use as AI costs exceeded budgets.
"The market will ultimately judge Anthropic on whether the economics of AI are holding up," Armina Rosenberg of Minotaur Capital told Reuters. "A scientific breakthrough would not, on its own, resolve that question."
What's Next
The decisive development will be an independent technical assessment of the Lean proof. No firm validation timetable has been established. If validated, the result would resolve a central mathematical conjecture and provide unusually strong evidence for AI-assisted original mathematics. If a flaw emerges, the impact will depend on whether it lies in a repairable argument, the formal statement, or an essential assumption.
For now, the most accurate description is an extraordinary, formally presented claim awaiting independent verification—not yet an accepted resolution of the Riemann Hypothesis.