Anthropic Scientific Breakthrough and IPO Delay: Claude Proves Fermat's Last Theorem as Listing Slips to Autumn
Anthropic has announced a historic breakthrough in scientific AI, revealing that a coordinated multi-agent system powered by Claude successfully formalized and verified Fermat's Last Theorem in the Lean theorem-proving language. Operating over an intensive 11-day period, the multi-agent collaboration wrote more than 13 million lines of machine-verified proof code and proved tens of thousands of supporting mathematical theorems. The achievement consumed approximately 6 billion output tokens from an internal research model with capabilities equivalent to Anthropic Releases Claude Fable 5.1 and Mythos 5.1, Cutting Cache-Read Costs by 75%.
IPO Prospectus Delayed to Late September
While Anthropic dominates scientific benchmarks, the company is adjusting its aggressive public-market timeline. Originally expected to make its IPO prospectus public in early September 2026, Anthropic has pushed the filing back to late September, with marketing of the offering now slated to begin no earlier than mid-October 2026.
The revised timeline pushes the blockbuster flotation—which is targeting a record-breaking valuation of up to $2 trillion—to within days of the U.S. midterm elections in November. The delay is attributed to ongoing corporate preparations, including finalization of a massive $15 billion revolving credit facility led by Morgan Stanley, Goldman Sachs, JPMorgan, and Citi.
Additionally, Anthropic is rapidly advancing its international footprint and is in the final stages of completing its $6 billion to $7 billion acquisition of Israeli AI efficiency startup Decart, which specializes in lean language models. Decart shareholders are heavily betting that Anthropic's post-IPO share value will surge as the company's annual revenue run rate reportedly tops $65 billion.