Tuesday discovered that tomorrow has a capital budget.

Europe put €3 billion behind having an AI company of its own

Mistral has raised €3 billion at a post-money valuation above €21 billion, roughly $24 billion. Samsung Electronics led the round, with the EU-backed Scaleup Europe Fund and PSG Equity among the co-leads. Mistral calls it the largest private technology equity round completed in Europe.

The valuation is impressive. The strategic motive is more revealing. Mistral has positioned itself as Europe’s best chance to retain control over more of the AI stack: models, enterprise products, compute and deployment.

That helps explain why a company still far smaller than its largest American competitors can attract capital out of proportion to its current scale. Investors are buying revenue potential and optionality for technological sovereignty.

Europe finally has enough confidence in its AI champion to fund it like strategic infrastructure—or has decided it cannot afford not to.

Claude wrote 13 million lines explaining why Fermat was right

Anthropic says dozens of Claude agents produced the first complete computer-checked formalization of Fermat’s Last Theorem in Lean. The run took 11 days and generated roughly 13 million lines of code and 29,500 intermediate theorems used in the final proof.

Claude did not discover Fermat’s Last Theorem. Andrew Wiles proved it in the 1990s. The system translated the mathematical machinery into a formal language whose logical steps can be checked by software.

That distinction makes the result useful. AI can generate plausible scientific arguments faster than people can inspect difficult ones. Formal mathematics offers an unusually strong escape hatch: convert the argument into an artifact a proof checker accepts or rejects.

The remaining questions matter. Machine verification establishes that the formal artifact follows from its encoded foundations; it does not automatically tell us how useful, elegant or faithful every route through 13 million lines is.

The capability is not typing quickly. It is maintaining enough structure across an enormous formal system for the result to survive mechanical checking.

The next giant AI chip needs a bigger stencil

ASML and TSMC are organizing an industry transition from today’s 6-inch photomasks to 12-inch masks for High-NA extreme-ultraviolet lithography.

A photomask carries the pattern projected onto silicon. High-NA EUV can print smaller features, but the current mask format constrains the area that can be exposed at once. That is awkward when cutting-edge data-center processors are physically enormous.

The partners want a 12-inch mask pilot line by 2031 and broader advanced-node readiness by 2033. Larger masks are intended to support large dies while improving manufacturing productivity and reducing the need to stitch exposures together.

The timeline is why the story belongs in HGI. The machinery for 2033 chips has to become an engineering program in 2026.

Software ate the world. Now it needs the world to build more machines.