# AI capex — X 热门讨论 (2026-09-29 06:01 UTC)
## @SloCan68 (Nostra Thomas 🇨🇦) · 09-29 01:32 · ♥30 ↻4 💬2 👀
Big Tech bought the #nuclear reactors. Nobody locked up the fuel. Erik Townsend’s call: within 12 months a hyperscaler announces a #uranium deal big enough to shock the market — offtake, equity + offtake, or a producer. First deal has to be huge. Second one costs a lot more. Meta, Amazon, Google, Microsoft have lined up multi-GW nuclear (Natrium, Oklo, X-energy, Kairos, TMI restart). Public fuel-chain deals covering mine → conversion → enrichment: none. Most of those designs need HALEU. No Western commercial-scale HALEU. First cores eat multiples of an AP1000’s uranium and SWU per kW. TerraPower already slipped because the first core was predicated on Russia. For a hyperscaler, even tripled U3O8 is a rounding error next to $50–60B per GW of AI capex. For a merchant utility, the same move can wipe the margin. That’s the stampede setup Alkin has been describing: fuel buyers don’t care until they do — then they run people over. Cameco is the conservative yardstick (Western pounds, Port Hope, 49% Westinghouse, 49% GLE). Ownership caps mean a contract/stake is likelier than a full bid. Other producers can be bought outright. First mover gets the last pre-announcement price. After that, utilities realize they were front-run by companies that aren’t even in their industry. 💪 👇 > 引用 @ErikSTownsend: The Uranium Deal Nobody Sees Coming https://x.com/SloCan68/status/2104746116600172682