Catching up to the frontier needs ~1GW, not 7GW: critique of Europe's 14GW compute plan
anton_d_leicht · x · 2026-09-15
Critiquing a report proposing 14GW of compute in Europe, antondleicht argues 1GW — roughly the size of xAI's Colossus 2 — suffices to catch up to the frontier; just 2GW of the proposed 14GW (about 1M B300s) would be enough to match OpenAI/Anthropic's internal models.
He adds that compute built today must be at frontier scale for pretraining + RL + internal inference years later when it comes online, plus external inference headroom, and questions the report's 'decoupling' logic — the number is roughly 4x what OpenAI or Anthropic raised.
More from Infra
- effect-harness on Cloudflare: run agent loops, tools and credentials on separate VMs — samgoodwin89 · 2026-10-31
- Google's Project Suncatcher boots TPU in orbit and beams back first AI response — sundarpichai · 2026-10-31
- Together AI launches Post-Training platform for agentic RL with tool-use environments — togethercompute · 2026-10-31
- General Compute raises $100M for inference capacity as CoreWeave stock hits 12-month low — YvesMulkers · 2026-10-31
- Analyst warns of unstable AI compute feedback loop as Kioxia's NAND price warning triggers flash crash — bookwormengr · 2026-10-31
- SOMA launches enterprise prompt-compression layer claiming 20%+ inference savings — markjeffrey · 2026-10-31