Back-of-envelope math: 120B DeepSeek-class tokens per person by end of 2027
bookwormengr · x · 2026-10-28
With 25GW of inference compute and Vera Rubin-class servers, the author estimates that by end of 2027 the world could generate 120B DeepSeek 4.1 Flash (1017)-quality tokens per living person, plus 1.2B Kimi 3.1 Pro-quality tokens. Assuming only 1B serious users, that's 960B and 9.6B tokens respectively — huge slack. This excludes another 25GW reserved for training/post-training (SemiAnalysis estimates 50GW online by end of 2027), and the author doubts training needs that much given Anthropic has spent only $15B on training compute to date.
More from Infra
- Palmer Luckey Unveils BlueChip, Billed as First American-Made Computer in Decades — whurley · 2026-10-28
- OpenRouter's entire DeepSeek traffic could run on a single GB300 rack, worth just $6-7M — teortaxesTex · 2026-10-28
- Open-source ai-phone: local LLM answers real calls on your Mac, nothing leaves the machine — Agent007_MI9 · 2026-10-28
- Agents built a custom inference engine: AMD MI350X serves GLM 5.3 FP8 37% faster than NVIDIA B200 — yisongyue · 2026-10-28
- Why RTX 5090s cost $10k: leaked factory footage shows cards converted into AI accelerators — Different_Fix_2217 · 2026-10-28
- Anthropic's S-1 admits its data centers rely on natural gas, exposing it to fuel costs — TobyWalsh · 2026-10-28