One Megawatt of AI Factory Costs $59M Up Front, Pays Back in Just Over Four Years
le_james94 · x · 2026-09-26
Summarizing Lecture 3 with Crusoe CEO Chase Lochmiller: one megawatt of AI factory costs about $59M up front — $19.2M building, $40M IT, of which $30M is GPUs. Rented as raw infrastructure it returns $15M a year, with payback landing a little over four years. The unit of account has shifted from dollars-per-GPU to dollars-per-megawatt.
The quoted Lecture 2 adds context: cost per token fell 99% in two and a half years while total spend went up — reasoning models and agents burn orders of magnitude more tokens, and H100 spot prices climbed through the whole decline. A fast chip is not a business.
More from Infra
- Altman says OpenAI aims to replace Nvidia in the medium to long term — zephyr_z9 · 2026-11-11
- Chinese University Bought Nvidia H200 Chips Despite US Blacklist — pstAsiatech · 2026-11-11
- vLLM Crosses 4,000 GitHub Contributors on Its Inference Engine — NVIDIAAI · 2026-11-11
- Baidu's ERNIE Evolve tops public bests on NanoGPT Speedrun, brings RSI to real inference optimization — xiaohu · 2026-11-11
- Baidu says RSI auto-optimized ERNIE inference, 6.24x decode speedup on Kunxin P800 — xiaohu · 2026-11-11
- Nvidia's supply commitments jump from $119B to $279B in one quarter, locking up 2027-28 memory — tengyanAI · 2026-11-11