← Index
Thread
Capital commitments
Long-dated spending on compute, and the risk of fixing costs ahead of demand. 5 notes so far.
- November 2026InfrastructureThe AI boom has co-signersNvidia, Broadcom, Alphabet and Meta have promised to pay about $270bn if other parties' AI data centres and chips go bad, up from about $45bn disclosed at the end of 2025.~$45bn → ~$270bn of guarantees in nine months
- November 2026InfrastructureThe neocloud GTM reset: sell the GPU-hour and you give away the engineeringSix providers charge exactly the same to run OpenAI's gpt-oss-120b; the fastest is eight times quicker than the slowest. Yet 98% of CoreWeave's revenue is still rented capacity, and capacity is getting dearer.Same price: 46 → 394 tokens a second
- November 2026InfrastructureThe collateral is Broadcom: what a $63bn chip lease says about AI creditFour of Wall Street's biggest names will fund $63bn of chips for Anthropic. They signed only after the chipmaker promised to cover up to $43bn if the chips lose their value.$50bn → $451bn of leases in nine months
- November 2026MarketsAnthropic's S-1: revenue per compute dollar rose from $0.40 to $2.08Revenue grew 14 times in a year; compute cost grew 3.8 times. The AI race is becoming a capital allocation problem.$0.40 → $2.08, then $1.72 per $1 of compute
- January 2026InfrastructureToken bills grow with every step an agent takes. A cluster you own doesn't.Metered inference costs 2.4¢ a case and owned capacity 8.7¢. Once agents use about 3.6 times the tokens, owning is cheaper, and agents typically use four times.Owning wins past ~3.6× tokens per case