Dwarkesh says frontier-model monetization could drive H100 rents far above today’s prices

inductionheads · x · 2026-07-29

Dwarkesh Patel argues that if frontier labs keep scaling revenue while compute only grows about 3× a year, compute prices may have to rise sharply.

• He uses Anthropic’s reported revenue growth, margin expansion, and the increasing share of spend going to inference to argue that the current pricing regime may not hold.
• The core claim is that a human-level software engineer running on H100-class hardware could justify more than $250k/year in rent, which would imply compute being far more expensive than today’s spot prices.
• The post frames this as an AI infrastructure economics problem: better models may simply monetize the same compute more efficiently, pushing up the value of scarce accelerators.

Original post →

More from Venture

• microagi pays people to film housework, then turns the footage into factory robot policies — alejandroll10 · 2026-07-30
• Crypto still looks strong as Bitcoin holds 60k and ETH fundamentals improve — JOBhakdi · 2026-07-30
• 8090 says it raised a Series A as Chamath frames AI as a threat to outsourced hours — vaibhavbetter · 2026-07-30
• AI Infrastructure Stocks Cool Down: CRWV at 52-Week Low, NVDA Down 8.69% — GaryMarcus · 2026-07-30
• AI subscriptions would probably be $500 a month without Chinese rivals — iamfakhrealam · 2026-07-30
• Doctronic buys Summer Health after 30 million patient encounters in two years — HealthcareAIGuy · 2026-07-30

Venture channel →