Will cheap open-source models kill OpenAI and Anthropic's margins? Fin argues no
dotey · x · 2026-10-21
Dotey summarizes part IV of Fin's 'AI Semiconductor Endgame' series, tackling the debate over whether cheap open-weight models will leave OpenAI and Anthropic unable to profit or buy GPUs. Fin says no, on three grounds:
- Prices and costs fall in lockstep: each chip generation roughly doubles tokens per watt, software optimization halves costs yearly. A 1GW datacenter running frontier models could sell $80B/year while hardware amortization is only $8B.
- Two separate markets: open-source competes only at the low end; the frontier tier is contested only by OpenAI and Anthropic, keeping prices stable.
- Open-source is small in dollars: 20%+ of token volume but only 5–10% of actual spend; stronger models enable longer tasks, so demand keeps growing.
More from Venture
- Oxide raises $445M Series D just months after its $200M Series C — Sethwinterroth · 2026-10-21
- Big day for AI in bio: OpenAI-Orchestra Bio deal, Holon-S1 cancer model, more — iScienceLuvr · 2026-10-21
- Open/custom model inference is a market as big as closed intelligence, argues VC saranormous — saranormous · 2026-10-21
- Google: Southeast Asia's digital economy to hit $368 billion in 2026 — Google AI (The Keyword) · 2026-10-21
- Blogger's top 3 traffic sources are all ChatGPT, with Google falling behind — tinyfool · 2026-10-21
- Investor: market underestimates compute demand 1,000x, heading terawatt-scale — inductionheads · 2026-10-21