LessWrong (30+ Karma) “Systematically dismantle the AI compute supply chain.” by David Scott Krueger (formerly: capybaralet)
Apr 2, 2026
A critique of a documentary's narrow solution framing leads into a third way: removing advanced AI compute from the equation. The conversation maps the global compute supply chain and pinpoints chokepoints from fabs to data centers. Concrete strategies are proposed for shutting down or buying out production and for international monitoring to prevent covert restarts.
AI Snips
Chapters
Transcript
Episode notes
Compute Concentration Creates A Third Path
- The scaling paradigm means AI progress depends on massive investment in specialized hardware rather than lone geniuses.
- David Scott Krueger argues that removing that hardware creates a third option beyond 'lock it down' or 'let it rip'.
Key Chokepoints In The AI Compute Chain
- The compute supply chain is highly concentrated: TSMC fabs, ASML EUV machines, Zeiss lenses, Spruce Pine silicon and US data centers.
- Krueger lists these chokepoints to show where interventions could stop advanced AI progress.
Dismantle Chips Through Coordinated Government Action
- Remove AI chips and their fabs to halt advanced AI progress via government action, buyouts, and buybacks.
- Krueger suggests international commitment, compensated shutdowns, chip buybacks, and surveillance limited to industry insiders to prevent secret restarts.