Sam Altman calls local inference 'a terrible allocation of scarce resources', gets mocked
ResearchCrafty1804 · reddit · 2026-10-30
Sam Altman argued in an interview that local inference is a terrible allocation of scarce resources: open-source models make people assume they can run everything on their own machines, but having the world's limited DRAM supply sitting idle in personal computers 95% of the time is inefficient — cloud utilization should be maximized in a supply-constrained environment.
Reddit commenters distilled the argument bluntly: 'Guy selling product explains why owning your own product instead of renting it from him is a terrible idea,' pointing out the obvious conflict of interest with OpenAI's cloud business.
More from Infra
- Follow-up chart: AI compute sales doubling every 9 months per Epoch AI — SamuelAlbanie · 2026-10-30
- AI compute sold in Q2 2026 topped all of 2024 as capacity doubles every 9 months — SamuelAlbanie · 2026-10-30
- NY data center moratorium was summer's most-lobbied bill as spending hit $64.5M — news-10 · 2026-10-30
- SGLang RFC decouples model code from parallelism so new models skip reimplementation — BanghuaZ · 2026-10-30
- Jassy says AWS will be a $1T revenue business — after years of heavy AI capex — TechCrunch AI · 2026-10-30
- Even 2020-era GPUs keep gaining rental value, undercutting the AI bubble thesis — JFPuget · 2026-10-30