AI capability could get 32 times cheaper in 5 years, says one analyst
jefrankle · x · 2026-07-25
In reply to Nat Lambert, the post argues that the best Moore’s-law analogue for AI is not scaling laws, but intelligence efficiency: how much cheaper a given capability becomes each year as models help train better models.
- If training/inference efficiency improves by 2x per year, the same performance would cost 32x less in 5 years and 1,000x less in 10 years.
- The implication is that intelligence becomes a commodity much faster than many people assume.
- The linked image adds the same theme: if capability costs fall by 4x per year, today’s expensive models become dramatically cheaper over time.
Related event: AI Intelligence Efficiency Predicted to Follow Moore's Law(2 posts)→
More from AGI Musings
- An open AI systems stack could become a much bigger economy than closed model ecosystems — rakyll · 2026-07-25
- The Verge says the Google Zero era is no longer something the web can ignore — gaganghotra_ · 2026-07-25
- ClickHouse and Nvidia argue open models will matter across safety, cost and performance — mattturck · 2026-07-25
- Nearly 60% of academic economists say they are already using AI agents in research — soumitrashukla9 · 2026-07-25
- AI Search Dries Up Web Traffic: The Era of Google Zero Is Here — The Verge AI · 2026-07-25
- Claude Opus 5 rates its own moral patienthood at 41% in automated interviews — Sauers_ · 2026-07-25