Dwarkesh Explains AI Economies of Scale: 10x Revenue vs 3x Compute Growth
morqon · x · 2026-07-30
Podcast host Dwarkesh analyzed the strong economies of scale in the AI model business. He noted that while top AI labs are only 3x-ing their compute year over year, their revenue has been 10x-ing.
This phenomenon of revenue vastly outpacing compute growth is driven by three main factors:
- Increasing Margins: Maturation of model commercialization leading to lower marginal costs.
- Rising Compute Prices: For instance, Google is reportedly paying SpaceX around $900 million/month for 110K GPUs—roughly 2x the current spot price, which itself is up 40% since February.
- Shift to Inference: Labs are spending a larger fraction of their compute budget on highly profitable inference rather than just training.
He explained that this makes logical sense: the one-time cost of training a model can be amortized across millions of users, unlike human labor which must be trained from scratch. However, he expressed concern that these strong economies of scale might lead to an undesirable concentration of power.
Related event: GPU Rental Prices Surge, Cloud Profits Underestimated(8 posts)→
More from AGI Musings
- Miles Cranmer's talk: how to avoid 'AI Brain Rot' and intelligence atrophy — Scobleizer · 2026-10-16
- AI's fundamental design flaws are fully hidden by its products' superficial design flaws — lescarr · 2026-10-16
- IMF Official: Markets' Surprising Resilience Largely Rests on AI Optimism — kernelangus420 · 2026-10-16
- Can AI trained only on human content infer the shape of the human brain? — rickasaurus · 2026-10-16
- Vals RSI Index extrapolates Google models hitting 130% autonomous AI R&D by Oct 2027 — HaydnBelfield · 2026-10-16
- Gowers' tl;dr: no p(doom), but can't confidently say it's close to zero — michael_nielsen · 2026-10-16