AI and compute economics
2026-04-26
11 minute read
21 sources
AI inference economics in 2026: GPT, Claude, Gemini, and the pricing war that is rewriting the application stack
Token prices are falling roughly 10x per year at constant capability, the marginal frontier provider is now a Chinese open weight lab, and hyperscaler capex is committing 325 billion dollars to a market where the unit price keeps collapsing.
Inference is the largest unsolved cost line in enterprise AI. Output token prices for frontier general capability have fallen from 60 dollars per million in 2023 to between one and three dollars in 2026, a roughly 20 to 60 fold compression. Epoch AI estimates a 10x annual decline at constant capability, sustained for three years, driven b...