Why Chinese Open-Weight Models Are Changing AI Economics
Stratechery · rss · 2026-07-20
Stratechery argues that Chinese open-weights models are forcing a return to first principles in AI economics.
- Open weights are not free to serve: R&D is a fixed cost, but inference is real COGS, and it scales with revenue.
- Tokens are not fungible: in the reasoning era, different models may require very different numbers of internal tokens to reach the same answer, so raw token pricing can be misleading.
- What matters is the cost of intelligence: model footprint, inference efficiency, memory efficiency, serving efficiency, and token efficiency all shape the true economics.
- Industry implication: for many useful tasks, intelligence is becoming a commodity, so profitability depends less on charging premium prices and more on having a better cost structure.
The piece uses Kimi K3 as an example and frames the competitive battle as one over cost efficiency rather than model quality alone.
More from AGI Musings
- Gary Marcus says LLMs still cannot really do math on their own — GaryMarcus · 2026-07-22
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- AI may make digital work infinitely leveraged while offline life gets more human — illscience · 2026-07-22
- Better AI math could save researchers time by killing false conjectures earlier — prateekj · 2026-07-22
- AI’s economic forecasts are split by nearly a quadrillion dollars by 2035 — bittingthembits · 2026-07-22
- Open source is becoming tech’s soft power, says Kevin Xu — kevinsxu · 2026-07-22