Compute Crunch Behind the Subscription Limit Wars
量子位 · wechat · 2026-07-13
Anthropic has again extended the subscription period for Claude Fable 5, while OpenAI temporarily removed the 5-hour limit for Codex. Both are using more aggressive quota strategies to win over developers.
The article cites Benedict Evans's analysis:
- The most certain thing in the AI industry right now is a "severe supply shortage," which is unsustainable.
- Inference businesses have a gross margin of about 40%–50%, but this doesn't account for the massive annual spending on training next-gen models.
- Over $1 trillion in data center CapEx is under construction. While inference efficiency is improving rapidly, demand for compute from new models is also surging.
The author believes a specific reason for the compute crunch in H1 2026 is that software development, a single use case, suddenly achieved product-market fit. If similar scenarios emerge for hundreds of millions of DAU, the existing global infrastructure won't hold up.
Regarding the competitive landscape, the article suggests that the methods, data, and capabilities of various frontier models are converging, with no clear network effects or winner-take-all moats yet. Analogous to the mobile communications industry, the underlying infrastructure might eventually become a low-margin commodity, with value shifting upstream in the industry chain.
More from Companies & People
- RSS launches under OMSF to push structural biology data modeling at scale — MoAlQuraishi · 2026-07-22
- Moonshot AI reportedly targets a $50B round ahead of Hong Kong listing — ctjlewis · 2026-07-22
- Annotated transcript of a Claude Code team interview is now available — trq212 · 2026-07-22
- AI industry astroturfing roundup tracks the sector’s fake-grassroots problem — ShakeelHashim · 2026-07-22
- Enterprise AI’s real value is in agentic workflows, not generic chatbots — irfan__77 · 2026-07-22
- Tesla expands Robotaxi rides to seven areas, including new Orlando and Tampa zones — elonmusk · 2026-07-22