Frontier AI labs may stop selling their best models. The core argument: when five, six, or seven labs reach roughly equivalent capability, selling model access by the token becomes a race to the bottom. The profit logic shifts entirely toward owning the product layer, not licensing the intelligence underneath it.
The strategic math is direct. A lab with a commanding, durable lead sells model access at high markup. A lab in a crowded, near-tied field builds products its competitors cannot replicate with last-generation models, then pulls up the ladder. OpenAI and Anthropic sit in that second position today. The evidence is already visible: Anthropic has shipped Claude Code and Claude Design, moving beyond pure API revenue. OpenAI is pushing ChatGPT as a superapp. These are not side projects.
What makes the full piece worth reading is not the conclusion but the mechanism: how open-weight competition from China, plus profit-insensitive scaling from Google, Meta, and SpaceX, collapsed the two-lab monopoly assumption that underpinned most AI business models built since 2022. If that assumption is gone, every SaaS company, every API-dependent startup, and every enterprise AI buyer needs a new plan.
[READ ORIGINAL →]