Why Soaring Training Costs Haven't Caused Industry Consolidation

natolambert · x · 2026-08-02

AI researcher Nathan Lambert discusses why the AI industry hasn't consolidated into a monopoly. Despite exponential increases in training costs, more companies are investing hundreds of millions to billions of dollars to train strong models and releasing them openly.

The core driver is the incredibly high demand for tokens. As models become more efficient and unlock more use cases, labs realize that building 'token machines' is a viable path to value. An increasing number of organizations are identifying this value source, driving the continued prosperity of open-source models on the Pareto frontier.

Original post →

More from AGI Musings

AGI Musings channel →