This is a fallacy. OpenAI and Anthropic would not continue to make money indefinitely by simply sitting still for the simple fact that their models can easily be distilled by competitors. Their value is contingent on sitting at the top of the leaderboards and staying there such that the marginal value of their AI is better than the mostly Chinese competitors.

And b/c these chinese competitors are open weight, the layer below frontier class AI is totally commoditized.

If there were a recession, the first thing enterprise customers would do is setup Kimi or Deepseek rigs. It would be a race to the bottom and no one would be profitable.

A similar phenomenon happened with rail lines in the 1850s where irrational exuberance led to a massive overbuild of rail lines, followed by a race to the bottom and the bankruptcy of almost all players. In the end, banks ended up absorbing the few companies that survived.