How far are we from OpenAI being “too big to fail”?

Eventually this bubble will burst. Question is what’ll do it.

(I’ll say I don’t use OpenAI after the DoD stuff, so don’t misconstrue this as approval.)