I just skimmed the tweet you shared, so not 100% sure I interpreted it perfectly, but I think you’re misunderstanding where the demand is coming from in this scenario.
Until there’s a new breakthrough with underlying architecture, the only way current leading models can improve is via more parameters and more data. It doesn’t necessarily mean that there’s more demand from users or power users.
This is primarily how it could end up being a economic bubble imo, and a major reason why I respectfully disagree with
@Rastafarian about these leading AI companies ability to quickly turn enough profits to justify their valuations.