I see "gold rush" the company selling shovels is making out like a bandit, everyone else is make a profit on the previous gen but requires a 10x cost increase for the next gen. And thus 10x more shovels.. As soon as 10x more shovels stops giving 10x+ improvements this is the wrong investment.
Hints are we already reached this point.
Some AI companies will pivot and improve in other ways with more linear costs/results.. The ones hoping the line continues to the moon.. I think they overshot.. I just don't know when it will fall back..