of the many companies trying to adding ai to anything and everything, ai powered butt scratchers etc, there might be 2 or 3 that will survive this nonsense.
AI is good at learning patterns, adaptive control, and replacing heuristic-based logic. Problems which don't have a good closed-form solution (which is a lot of real world problems) are often better solved through AI than having a programmer try to think up some ad hoc rules.
So, I don't agree that it doesn't have a role to play in a lot of contexts. Is it being oversold? Of course. Tech trends are always like that. That doesn't mean there's no meat to the claims. So, I partially agree that there will probably be point where the industry turns sour on AI, but it won't go away.
Furthermore, I'd say the only one feeling a hard crash should be Nvidia, but that's partly because it will be facing increasing competition on top of a decrease in demand. A lot of its biggest customers, like Google, MS, and OpenAI are building their own solutions, while AMD, Cerebras, and Intel are catching up to it on the open market.