Nvidia is interesting on multiple fronts. For one, GPUs aren't just useful for ML. Secondly, ML isn't just LLMs. I also find their bet on USD to have been vindicated by other industry players, and their Omniverse/digital twin concepts are going to be important tools for media, robotics, research, among other things. I'm not so sure their P/E isn't absurd, because it probably is, but you're not betting on just one horse with their stock.

You can easily make maybes in the other direction too - maybe they maintain their absolute market dominace, like Intel for years and years. Maybe they come with a technology jump that will be hard to beat for a long time, like M1. We are probably in the early adopter stage of current AI too.

Predicting the future is hard. Stock prices are based on extrapolated current information and mostly hype.