I'm not sure that's infinitely true as long as AI costs to the user are proportional to the cost it takes to run the model. Even if user costs are heavily subsidized by investment, as long as they are non-zero and go up when models cost more, there will be at least some pressure for cheaper models and not just more capable ones and that pressure will go up with costs. AI is a crazy industry, but it's not totally immune to the law of supply and demand.

The real question though is how close are we to the point where the pressure is more for efficiency rather than capability. Anecdotally I think it's a ways off. Right now the general vibe I get is that people feel AI is very impressive for how cheap it is to use, which suggests to me that a lot of users would be very willing to pay more for more capable models. So the tipping point where AI hardware demand might slow down seems a ways off.

If models become more efficient we will move more of the work to local devices instead of using SaaS models. We’re still in the mainframe era of LLM.

Jevons paradox https://en.wikipedia.org/wiki/Jevons_paradox

I disagree. I think a sharp drop in memory requirements of at least an order of magnitude will cause demand to adjust accordingly.