I think the author is incorrect to look at the cycle of market bubbles and should instead be looking at the Gartner Hype Cycle. We are currently at or just past the peak and moving into the trough of disillusionment (to wit: this article). Eventually the companies that are actually building useful technology will emerge and win out in the long run, the others will fail. Just the same as any other new hyped technology.

https://en.m.wikipedia.org/wiki/Gartner_hype_cycle

Not sure any of the latest hyped technologies have really recovered out of the trough. I don’t see blockchains, hyperloops or NFTs getting any meaningful adoption.

We had the dot com bubble but it didn’t end the tech sector which is bigger than it ever was during that bubble. The same is likely with AI. If the reality doesn’t match the hype then the bubble may burst but the technology will eventually find its place.

The dot com bubble ended when the ecosystem of fairy tales and lies which built up around a technology were generally decided to be nothing but hot air. It is true that the technology sticks around but the wildly inappropriate estimates of future finance, social behaviour and technological capability subside after these bubbles burst. Plus a bunch of too-willing-to-believe investors lose a bunch of money.

This makes the bubble dangerous for investors though. For example those who think nvidia will keep going up.

Most bubbles are based on over-investment and mania about worthwhile ideas but that doesn’t mean any bubble companies will survive.

AI and the changes it makes are here to stay. The financial markets will go through cycles trying to adjust to these changes. He's right. This is a predictable pattern. But he concludes implying the downcycle will be permanent. The web survived and thrived after the dot.com crash. AI will survive and thrive after the ai.com crash. It won't end in the planet covered in data center as he ridiculous jokes at the end of the article.

At the moment we could make only one conclusion: AI already disruptive technology and it's using already good enough for wide adoption of already existing models; from other side, it is not universal tech, and nobody knows when it become universal, and in current staTE it is impossible to sustain growth to universal.

To be more clear, AI is just semiconductor application, which is now growing faster them Moore law, nothing more.

And it is possible to change state to highly regulated, so AI will be monopolized and limited to academy environment and to big corporations (which could pass regulations), and this could make it profitable, at cost of growth slowing.

But tough regulations are against market competition, this will not be done, I believe.