We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten.

Every new generation changing technology is followed by a frenzy of infrastructure build out.

Running up to the dotcom bubble lot of money was spent on building undersea cables and Internet infra because the assumption was that the demand for websites was going to be a straight line if it not exponential. There was an over capacity of expensive infra. Then the market crashed and the same infra went for cheap and laid foundation for Internet as we know today.

Same thing happened during the railroad frenzy in the 1800s.

And same thing is happening today. There is assumption that the pace of AI improvement and demand is going to be straight line if not exponential. So lot of money is being spent on data centres looking at “future”. There is going to be an oversupply of expensive hardware and models because who doesn’t want that sweet AI money.

Sooner or later the market is going to reset because nothing can keep going up forever. It’s only after the reset we are going to see the upsides of AI.

The railroad bubble and the dotcom bubble feature prominently as disastrous misallocations of capital in one of my favorite books: "Devil Take the Hindmost: A History of Financial Speculation."

Pitching AI as in the same vein is a very dark prognostication. As in, probably going to cause a wreck so bad we won't see breakeven in our lifetimes. The railroad bubble and the dotcom bust were generational economic calamities. For Gen X and Millenials, combine that with the 2008 GFC and it's been nonstop waves of speculation-driven disaster drowning us our whole lives.

The comparison with the internet is not valid, and I can immediately give you one reason. I was there (3000 years ago) when the internet was in its infancy. When the dot com bubble was going on, the internet was a fringe thing for most people. People didn't really use it, didn't know how to use it, didn't trust it (for purchases, for information). The investors were still right, but they were too ahead of their time, by almost a decade.

AI is not like that today already, both by common people or by companies. They're using AI at capacity. Demand is higher than supply. Not the other way around, like it was for the internet.