NewScientist has an unusually well-balanced take on the Altman Bubble. They are more balanced than usual because some of the Bubble Lords have gone over to the Official Dark Side, but nevertheless the article is fair. It points out that most of the data center investment is happening in unregulated “shadow banking” circles, so the backers don’t have the usual caution about risks.
The last part is worth quoting:
[Popping the bubble] could force us to face the reality that using a vastly expensive tool like GPT-5 to write an email is like using a sledgehammer to crack a nut, and that the true cost of using it had previously been hidden by the frenzied AI arms race. “At the minute, there’s a lot of free lunch, but at some point these companies have got to make a profit,” says Poskett.
The author compares AI to the railroad bubble of the 1880s and the dot-com bubble of the 1990s. Both of those bubbles destroyed a lot of fortunes but left some permanent infrastructure behind in the form of tracks and cables. Not incidentally, railroads had a hugely sophisticated electrical data web in the 1880s, a bigger advance in web tech than Comcast cables.
AI data centers are NOT permanent infrastructure. They depend on chips and boards that last only a few years, and they CONSUME massive amounts of electricity in return for DESTROYING massive parts of the real economy and real civilization. If the bubble continues a bit longer, it will give us more nuclear power, especially small modular reactors. Nuclear power PRODUCES massive amounts of clean electricity which can be used for REAL purposes. Reactors WILL be permanent because decommissioning a reactor is even more trouble than building it.
Polistra's Mill 