It is certainly a fascinating industry to watch, and I think that perhaps there will be some money to be made betting on the downside.

Rephrased from an article in the FT:

A recent analysis from from the Brookings Institution notes that spending on the AI services produced by all those data centres would need to account for an astounding 9.2 per cent of US GDP by 2032 for their operators to make a decent economic return. Customers would need to spend $3.7tn a year.
Gartner came up with a similar calculation, estimating that $8.5tn of AI revenue will be needed between now and 2030 to produce decent investment returns on all the new data centres. The actual AI revenue Gartner forecasts over that period: $3.6tn.
Already, the big 5 are in aggregate cash flow negative, which is pretty impressive given that without that spending it would be +$700bn of cash flow this year.
Financial Times: Content 5dd8b17e b0a3 4b14 9d83 aba2627e635d

It seems pretty clear to me that this revenue is not going to happen, and something has to give. The building plans are going to slow massively at some point, and pretty likely to go into reverse.

My greedy thinking:
This implausible 9%-of-GDP calculation is based on the current forecasts of capex. That is a mix of the Oh-MAMA biggies (Oracle, Meta, Alphabet, Microsoft and Amazon) and a lot of relative also-rans, which includes firms that are slicing the capex in different ways (power, buildings, equipment, CPUs, any mix of those, and any mix of renting/leasing/financing those). If a bunch of projects get cancelled, then it's true that the plausibility of aggregate revenues and the aggregate economic returns from the global build-out will perhaps improve, but the firms that cancel plans may be in trouble as their business plans go up in smoke. Offhand, I would see the weak hands to be Oracle and the more frothy firms that are smaller than the Oh-MAMA set but with very speculative valuations. Some of which are public firms with options available.

I can imagine putting 1% of my portfolio, gradually, into a random smattering of put options at various low strikes and longish dates, with a focus on those without other sources of income.

I might not start quite yet, as it's extremely rare to see big and lasting widespread drops in equity prices very soon after a fresh headline market high. The last one was only 7 trading days ago. But I think winter is coming.

Jim