With each passing day, the probability is rising that the AI boom is coming to an end.
Nvidia reported good earnings yesterday and the guidance was upbeat. At the very least, there was no sign of trouble in the earnings numbers. Shares initially cheered, rising more than 5% after hours but that disappeared at the open and NVDA is now down 2.3% on the day.
The inability to bounce on good news despite a 38% drop in the past month and an undemanding 9x multiple is worrisome.
The price action in Micron and CoreWeave since recent earnings reports makes it look like a pattern.
The optimistic take is that the profits from AI are shifting away from chip and equipment makers to the users, and potentially the labs. Competition is coming for chips and one of the killer apps for AI will surely be designing custom chips and hyper-efficient hardware. So while the profits will likely remain high this year and next, there’s a rub pull coming with profits inevitably cratering just as more capacity comes online.
The pessimistic take is that the economics of AI make no sense. Institutional Investor this month laid out the case that AI infrastructure investments could require economic value equivalent to 20% of all global corporate operating profits to justify the spending. And if investment keeps growing at its current assumed pace, the requirement could eventually approach 50%.
That’s ludicrous and obvious bubble stuff. It’s based on $9 trillion in capex from 2026-2030 and the assumption that every $1 trillion in capex must ultimately create about $1.2 trillion of annual value for customers.
Similarly, a BCG report is particularly interesting because it estimates that current annual AI revenue for the largest hyperscalers is only around $100 billion, versus the $1.6–2 trillion eventually required in its model. That’s a 16–20-fold increase.
So the bear case on the whole boom is that people are doing the math and finding it difficult to envision an economic return.
That brings us to the second bullet in this week’s AI trade– the Anthropic IPO. It looks to be pricing well and that should be giving markets a lift but it doesn’t seem to be. The bull case is that Anthropic will capture a great deal of the economic value of AI, likely by disrupting incumbents but — if you believe the IPO hype — also by curing diseases.
The problem is the starting point. Investors putting a $1.25-$1.75 trillion valuation on it now are already pricing it as one of the most-valuable companies in the world, despite intense competition from open source models, Google, Meta and OpenAI. To score a double, it would need to surpass Amazon in valuation, for a triple it would need to pass Microsoft. There’s no path to a 10x in any realistic scenario while the odds of getting wiped out are significant.
That said, it looks like the IPO is going to go well. If that’s the case, it needs to be something that calms down the rest of the market because right now it looks like it could be a tipping point towards a rout.
This article was written by Adam Button at investinglive.com.