TechTakes

Yes, but the article's not actually about that. It's about Microsoft returning to the same datacenter-building schedule from a decade ago. Datacenters have a lag of about 3-5yrs depending on what's inside them and where they're located, so what we're actually seeing is Microsoft projecting a relative reduction in overall usage. Note that among all the cancellations of notes and prospective claims, Microsoft isn't walking back their two-decade nuclear-power deal with Westinghouse; they're not destroying or reducing any existing capacity, just planning to build less. At risk of quoting Bloomberg:

After a frantic expansion to support OpenAI and other artificial intelligence projects, [Microsoft] expects spending to shift from new construction to fitting out data centers with servers and other equipment.

To the extent that the bubble is popping, Microsoft and other datacenter owners have to guess half a decade in advance when the bubble will pop, and if you take them at their word — that is, if we assume that they canceled these contracts with perfect foresight — then the bubble must have already popped in 2023-2024, and the market is experiencing coyote time because…? More likely, this is fallout from their ongoing breakup with OpenAI, who almost certainly begged Microsoft for so much compute (and definitely begged for too many nVidia GPUs!) that Microsoft had to adjust their datacenter plans. The bubble's not done until OpenAI has exhausted all possible funding, say in late 2025 or early 2026 when Softbank and the Saudis realize that they've made a hilarious mistake.

We've discussed this previously on awful.systems, both the value of nuclear-energy contracts and Microsoft's retraction of intents.

You're dead right on that.

On a wider front, I expect the bubble's burst will deal some long-lasting damage to the tech industry, on two major fronts:

1. Evaluations/stock prices will go through the floor, as investor assumptions of Endless Growth^tm^ are shattered, and the tech industry comes to be seen as "a stable, mature industry whose days of endless growth are behind it" (quoting Baldur Bjarnason), at best and deep in the throes of a serious malaise era at worst. One outcome is obviously better than the other, but both will lead to a lotta pain.

2. The tech industry's public image, already in the shitter for a litany of reasons, will likely take another significant blow as the industry's ability to generate hype for AI and/or counteract the boiling resentment towards it completely falls apart.