Overview
There are two sections today: a longish Rough Notes update on my recent data center comments, from cost of capital to futures, followed by some related (and unrelated) reading in Rougher Notes.
Rough Notes
The Kanye/Data Center Crossover
I promise to move on to other topics soon—really!—but I've been overwhelmed with interest in my prior two pieces (here and here) on AI-related capex. I have done a dozen media interviews, some podcasts (at least one of which is out), had various conference speaking invites (yeah, no), and been reminded why I only do selective consulting to banks and hedge funds. There have also been a bazillion glosses elsewhere on the piece, some of which I've had a chance to read, but not all.
My takeaway is that the level of interest in this topic is, as they say in the Tour de France, hors categorie, outside of the categorization system, crossing from business and geek fixations into popular culture fascination. To put in a kind of context, I sometimes compare things in units of "Ye"s. Is the topic more interesting to people than is Kanye West?
Well, data centers are about 1.25 Ye's, which is effing nuts:
Given the data center/Ye crossover, this seems like a good time to summarize the most common questions I get and my answers to them.
Why did you write about AI capex?
Like when writing about most things, I wrote it because no one else would. Writing is annoying, and so, having noticed that AI-related capital expenditures had become large enough to affect economic data, I assumed someone else would say it first. When no one did, or at least I didn't notice anyone doing so, I did.
Of course, there is more to it than that. I wrote about AI capex because it seemed important that so much money was flowing so quickly into such a narrow slice of the economy. The size is rapidly exceeding prior capital expenditure waves, from the fiber buildout in the dot-com era, to the railroads in the 19th century. That is remarkable.
