First off, this is a stolen post by Ben Solar on Shrewdm.com.
Krugman: The last number you had was a little bit bigger as a share of GDP than the telecoms boom of the 90s. But presumably you think it’s higher than that now?
Kedrosky: I do. It’s something like, nonresidential fixed investment, probably around 14% now. So, we’re considerably ahead of where we were in the telecom bubble, we are somewhere between rural electrification and World War two rearmament.
Krugman: But not yet like railroads in the 19th century.
Kedrosky: Not yet like railroads, but on a path to a similar place. Given that—and this is really important—we’re in that point where there’s a financial flywheel now, where increasingly the financing of these data centers is in a place where it’s somewhat divorced from what goes on inside the data center, because we’ve created a financing template for how to finance data centers, where you have these SPV, these special purpose vehicles into which a third party contributes capital and the tech company contributes technology. And out the other side magically pops these securities that have great income and yield characteristics that are hugely attractive to investors. They look at it as almost like a synthetic security, where I understand it’s the SPV, the data center that’s producing this.
But on the other side of this is Meta and Google and they’re a prime credit. They’re a really strong credit. So they’re going to keep paying on this. I don’t really care what goes on inside the data center, because I have a lot of confidence in the counterparty in this. We know where all of this kind of thing leads when you have these financing flywheels driven by securitization and high yields and people not caring what goes on inside the actual structure itself: it leads to a lot more construction and eventually over-building.
Krugman: Oh, God. I’m having flashbacks to 2008, 2009. All of the stuff that was “perfectly safe” because after all, AIG was backing it, right?
Kedrosky: Right, exactly.
XXXXXXXXXXXXXXXXXXXXXX
Kedrosky historically, the U.S. has been very good at speculative bubbles. This is one of our main core competencies here. They tend to be about real estate, or they tend to be about technology, or they tend to be about loose credit, and sometimes they even have a government role with respect to some kind of perverse incentive that was created. This is the first bubble to have all four. We’ve got a real estate component, we have a loose credit component, we have a technology component, we have a huge government component, because we’re told “we’re in an existential crisis with China, and we must win this at any cost.”
Kedrosky: The sum of all bubbles.
tj-Ah the MOTHER of all bubbles.
Kedrosky: We have this prodigious amount of spending going on,
scarcely a ripple in terms of people saying, “hello, we’re running a giant private sector stimulus program that’s keeping the US out of recession,” and yet no one’s talking about it in those terms.
And while the bubble is expanding & sucking in most of US credit sources other business sectors are in starvation mode detrimental to them. The reindustrialization of the USA by re on shoring manufacturing has just become almost impossible.
This interview revealed another factoid. Since the data center is running 24/7 the chips burn out in ~2 to 3 years. So constant replacement of those are needed.
And then Kedrosky points bad government data on AI.
Kedrosky the Census Bureau added a line on AI adoption, trying to be helpful back in 2022. It’s showing that AI adoptions actually began plateauing at around 18% of large corporations already in the third quarter of 2025, which seems ridiculous, obviously, for a host of reasons.
An ongoing problem in general is that a lot of our economic statistics are really designed for the economy of 1929.
I pointed this out about measuring poverty in my post “The Measuring Stick is Outdated”
How can a government run a country its data is outdated. Garbage In/Garbage Out.