“The Economics Are Abysmal”

Submitted by QTR's Fringe Finance

I’ve long admired Harris Kupperman, the founder of Praetorian Capital, for his ability to cut through noise and spot big-picture themes before they become consensus.

Harris just released his thoughts on markets, AI and his positions in his Q3 investor letter, which I’m happy to share with my subscribers. Harris warns that the long-awaited AI reckoning is finally here, and a broader market crash may be just beginning.

In his latest Q3 letter, Harris explores the increasingly unsustainable economics of the AI boom, the potential for a wave of bankruptcies to ripple through financial markets and the broader economy, and why he’s positioning his portfolio for a major downturn by raising cash and turning to U.S. government bonds.

He also reviews his current long positions and explains why he remains bullish on precious metals, sees significant upside in one little talked about sector as AI reshapes the labor market, and talks undervalued real estate assets and emerging markets in the future. (Photos, charts and article links added for emphasis by QTR).

“Highly Unlikely” AI Reaches Profitability As Is

I like to say that most things in finance are rather obvious; it’s the timing that’s difficult. In that vein, it’s always been rather obvious that the economics of AI are absolutely abysmal.

I think it’s highly unlikely for the industry to ever reach profitability as currently configured, as it would need a few percent of US GDP annually in revenue to break even on the capital already invested, and may need as much as ten percent of GDP to break even on the capital that many analysts expected to be invested by 2030. The numbers just cannot work—especially as tech tends to be deflationary.

How would anyone ever expect the US to be spending a few trillion dollars annually on AI, when software, which is already ubiquitous, and maturing, is currently around half a trillion of US spending?

Think back to the railroads; they became dominant modes of transport, but they periodically destabilized the US economy as they went bankrupt in repeated waves during the 1800s. Then, someone picked up the pieces, and railroad construction continued. I think we’re at a similar point in the AI journey. I definitely don’t think that AI is over, or that the buildout won’t continue. I just think a phase of rationalization is upon us.

It’s said that you go broke slowly, then all at once. I think this is a good mindset for how the AI sector will also fail. Slowly, then all at once.

The first signpost will be that the funding gets cut off, and I think we’re finally witnessing that. As funding grows more scarce and expensive, the AI sector will struggle to support unprofitable operations, and I expect to see many bankruptcies of neoclouds, followed by the large frontier AI labs. While they may have interesting assets, they also have hundreds of billions of contingent liabilities along with purchase commitments that I cannot imagine anyone wants to fund, especially if chip costs come crashing down with reduced demand.

Following bankruptcy and the rejection of these liabilities, I then assume that they get folded into the hyper-scalers.

However, these will be incredibly messy and complex bankruptcies. Just think back to how messy Enron was, with hundreds of SPVs and all sorts of future commitments over pipelines, utilities and whatnot. This will be worse, as many of the SPVs have...(READ THIS FULL LETTER HERE).