Some disconcerting facts about AI and banking that may have profound consequences

Disturbing new numbers

Haunting tweet by the author, 13 months ago

Back in November, I wrote an essay called A trillion dollars is a terrible thing to waste. It was already clear by then that generative AI was probably going to lose a lot of money. Although there is some clear value in it, the costs are high, and relative to those costs, the revenue has been modest. The big problem, which has persisted since the beginning of LLMs, is that the reliability is not great, and hence returns on investment for most big corporate customers have been low (as multiple studies have now shown).

What I couldn’t figure out then was how bad the consequences might be if things fell apart. I pointed to the exposure of the banking industry as a key unknown (boldface added):

But it’s not just that a trillion dollars or more might go down the drain, but that there might be considerable collateral damage, to the rest of society, both economic and otherwise (e.g., in terms of how LLMs have undermined college education). As Rogé Karma put in a recent article in The Atlantic, “The entire U.S. economy is being propped up by the promise of productivity gains that seem very far from materializing.”

To be fair, nobody knows for sure what the blast radius would be. If LLM-powered AI didn’t meet expectations and became valued less, who would take the hit? Would it just be the “limited partners” like pension funds who entrusted their money with VC firms? Or might the consequences be much broader? Might banks go down with the ship, in 2008-style liquidity crisis, possibly forcing taxpayers to bail them out?

Would it be just investors who are screwed, or would it be everyone, indirectly? The question hinged in part on how much banks stood to lose if Generative AI lost steam.

This morning, Steven Rosenbush, in his newsletter for The Wall Street Journal, put together a bunch of facts about AI and debt that should concern everyone, quoting something from yesterday in Bloomberg that I missed (italics) and putting context around it:

As Bloomberg reported Monday:

Morgan Stanley expects hyperscalers to borrow $400 billion this year, up from $165 billion in 2025. The offering spree will likely drive high-grade debt issuance to a record $2.25 trillion this year, Vishwas Patkar, head of US credit strategy at the bank, wrote in a note on Monday.

….
As the WSJ said on Saturday, the data center buildout is “bigger than the railroad expansion of the 1850s, the Apollo space program that put astronauts on the moon in the 1960s and the decadeslong build-out of the U.S. interstate highway system that ended in the 1970s.”

I have a hard time seeing how this doesn’t end badly. A recession, maybe worse.

Let me add that I do think that some form of AI will eventually actually yield the returns people were anticipating for Generative AI. But the unreliable statistical mimicry of Generative AI is unlikely to be the ticket to those massive productivity gain. Instead, the misguided obsession with large language models will likely cause a lot of pain over the next few years, economically and socially, until a new, far more reliable form of AI is developed.