An Unusually Prudent Bet on Obsolescence

It is a well-documented quirk of the human condition that whenever we find ourselves in possession of a truly staggering amount of money, our first instinct is to find someone else who will let us borrow even more. This is currently the state of play in the world of "hyperscalers"—those vast, cloud-dwelling entities that have decided the only thing standing between humanity and a digital utopia is a few hundred billion dollars' worth of graphics cards and a cooling bill that could power a small moon.

JPMorgan, an institution that has spent the better part of two centuries watching people make questionable decisions with large sums of money, has recently introduced a rather elegant solution for those who suspect the current AI boom might be less of a "new era" and more of a "very expensive mistake." They have launched a Credit Default Swap (CDS) basket. For the uninitiated, a CDS is essentially a way of taking out an insurance policy on someone else’s debt. It is the financial equivalent of betting that your neighbour’s new, gold-plated conservatory will eventually cause their house to fall into a sinkhole.

The neighbours in this instance are the five horsemen of the silicon apocalypse: the hyperscalers who are currently engaged in a borrowing spree that would make a lottery winner blush. They are taking on debt at an unprecedented rate to finance the construction of data centres—vast, windowless cathedrals dedicated to the worship of the Large Language Model. These buildings are filled with rows upon rows of GPUs, humming with the quiet, expensive intensity of a thousand hair dryers, all working together to ensure that a teenager in Nebraska can generate a picture of a cat wearing a tuxedo in under three seconds.

There is something inherently whimsical about the scale of this ambition. We are, as a species, currently betting the stability of the global financial system on the hope that if we feed enough electricity into a box, it will eventually start saying things that are both true and profitable. It is a bold strategy, akin to trying to teach a toaster to write poetry by plugging it into a nuclear reactor.

JPMorgan’s new basket allows investors to hedge against the risk that these companies might, at some point, find themselves unable to pay back the money they borrowed to buy the boxes. It is a remarkably polite way of acknowledging that while we all hope the digital future is bright, it wouldn’t hurt to have a sturdy umbrella in case it turns out to be made of lead.

The narrator finds it particularly charming that we have reached a point where the "risk" being hedged is not that the technology won't work—it clearly does, in its own hallucinatory way—but that the sheer cost of maintaining the hallucination might eventually outstrip the desire of anyone to pay for it. It is the ultimate triumph of the accountant over the visionary. The visionary sees a world transformed by intelligence; the accountant sees a quarterly interest payment that is starting to look a bit like a telephone number.

One cannot help but admire the bureaucratic neatness of it all. On one side of the ledger, we have the engineers, frantically trying to prevent their servers from melting through the floor. On the other, we have the financiers, sitting in quiet, air-conditioned rooms, carefully constructing baskets of risk that allow them to profit whether the servers melt or not. It is a beautifully balanced ecosystem of mutual suspicion.

There is a certain comfort in knowing that no matter how advanced our technology becomes, the fundamental mechanics of human greed and anxiety remain unchanged. We will build the future, yes, but we will also make sure there is a secondary market for betting on its failure. It is the most human thing about the entire AI revolution.

As we move forward into this brave new world of agentic spreadsheets and sentient search engines, we should perhaps take a moment to appreciate the humble CDS basket. It is a reminder that in the grand theatre of progress, the most important person is often the one selling the insurance policies in the lobby. After all, if the play turns out to be a disaster, at least someone will be able to afford the taxi home.

Fictionalised reflective observation: I once saw a man try to insure a sandwich against the possibility of it being eaten by a seagull. The premium was higher than the cost of the sandwich, but he felt a profound sense of peace knowing that his lunch was protected by the full weight of the insurance industry.

The current AI infrastructure boom feels remarkably similar. We are building a sandwich of unimaginable complexity, and JPMorgan is simply providing the seagull insurance. It is a prudent, if slightly cynical, arrangement. And in a world where algorithms are increasingly in charge of our destinies, a little bit of cynicism is probably the only thing keeping us sane.