AI Bubble Burst Bonanza

Listen to Free Audio Version Here

I have watched a lot of gold rushes from the cheap seats, and they all end the same way. Somebody finally asks who is paying for the shovels. In the AI boom, that question just got asked out loud by the man with the most shovels of all.

On a Sunday in late October 2026, Elon Musk answered a post on X from venture capitalist Chamath Palihapitiya, who had been grumbling that the specialty AI cloud companies were “disorganized and poorly capitalized.” Musk did not argue. He wrote, “As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers.” Then came the line that will be quoted at every funeral in the industry: “That means the 200+ neoclouds are going to see guppy-level death and consolidation. Some will make it through, but not many.”

For those who do not spend their evenings reading credit spreads, a neocloud is a company that borrows a mountain of money, buys a mountain of Nvidia chips, stacks them in a warehouse the size of a small county, and rents the brainpower out to the AI labs. It is a fine business as long as two things stay true: the renters keep paying, and the bankers keep lending. Musk is saying the second leg of that stool just got sawed off.

The kicker, and there is always a kicker, is that he said all this in the same breath as announcing that SpaceX had closed forty billion dollars of debt to buy its own GPUs. The lenders were happy to write that check, he explained, because SpaceX is investment grade and has Starlink and rocket launches paying the bills. So the bank door is not locked. It just has a velvet rope now, and the bouncer only waves in people who already own a rocket. When someone online suggested he was talking the neoclouds down so he could scoop them up cheap, he shot back that SpaceX is not buying any neoclouds. “Pointless.” That is the corporate version of looking at a drowning man and remarking on the weather.

I want to be careful here, because I do not guess. Musk did not say a group of banks held a meeting and blacklisted the cloud companies. He said debt markets, which is the whole lending machine: banks, bond buyers and the private credit funds that have been bankrolling this buildout. And the warning signs were there before his post. Banks such as JPMorgan, Morgan Stanley and SMBC have spent months trying to slice up and sell off their data center loans because, as one risk manager told the Financial Times, at these sizes “banks very quickly start choking.” In mid-October, CoreWeave, the poster child of the neocloud crowd with roughly fifty billion dollars of debt, watched its bonds slide to distressed levels, with yields near fifteen or sixteen percent and the cost of insuring its debt against default climbing to about twelve percent a year. When the insurance on your loan costs more than a credit card, the market is trying to tell you something politely.

The stocks heard the message. CoreWeave, Nebius and IREN each dropped more than four percent before the bell on the Monday after Musk posted, piling onto a month in which CoreWeave and Nebius had already shed more than thirty percent and IREN about forty. Gartner had already forecast that more than eighty percent of these specialist providers will stop operating independently by 2030. Chamath, for his part, predicts a year of “partially stranded compute,” chips bought and paid for and sitting in boxes, collecting dust instead of revenue. There is nothing quite like a processor that costs more than a car being used as a paperweight.

So is this the AI bubble bursting? Plenty of loud voices online are saying so. This is where high-end pondering earns its keep, because I think the honest answer is: not the whole bubble, just the bubbles inside the bubble. The giants, Amazon, Microsoft, Google, all have other businesses throwing off cash, and they will keep building. What is popping is the idea that anybody with a pitch deck and a loan officer could become a cloud company. Even one careful analysis making the rounds concluded that calling the whole AI bubble burst goes further than the record supports. The IMF has warned that the money linking these companies together can carry a shock from one to the next, but a warning about a fire is not the same as the fire.

Here is what I keep coming back to. The machines themselves did not fail. The AI got smarter right through all of this. What failed was the old human habit of borrowing against tomorrow as if tomorrow had already signed the paperwork. We built thinking machines on the oldest dumb idea in finance, and the thinking machines were not consulted.

There will be consolidation, which is a polite word for the big fish eating the little fish. There will be fire sales of chips, and some lucky bargain hunter will build something remarkable on hardware bought for pennies on the dollar. That is how every boom ends, from railroads to the dot-com crash. The tracks stay. The companies that laid them go under. The future gets built anyway, just with a different name on the deed.

So if you own one of these neocloud stocks, keep the little net handy. And if you are a guppy, swim close to a rocket.

C. Rich