CoreWeave had a bad week last week. Its bonds have slipped into distressed territory by a common definition: when investors demand a return 1,000 basis points higher than they would get from government bonds. Last Tuesday, Chamath “SPAC King” Palihapitiya noticed the distressed debt and also questioned whether AI development was a private credit bubble. Its shares are down more than 20 percent since Tuesday’s close, which is a lot! There even appears to be collateral damage in the market for other AI companies.
Now I have Also asked whether the AI industry is a bubble, but when the SPAC King echoes the concerns expressed by a lady whose constant question on AI is “How do we plan to afford it, champions?”, that’s a very different signal. When I say “yeah,” that’s one thing. When Elon Musk says it, it’s something else.
When you talk about wanting to tap into the huge amounts of cash that insurers have, that’s the kind of thing that makes investors worry that CoreWeave will borrow even more.
To make CoreWeave’s situation worse, Brannin McBee, one of the company’s co-founders, told the Financial Times — Also last week, the same day as Palihapitiya’s message – that the company wants its debt to be investment grade, rather than junk, so it can borrow from insurers because their capital is, in his words, “unlimited.” That’s not entirely unreasonable: Earlier this year, CoreWeave secured its first investment-grade chip loan. But when it comes to wanting to tap into the huge sums of cash available to insurers, it’s the kind of thing that makes investors fear that CoreWeave – already a massive debtor – will borrow even more.
“The problem is not that credit profiles have changed over the last year, but that we still have a year and a half of supply to come,” Bloomberg John Yovanovic, co-head of leveraged finance at MetLife Investment Management, was quoted as saying.
I’ve seen people suggest that this means the AI bubble is bursting. I’m not so sure. A fall in stock prices is not necessarily a financial catastrophe. Stocks are an expression of confidence in the future of a company and not a reflection of day-to-day activities. But CoreWeave has a lot of very expensive debt, and if that debt becomes more expensive, that makes it even harder for CoreWeave to turn a profit.
CoreWeave is the poster child for AI debt, and as of its second-quarter earnings, it owed more than $35 billion. That’s partly because it was the first to use GPUs as collateral for loans, and partly because CoreWeave has no real business beyond selling compute, a business it doesn’t make money on — the company lost more than $1 billion in the first half.
Investor concerns around CoreWeave are in fact present for all neoclouds
So why am I not ready to take a victory lap on this? After all, I suggested that this business was not sustainable in the long term. The answer is “the remaining performance obligations” of its customers – including $103.7 billion in the second quarter. About 40% of this money is due to CoreWeave in 2028, and companies that agreed to these contracts can only cancel them in very limited circumstances. Its biggest customers are Microsoft, OpenAI and probably Nvidia, although CoreWeave does not reveal who its third customer is. These three companies generated almost three-quarters of the company’s revenue in the second quarter.
Microsoft and Nvidia are probably good value for money – at least, as long as it benefits them. (I suspect they realize that if CoreWeave goes bankrupt, there will be a lot of collateral damage in the market, so they’re unlikely to push the company over the edge themselves.) OpenAI is more fragile, but unless OpenAI is in much greater financial distress than we all think, CoreWeave will probably be fine for a while. As long as she can continue to repay her debt, the music continues. Only when the AI boom peters out will we know whether this business is truly viable.
Still, the angst around CoreWeave has made the situation fragile for other companies that have taken on significant debt as part of AI development — Oracle shares, for example, fell nearly 10% during the same period. There was also a sell-off in Oracle bonds. Investor concerns around CoreWeave are in fact present for all neoclouds. It’s just not clear whether the bare metal business is a good business.
Either way, some people buy the dip.
Gn bussni
