CoreWeave’s sudden sucker punch from the financial markets has come as a surprise to many. Not to AI critic Ed Zitron, who has been predicting a decline in either CoreWeave or Oracle (that’s already started) for well over a year. If you weren’t paying attention to the stock market this week, the compute provider’s stock dropped 15% this week (at the time of writing), despite Nvidia helping bump it back up by 6%.
This has led to other outlets, which tend to boost AI, taking note of how financially unsustainable the whole enterprise is. I’ve been pointing this out for some time. A trillion dollars in expenditure (this year alone) for an… interesting technology only makes sense if it sees massive (paid) adoption. LLMs in general can’t command that sort of usage. Outside of some very specific areas, most folks just aren’t interested in what AI has to offer.
CoreWeave doesn’t have any involvement in the training, development, and release of artificial intelligence. Nor does Oracle. Both companies build infrastructure. Both are also the first at risk for any payment shortfalls for what they’re creating. The reason for this is debt. Lots and lots of debt.
CoreWeave another tale
Reuters quotes Max Frumes, head of distressed and leveraged finance at 9fin, as calling CoreWeave “the canary in the coal mine.” The thing about the canary is that it’s supposed to die first.
“This is going to be a turning point in the pricing and availability and structure of credit that’s supporting the AI boom. It’s going to get more expensive. There will be a lot more scrutiny and probably some pushback from investors.”
The bulk of the AI industry, which means OpenAI and Anthropic, is propped up by investor money. Advertising, subscriptions, and other income streams barely make a dent in the overall cost needed to keep your favourite AI model up and running. CoreWeave has been identified as a possible point of failure for ages. Now, the value of the company’s bonds has dropped sharply, which means it’ll cost more money to complete the projects it’s contracted for. As with Oracle’s recent ‘force majeure’ declaration, this will have a knock-on effect.
When OpenAI attempts to fundraise, it’s going to cost more. Anything new that AI companies offer will similarly be more expensive. These costs, much like the industry’s circular financing, will be passed around until everything gets more expensive for very little tangible benefit.
AI go boom
CoreWeave isn’t dead yet, but it’s not far away from it either. The current decline means it’ll have more trouble raising debt. The money it does raise will come at a greater cost. And if any of its major clients fail to raise the funds they need to operate, CoreWeave will explode. Explaining exactly why will take entirely too long. Subscribe to Ed’s newsletter for an exhaustive breakdown of exactly why CoreWeave and Oracle are ticking time bombs. If either company — most likely it’s both — goes down, they’ll probably be beyond rescue. Nothing will be left.
If that happens, OpenAI and Anthropic (one will have started this mess by missing payments to these compute providers) are in massive danger. There’s a chance other industry players will come to the AI giants’ rescue, but that too puts them at risk. Google and Microsoft might take a shot anyway, but they could also choose to let their competitors die and pick over the carcasses. Meta probably won’t rescue either AI company. Amazon definitely won’t.
In classic bubble fashion, entire industries will be wiped out. SoftBank, which has invested extremely heavily in OpenAI, will almost certainly disappear. Smaller companies — Character AI is already cooked — built on AI will disappear. Other industries that have shifted to heavy AI reliance will also take a serious knock. The financial implications are enormous, but also uncertain. You can’t have a handful of companies spending trillions of dollars on a technology without attempting to rescue that investment. What that rescue looks like is the tricky bit.
OpenAI goes hunting
OpenAI has launched into a high-speed funding round, according to unnamed sources. The reason given for this usual $30 billion round, according to those sources, is to get ahead of competitor Anthropic’s pending initial public offering. That’s almost certainly part of it, but the recent CoreWeave performance is probably also a factor.
If the value of CoreWeave’s debt falls, expenses (and rates) for OpenAI also go up. If that happens, it makes paying CoreWeave harder, which could hasten the compute provider’s downfall. There’s one more risk — if Anthropic’s IPO goes poorly, investors will be less likely to hand over cash to OpenAI.
A very similar situation is playing out in China over humanoid robots. The country’s darling, Unitree, recently listed on that country’s stock market. Initial enthusiasm has been dampened by falling stock prices, eventually leading the country’s regulator to slam the brakes on similar IPOs. The approach might work for physical tech like robotics, but AI development is far too resource-intensive to allow the money hose to stop pumping even for a short while.
CoreWeave probably isn’t in immediate danger, but the warning signs are blinking red at present. And if that company goes, others will follow. Oracle, SoftBank, and other over-leveraged entities are on the chopping block. Anthropic and OpenAI are also at serious risk. Google and the rest of the big players stand to take serious hits, but they’ll probably emerge from the other side. What they’ll look like when they do, who can say?
