LoomCoreWeave's bad day started in September
On October 14, CoreWeave fell 15.5% in a single session. The reasons had been in the news for a month, scattered across stories that didn't look connected.
What happened on October 14
CoreWeave (CRWV) closed at $65.78, down from $77.81 the day before. Over the week, the stock lost 25.6%.
The trigger was the company's bonds, not its stock. Notes CoreWeave had sold only weeks earlier traded down to about 81 cents on the dollar, with yields above 15%. Bloomberg described them as distressed. That evening, Gartner added a second blow: a forecast that 80% of neocloud providers, the GPU-rental companies CoreWeave leads, could disappear by 2030.
A bond yield above 15% for a company this size says something specific. Lenders are no longer pricing CoreWeave like a fast-growing infrastructure business. They're pricing in real doubt about getting paid back. Shareholders stand behind bondholders in line, so when the debt market gets nervous, the stock usually follows.
How the debt got there
CoreWeave builds and rents out data centers full of Nvidia GPUs. That takes a lot of money up front, long before the rental income comes in, and CoreWeave has paid for much of it by borrowing.
In mid-September it borrowed again. On September 17 it launched a $3 billion convertible bond sale. A convertible is a bond that can turn into shares later, so it adds both debt and the chance of future dilution. Demand was strong. The deal priced at $3.7 billion the next day, at a 2.875% coupon, and closed at $4.2 billion on September 22.
At the time, that looked like good news. Investors wanted in, and the deal got bigger. It also meant CoreWeave owed several billion dollars more than it had a week earlier.
Meanwhile, money got more expensive
Two other things happened that same week. Neither was about CoreWeave, but both hit the thing it depends on most: cheap, easy borrowing.
The Fed raised rates on September 17, the same day CoreWeave launched its bond sale. The Information wrote that the hike changed the funding story for AI infrastructure. The logic is simple: companies that grow by borrowing get hit hardest when borrowing gets more expensive.
Three days later, a similar deal ran into trouble. Bonds sold to pay for a new data center, one that the trading firm Jane Street had agreed to rent, struggled to find buyers. CoreWeave had nothing to do with that deal. But it was the same kind of debt, paying for the same kind of building, sold to the same kind of investors. It was an early sign that lenders' appetite for AI data center debt had limits.
It's easy to read stories like these as background and move on. I nearly did. What stopped me was the app I build, Loom: it links each news story to the stocks it's likely to affect, and it put both of these next to CoreWeave. Seeing them side by side with the bond sale made the pattern hard to miss.
Then the stock went up
If those stories were a warning, the market didn't treat them as one. Not right away.
From $83.35 before the bond sale, CRWV climbed to $91.72 by October 6, a 10% gain. On September 24, JPMorgan started covering the stock with a Buy rating. Its argument was that CoreWeave's shift toward shorter contracts would ease fears about how much it was spending. The bull case was alive and well, and for three weeks it was winning.
- 1Sep 17$3B bond sale launched. Fed raises rates.
- 2Sep 20Jane Street data center bonds struggle to sell.
- 3Sep 22Bond sale closes at $4.2B.
- 4Sep 24JPMorgan starts coverage at Buy.
- 5Oct 6High: $91.72.
- 6Oct 9OpenAI revenue recount hits AI names.
- 7Oct 14Bonds turn distressed. Stock −15.5%.
That's worth remembering whenever a story like this gets told backwards. In hindsight the debt looks obvious. At the time it was one argument among several, and the price was saying the other side was ahead.
What changed in October
The first crack showed on October 9. A Bloomberg report put OpenAI's annualized revenue near $50 billion, not the $68 billion that had been widely reported. The gap came from accounting, not lost sales. AI infrastructure names sold off anyway, CoreWeave among them. It was a reminder of how much of this sector depends on a few very large customers growing into very large commitments.
Then came October 14: the bonds, the Gartner forecast, and a 15.5% day.
What I take from it
Big one-day moves rarely come from one day of news. CoreWeave's fall on October 14 was the debt market catching up with a story that had been building since mid-September: more borrowing, more expensive money, and a first sign that lenders were getting picky about AI data center debt.
None of those stories said the stock would fall, or when. The stock rose for three weeks after them. But someone who had read them together would have understood October 14 the moment it happened, instead of finding out the reasons afterwards.
Sources
- CNBC, Sep 17: CoreWeave launches $3 billion convertible debt sale
- The Information, Sep 17: Fed hike changes AI funding story
- The Information, Sep 18: CoreWeave prices $3.7 billion convertible bond offering
- The Information, Sep 20: Jane Street-linked data center debt sours
- The Information, Sep 22: CoreWeave completes $4.2 billion convertible offering
- CNBC, Sep 24: Buy CoreWeave as it leans into short-term contracts, JPMorgan says
- Yahoo Finance, Oct 9: OpenAI's revenue appears $20 billion lighter than reported
- Bloomberg, Oct 14: CoreWeave bonds tumble into distress, drawing "yikes" from Musk
- MarketWatch, Oct 14: CoreWeave sparks worry about broader credit market stress as its bonds tumble
- The Motley Fool, Oct 14: CoreWeave plummets on Gartner neocloud warning and bond yield surge
Prices are daily closes. This is not investment advice.