CoreWeave’s credit market is flashing a warning that would have been hard to imagine a year ago. On Oct. 15, the cost of insuring the AI cloud provider’s debt against default for five years hit a record of roughly 1,195 basis points, or about $119,500 a year to protect $1 million of bonds. Market models translate that into better-than-60% odds of a default within five years. The next day the stock fell almost 12%, to $61.39.
Our read: the market is pricing a balance-sheet problem that is real, but it is also caught in a feedback loop that makes the danger look bigger than it is. Below, we explain why AI sentiment turned against CoreWeave in particular, what its finances actually look like, how the “doom loop” between its stock, bonds and credit insurance works, and what would have to happen for the default math to come true.
What CoreWeave does, and why it is in the crosshairs
CoreWeave is the largest of the so-called neoclouds: companies that buy Nvidia’s AI chips by the tens of thousands, put them in data centers and rent the computing power to AI labs and big tech companies. It reported a revenue backlog of more than $100 billion at the end of June, and Nvidia, its main supplier, also owns about 10% of the company.
Where that capacity sits matters. Data-center markets are graded by tier: Tier 1 is a major metro area, Tier 2 a secondary or suburban market, and Tiers 3 and 4 are rural sites with cheap land and power but more distance, and therefore more delay, between the servers and the people using them. That delay barely matters for training a model, which can run for weeks; it matters much more for answering users in real time.
In a September note, Bernstein estimated that about 25% of CoreWeave’s active U.S. power and about 74% of its contracted power sits in Tier 3 and Tier 4 markets, and called it the most exposed name in its coverage if spending on training new AI models slows. That note followed public comments from AI leaders, including Anthropic’s chief executive, in favor of a slower pace of frontier AI development.
Two shifts in AI that work against CoreWeave
The first shift is training fatigue. If the biggest labs train fewer giant models, or train them less often, the rural capacity built for training is the first to feel it. We think training demand will stay large for years, but markets trade the direction of the next quarter, not the next decade.
The second is the move toward AI agents and running models in-house. Much of the industry’s excitement has moved from building ever-bigger models to agents that carry out tasks, and many of those tasks use small models that run on a laptop or a single workstation. Even large open models that once needed a cluster of data-center GPUs can now be run, more slowly, on a handful of high-memory desktop machines that cost tens of thousands of dollars rather than hundreds of thousands. Every company that decides to run AI on hardware it owns is one less customer renting GPUs by the hour.
Neither shift means demand for rented AI computing disappears. Together, they mean investors are less willing to pay up for the kind of capacity CoreWeave has the most of, at the moment its finances leave the least room for disappointment. The data below shows how the stock has traded over the past six months and when company insiders sold.
The chart below tracks CRWV's daily closing price from April 22, 2026 to Oct. 16, 2026. Over that stretch it is down 49.9%, from $122.54 to $61.39, with a high of $137.98 on May 6, 2026 and a low of $60.82 on July 29, 2026. Each red dot marks a day company insiders sold shares on the open market (74 days in this period). There are no green dots: no insider bought on the open market in this window.
The largest open-market trades by CoreWeave, Inc. insiders in the last 90 days, from their SEC Form 4 filings:
| Date | Insider | Role | Trade | Filing |
|---|---|---|---|---|
| 2026-08-13 | Magnetar Financial LLC | 10% Owner | sold $45.9M | SEC filing |
| 2026-08-12 | Snyderman David J. | 10% Owner | sold $39.6M | SEC filing |
| 2026-08-13 | Magnetar Financial LLC | 10% Owner | sold $23.1M | SEC filing |
| 2026-08-12 | Snyderman David J. | 10% Owner | sold $22.6M | SEC filing |
| 2026-08-12 | Snyderman David J. | 10% Owner | sold $19.9M | SEC filing |
The balance sheet is the real issue
CoreWeave spends far more than its operations bring in and closes the gap by borrowing and selling stock. In the second quarter, according to Bernstein, it generated $679 million of operating cash flow against about $9.4 billion of adjusted capital spending, and its net interest expense of $640 million came to about a quarter of revenue. Bernstein counts about $35 billion of borrowings plus $16.5 billion of lease obligations at the end of June. The company’s 2026 capital spending budget is $31 billion to $35 billion, and in September it sold $3.7 billion of convertible notes.
That model works as long as two doors stay open: investors willing to buy new shares, and lenders willing to fund GPUs on acceptable terms. Bernstein, which rates the stock Underperform with a $74 price target, described the selloff as a sharper repricing of a financing risk that already existed rather than a reaction to new operating news. We agree with that framing. Nothing broke last week; the market simply became less willing to look past it.
Our dilution gauge, built from CoreWeave, Inc.'s SEC filings, reads high (95/100); the share count rose 41.4% over the past year.
CoreWeave, Inc.'s recent offering-related SEC filings:
| Date | Form | What | Filing |
|---|---|---|---|
| 2026-09-22 | 8-K | Convertible notes (private offering) · with a capped call that limits dilution up to a cap | SEC filing |
| 2026-09-17 | 8-K | Convertible notes (private offering) · announced, not yet priced · with a capped call that limits dilution up to a cap | SEC filing |
| 2026-09-17 | 424B5 | At-the-market program (common stock) (prospectus) | SEC filing |
The doom loop, step by step
A credit default swap, or CDS, is insurance on a company’s debt. The buyer pays a yearly premium and gets paid if the company defaults. Its price rises when the market worries more about a default, but it also rises when a lot of investors want that protection at the same time, whether or not anything changed at the company that week. Since 2008, rising CDS prices have carried an emotional charge that the numbers alone don’t justify.
For funds that want to hedge their exposure to the AI boom, CoreWeave’s CDS has become one of the most direct ways to do it. That creates a loop that feeds on itself:
- The stock falls on AI and funding worries.
- A lower share price makes raising money by selling stock more expensive: every dollar raised means giving up a bigger slice of the company.
- Bonds slip as investors price in more risk. On Oct. 15, CoreWeave’s 9% notes due 2031 traded as low as about 80 cents on the dollar, a yield of roughly 15% to 16%, which makes any new borrowing more expensive.
- Funds hedging AI exposure buy more protection through CDS, pushing its price to new records.
- Record CDS prices are reported as record default odds, which unsettles stock and bond holders, and the cycle starts again.
It is the GameStop story in reverse: instead of crowded buying pushing a stock far above what the fundamentals justify, crowded hedging is pushing the market’s estimate of default risk above what we think the fundamentals justify. The balance sheet is weak; the loop makes it look weaker.
Why we think 60% default odds are too high
A default probability backed out of CDS prices is not a forecast. It depends on an assumed recovery rate, and it includes the extra premium investors demand for uncertainty and for the scramble to hedge. When everyone wants the same insurance at once, the price of insurance overstates the odds of the fire.
CoreWeave also has more ways to raise money than the loop suggests. The stock still trades well above its $40 IPO price from March 2025, and a company worth tens of billions of dollars can still sell shares, even at prices existing holders won’t like. Much of its backlog is take-or-pay contracts, which customers must pay for whether or not they use the capacity. Its GPUs can secure asset-backed loans. And its largest strategic partner is publicly in its corner.
Speaking in Stockholm on Oct. 15, Nvidia chief executive Jensen Huang said CoreWeave’s strength is improving every day, that he is optimistic, and that its cost of capital is actually decreasing; he also said Nvidia does not need to backstop it. We would push back on the cost-of-capital point: when a company’s bonds trade near 80 cents on the dollar, new borrowing gets more expensive, not less. But the support matters. The funds that report the largest holdings are listed below.
The largest reported holders of CRWV in SEC Form 13F filings (Q2 2026):
| Fund | Value | Last quarter |
|---|---|---|
| Magnetar Financial LLC | $5.2B | cut (-24.1%) |
| Nvidia Corp | $4.7B | unchanged (0.0%) |
| BlackRock Inc. | $2.4B | added (+157.0%) |
| Goldman Sachs Group Inc | $1.8B | added (+1.3%) |
| Vanguard Portfolio Management LLC | $1.7B | added (+34.3%) |
| Vanguard Capital Management LLC | $1.6B | added (+26.9%) |
| Jane Street Group, LLC | $1.4B | added (+56.8%) |
| Invesco Ltd. | $1.2B | added (+1149.5%) |
| Alyeska Investment Group, L.P. | $1.1B | added (+55.7%) |
| Value Aligned Research Advisors, LLC | $1.1B | added (+31.5%) |
Our view: dilution is likelier than default
Before a company like CoreWeave defaults, it usually has a long list of calls to make: new share sales, convertible bonds, asset-backed loans on its GPUs, help from partners with deep pockets. Each of those has a cost, and most of that cost lands on shareholders, whose stake gets diluted, before bondholders lose money. That is the outcome we think the market is underpricing relative to bankruptcy.
If AI spending does turn into a broader market problem, we think the earlier warning will come from the wider economy, especially hiring and the labor market as AI agents improve, rather than from one company’s credit spreads. In that sense CoreWeave is a sideshow: a stressed borrower caught in a hedging loop, not the black swan that ends the AI cycle.
Read our related page on why CoreWeave, Nebius and IREN fell, and the live CRWV page for insider trades and filings as they are reported.