CoreWeave Default Swaps Top 1,000bp as AI Credit Stress Hits Neoclouds

The price of insuring CoreWeave's debt against default went through 1,000 basis points in mid-October 2026, and as of 10 November it had not come back. On 15 October Reuters reported that CoreWeave's five-year credit default swaps had blown out past 1,000 basis points. Reuters said that level implies a default risk of nearly 63% over five years, up from about 30% a year earlier. On Friday 6 November, after the midterm chip selloff, Reuters reported the spread was still above 1,000, according to Decrypt. CoreWeave reports third-quarter results after the close today, and those numbers now get read through the bond market as much as the stock.

What happened

The break came on Wednesday 14 October. CoreWeave shares fell 15% in regular trading and another 6% after hours, Reuters reported. Reuters tied the drop to hotter-than-expected US inflation data, rising rate expectations and doubts about AI profits. The same day brought a Gartner forecast that more than 80% of neocloud providers will stop operating independently by 2030. Reuters put CoreWeave's debt at nearly $50 billion, citing LSEG data. Its 2032 bonds traded at around 78 cents on the dollar, a yield of just over 16%.

The rest of AI credit moved too, but much less. Oracle's five-year CDS was at 288 basis points on 15 October, up from 260 a week earlier, and SoftBank's widened to 424. Reuters reported that Amazon, Alphabet, Meta and Nvidia CDS hit record highs, but rose far more slowly. Broadcom chief executive Hock Tan said that week that the investor rush into anything AI was like "casino gambling".

Stocks bounced the next week. Credit did not. On 19 October CNBC reported that CoreWeave's 9.625% bond due 2032 was yielding 14.9% and its CDS sat at about 1,082. Oracle's CDS had widened to 315 basis points, up 311% from a year earlier, on about $132 billion of net debt and a BBB- rating. On 23 October CoreWeave's 9.75% notes due 2031 yielded 15.637%, with the CDS at 1,129, per CNBC. Over the month, Reuters reported (as cited by Network World) that CoreWeave, Nebius and IREN shares each fell more than 30%.

The numbers

Why it matters

Equity investors can wait for a story to play out. Credit investors care about whether the cash arrives before the coupon is due. CoreWeave's own numbers show why they are nervous. Second-quarter revenue was $2.575 billion and the net loss was $626 million, with net interest expense of $640 million, up from $267 million a year earlier, according to the company's release. For the third quarter, management guided on the August call to adjusted operating income of $200 million to $260 million against interest expense of $860 million to $940 million, per MarketBeat's summary. At the guided ranges, interest costs are several times operating income.

The worry is not only about CoreWeave. David Rosenberg of Rosenberg Research told CNBC that Oracle "is the bridge between junk-rated neoclouds and investment-grade hyperscalers", and that "A downgrade to junk would be the clearest sign of contagion." Goldman Sachs credit strategist Amanda Lynam wrote that neoclouds and firms with heavy frontier-lab exposure are at the centre of the funding-cost debate, while investment-grade bonds have seen milder price moves.

What sceptics say

Holders of the bonds argue the market is overreacting. John Lloyd of Janus Henderson, whose fund owns CoreWeave 2031 bonds, told CNBC he thinks the moves are overdone, while admitting the company "need[s] the debt markets in order to grow." Demand has not gone away either. On 17 September CoreWeave said it was still contracting new capacity at higher prices, around $40 million per megawatt of annualised revenue. It said it had signed more than $25 billion of net new commitments early in the third quarter, and its contracted power had grown to about 4.2 gigawatts. Its second-quarter backlog was about $104 billion.

A CDS spread is a market price, not a forecast. The implied default probabilities depend on recovery-rate assumptions, and thin CDS markets can overshoot in both directions.

My take, as opinion: the detail that worries me in those September terms is "short-dated", three to six months. Short contracts at high prices are good for revenue this quarter. They do nothing to match revenue to debt that runs to 2031 and 2032.

What it means for engineering and DevOps teams

• Treat your GPU provider as a counterparty. If your inference or training runs on a neocloud, check its debt maturities and bond yields the way finance checks a supplier's credit. A provider paying 15% to borrow can run into trouble and change its terms quickly.
• Make leaving cheap. Keep model weights, checkpoints and datasets somewhere you control. Keep your deployment manifests portable across providers. Choosing between GPU clouds is much easier when moving is a config change.
• Watch prices both ways. Stress at lenders can push some providers to discount spare capacity for cash, while others raise prices to cover interest. Short commitments and per-hour self-hosting break-even maths give you room to take advantage of either.

FAQ

What does a CoreWeave CDS above 1,000 basis points mean?

It costs more than 10% of the insured amount per year to protect CoreWeave's debt against default for five years. Market practice treats spreads above 1,000 as distressed. Reuters said 1,000 implied nearly a 63% default risk over five years. That figure depends on recovery-rate assumptions, so it is a signal, not a prediction.

Is CoreWeave going bankrupt?

Nothing in its filings or public statements suggests that is imminent. It reports growing revenue, a backlog of about $104 billion and new contracts at higher prices. The risk the market is pricing is refinancing: whether it can keep borrowing at bearable rates while interest expense climbs faster than operating income.

When does CoreWeave report Q3 2026 earnings?

After the US market close on Tuesday 10 November 2026, with a call at 5:00 p.m. Eastern, according to the company's 27 October announcement.

Related: GPU Clouds in 2026: H100s at $2.59 an Hour, $104B of CoreWeave Backlog on $35B of Debt, and a Six-Year Bet on Silicon, Self-Hosting LLMs vs API Costs in 2026: The 62% Utilisation Break-Even and FinOps for AI in 2026: 98% Manage Token Spend, 3 in 4 Can't Prove the Value, GPUs Run at 5%, and FOCUS Gets a Token Column in 1.5.