CoreWeave’s default risk surged to a record in credit markets on Thursday, prompting Nvidia CEO Jensen Huang to publicly defend the company and say Nvidia does not need to provide financial support.
On Oct. 15 in U.S. Eastern Time, CoreWeave’s five-year credit default swap spread, a measure tied to the cost of insuring against a debt default, rose as much as 74.5 basis points to 1,194.5 basis points intraday. That marked an all-time high. At the same time, yields on some of the company’s bonds climbed into a range above 15% to more than 16%, a sign that investor concern over its ability to service debt had intensified.
Huang, speaking to reporters while attending an event in Sweden the same day, said, “CoreWeave gets stronger every day.” He said he remained optimistic about the company and added that its “cost of capital is actually declining.” Asked whether Nvidia might need to backstop CoreWeave financially, Huang said he did not think that was necessary.
CoreWeave shares, trading under ticker CRWV, opened down 6.6% on Thursday and at one point fell as much as 9.5% in early trading. After Huang’s comments, the stock cut losses quickly, turned positive by midday, and kept rising into the close. Late in the session, the stock was up a little more than 6% on the day, rebounding roughly 17% from its intraday low. By the close, it had gained more than 5.9%, snapping a three-day losing streak and narrowing its loss for the week to about 15%.
CDS near 1,200 basis points as CoreWeave bonds sell off
CoreWeave is facing pressure in both stocks and bonds.
Data from ICE Data Services showed the company’s five-year CDS spread climbed to 1,194.5 basis points intraday on Thursday, the highest level on record. Pricing at that level implied a cumulative default probability of more than 60% over the next five years. Separate figures from LSEG and S&P Global Market Intelligence showed CoreWeave’s CDS spread moved above 1,100 basis points during the session, implying roughly a 64% five-year default risk, up from about 30% a year earlier.
CDS contracts are credit derivatives used to hedge against default risk. Their spreads usually rise when the market grows more concerned about an issuer’s ability to repay debt. Still, CDS-implied default probabilities depend on pricing models and assumptions, and they do not amount to a precise forecast of whether a company will default in the next five years.
Bond trading painted the same picture. Bloomberg, citing market data, reported that a CoreWeave bond due in 2031 with a 9% coupon fell about 3.3 cents to 79.9 cents on the dollar, pushing its yield to 15.6%. Another 2031 bond carrying a 9.75% coupon fell about 2.6 cents to 78.65 cents on the dollar, sending its yield to 16.2%.
Reuters reported that a CoreWeave bond due in 2032 traded at about 78 cents on the dollar on the Tradeweb platform, with a yield above 16%. The bonds differ in maturity, coupon, and price, but all pointed to the same conclusion: investors are demanding much higher compensation for risk.
When bond prices fall far below par and yields jump this sharply, future borrowing or refinancing through the bond market can become more expensive.
CoreWeave co-founder Brannin McBee said in a recent interview with the Financial Times that the company wants an investment-grade credit rating, which would open broader funding channels and potentially let it access what he described as an “unlimited” pool of capital. That ambition stands in sharp contrast with current bond yields. With concerns over debt repayment still unresolved, the market has yet to decide whether the company can keep raising funds at lower cost.
Oracle, SoftBank and others also see higher AI-linked credit stress
CoreWeave is not the only AI-related company under pressure in credit markets.
During Thursday’s session, Oracle’s five-year CDS spread was quoted at 288 basis points, unchanged from the prior trading day but up 28 basis points from 260 basis points a week earlier. SoftBank’s five-year CDS spread widened 24 basis points in a single day to 424 basis points. Nvidia, STMicroelectronics, and SpaceX also saw their CDS spreads move noticeably higher.
According to LSEG and S&P Global Market Intelligence, CDS spreads for Amazon, Alphabet, and Meta also reached record highs, though the pace of widening was slower than at Oracle and CoreWeave. The cost of insuring debt at those large technology companies has also been affected by a broader reassessment of financing risk tied to AI investment.
CDS spreads measure what investors must pay to protect against a default. Wider spreads mean the market is asking for more risk compensation. They do not mean a company is about to default, and they should not be treated as directly comparable default probabilities across different issuers.
Rising credit insurance costs across several AI-linked companies show investors are rethinking how AI infrastructure expansion is being financed. Oracle’s large borrowing needs have come under scrutiny, while CoreWeave has become the clearest stress case because of its heavy debt load, continued losses, and dependence on external financing.
Reuters said stronger-than-expected U.S. inflation data and growing concern about the path of future interest rates have added to doubts over the payback period for massive AI capital spending. As more of the sector relies on debt rather than free cash flow to finance AI buildouts, the market’s focus is shifting. Demand for compute is still central, but the bigger question now is whether companies can generate enough profit and cash flow quickly enough to cover rising interest expense.
Nearly $50 billion in debt puts CoreWeave’s expansion model under the microscope
The core issue for CoreWeave is the balance between capital spending growth, cash flow, and access to financing.
Media reports said the AI cloud company is carrying close to $50 billion in debt and remains unprofitable. Its model depends on buying Nvidia GPUs at scale, building data centers, and expanding compute capacity to meet customer demand. That kind of investment usually requires large upfront spending well before revenue is fully realized.
CoreWeave has ramped up capital expenditures aggressively to capture a fast-growing AI compute market. In periods of strong demand, debt-funded expansion can help a company build infrastructure ahead of time and support investment plans with long-term customer contracts. But data centers take time to build and require heavy equipment spending. If financing costs rise, projects are delayed, or customer demand grows more slowly than expected, funding gaps and debt-servicing pressure can emerge quickly.
The market is not only questioning the size of CoreWeave’s debt. It is questioning whether future cash flow will be enough to cover capital expenditure, interest payments, and other operating costs. Rising bond yields make that problem harder by lifting the hurdle for new financing.
CoreWeave management has said the company has a strong backlog and that the market is misreading its financial position. But a backlog is not the same as cash already collected. Long-term contracts may provide some visibility on revenue, yet they do not remove risks tied to construction cost overruns, delivery delays, changing financing conditions, or customer concentration. Investors still have to judge whether those contracts can turn into enough cash flow and durable profit.
Why Huang stepped in as Nvidia’s ties to CoreWeave draw scrutiny
Huang’s public defense of CoreWeave reflects the close commercial and capital relationship between Nvidia and the AI cloud provider.
Nvidia owns about 10% of CoreWeave, making it the company’s second-largest shareholder. CoreWeave is also a major buyer of Nvidia GPUs. Its business model depends on purchasing Nvidia chips, building data centers, and then leasing compute power to AI developers and large technology companies.
If CoreWeave runs into financing trouble, future data center construction, GPU purchases, and compute delivery could all be affected. Existing orders may not be canceled right away, but higher funding costs could still alter the pace of project development and the expected return on capital. That, in turn, could influence how the market views future AI demand for Nvidia.
A more sensitive issue is the concern over what some investors see as circular financing. A supplier owns a stake in a customer, while that customer borrows to buy the supplier’s products.
That arrangement does not by itself imply wrongdoing. Even so, when a company depends on debt to buy a supplier’s products and the supplier also holds equity in that company, investors naturally ask how much of the growth in the value chain is coming from end-customer demand and how much is being sustained by financing-driven expansion.
By saying CoreWeave’s cost of capital is falling and that Nvidia does not need to step in, Huang was trying to support confidence in an important partner while also pushing back against any suggestion that Nvidia is responsible for CoreWeave’s debt risk.
His comments do not replace a review of CoreWeave’s balance sheet. Nvidia’s equity stake is not a debt guarantee, and Huang’s optimism does not directly solve the company’s financing needs, interest burden, or cash flow challenge.
AI infrastructure returns are becoming the market’s main test
CoreWeave’s credit turmoil is turning into a stress test for AI infrastructure investing.
As generative AI demand has grown rapidly, technology companies and compute operators have raced to build data centers in an effort to secure future market share. Revenue, however, often arrives later than the spending. That is especially true for newer cloud providers that depend on outside financing to buy GPUs, build server capacity, and pay energy and operating costs.
When rate expectations rise and bond investors demand more compensation for risk, that model comes under harder scrutiny. Higher financing costs can squeeze margins and force companies to slow expansion. Slower expansion can then affect future revenue growth and the return on investments already made.
Deutsche Bank strategist Jim Reid said this is not the first time markets have seen significant volatility tied to concerns about AI, but as the AI capital spending cycle is funded increasingly by debt rather than free cash flow, investors need to pay closer attention to signals from credit markets. “Overall, financing costs related to AI are not especially high yet, but there are worrying signs in some of the more marginal areas,” he said.
That view captures the industry’s central tension. Rising demand for compute does not automatically mean compute operators will earn enough profit, and growing order books do not automatically mean debt will be easy to repay.
For CoreWeave, the real test is whether it can control debt growth, keep access to funding open, and convert a large backlog into sustainable operating cash flow. For Nvidia and the wider AI supply chain, the market’s attention is shifting from how much compute will be needed to who will pay for the buildout, and whether those investments will produce returns large enough to justify the cost.
Huang tried to calm the market, but CoreWeave’s elevated bond yields, record credit insurance costs, and wider CDS spreads at Oracle, SoftBank, and other companies show that investors are digging deeper into the financing model behind AI expansion. Valuations across the AI supply chain may still depend on demand growth, but they are increasingly tied to whether companies can turn a capital spending race into real and sustainable cash flow at a financing cost they can bear.