CoreWeave Credit Risk, October 2026: CDS Markets Price 60% Default Probability

Bloomberg and CNBC reported this month that CoreWeave credit-default swaps hit distress levels with pricing implying 60% default odds, even as the company holds a reported $104 billion order backlog and completed a $4.2 billion convertible-note offering on September 22, 2026.

CoreWeave sits at the center of an unusually sharp disagreement between equity and credit markets. Bloomberg reported this month that the AI GPU-cloud operator's credit-default swaps have hit record levels, with pricing implying a 60% probability of default — a threshold CNBC separately characterized as distress territory. That signal arrives not in response to any disclosed operational failure, but alongside a reported order backlog of approximately $104 billion against liabilities of roughly $72 billion and third-quarter contract pricing for new capacity running at approximately $40 million per megawatt. The divergence is a useful lens on the central question in AI infrastructure finance: whether debt-funded buildouts at this scale can service their obligations before demand assumptions face a real test.

The credit concern has a clear structural source. Slicast's compilation of SEC XBRL data shows CoreWeave's capital expenditure reached $10.31 billion in FY2025 — up 18.5% from $8.70 billion in FY2024 — against revenue of $5.13 billion, a capex intensity of 201% and the highest absolute spending of any neocloud in the peer group. In the single quarter ending March 31, 2026, capex reached $7.70 billion. To bridge that financing gap, CoreWeave filed an 8-K on September 22, 2026, disclosing the completion of a $4.2 billion offering of 2.875% convertible senior notes due 2033, placed with qualified institutional buyers under Rule 144A in an upsized transaction. Reuters this month analyzed the AI contracts and lenders underpinning the broader buildout; Stock Titan separately mapped the debt maturity schedule. Rothschild Redburn downgraded CoreWeave to Sell in September 2026 with a $54 price target, placing the borrowing model at the center of its bear thesis: it warned that demand driven by venture-capital-funded AI companies may prove unsustainable, flagged what it described as a $3 trillion off-balance-sheet risk across the sector, and argued that, net of costs, CoreWeave is generating losses despite rising headline contract prices. Analysis published this month also highlighted so-called cash-trap provisions in the company's borrowing agreements as a structural constraint in adverse scenarios.

Against that backdrop, CoreWeave's strategic pipeline has continued to expand. The company reported net new customer commitments exceeding $25 billion in the third quarter, bringing total signed capacity to 4.2 gigawatts. OpenAI, a cornerstone customer, reportedly signed a $12 billion compute agreement. In September 2026, CoreWeave offered NVIDIA's Vera Rubin NVL72 GPU clusters at production scale, with Cognition as an early customer. The company also announced a 240-megawatt GPU cloud region in Navi Mumbai through a partnership with AdaniConneX, establishing its first presence in India. The launch of Forge, a full-stack AI cloud platform, represents a further step toward competing for enterprise workloads that have historically gone to hyperscalers.

Recent market action has been turbulent across the neocloud cohort. On October 8, AI infrastructure stocks including CoreWeave, NVIDIA, and Oracle fell broadly after reports of a downward revision to OpenAI's annualized revenue to approximately $50 billion; CoreWeave's shares dropped roughly 7.77% to approximately $81.58 on that session, according to reporting that also underscored why the company's borrowing model amplifies the stakes of any demand shortfall. Also reported this month, Firmus withdrew a planned $5 billion IPO, adding to sector-wide uncertainty about the durability of leveraged GPU infrastructure. Competitor IREN, by contrast, reportedly secured a multi-year compute agreement with Anthropic, illustrating that individual customer wins remain a meaningful differentiator within the cohort. Lambda Labs, backed by NVIDIA, is separately seeking $4 billion in funding ahead of a targeted 2027 IPO, signaling growing competitive supply. On October 14, CoreWeave closed at $69.7, up 6% on the session; no published report identifies the specific catalyst for that move.

The analyst community has not converged. Cantor Fitzgerald maintained a Buy rating with a $176 price target this month; JPMorgan upgraded the stock, flagging rising compute pricing as a positive re-rating factor — a view informed by Nebius raising its H100 hourly rates by 17% to 21% from October 1, seen as a potential positive read-across for the sector. Rothschild's $54 target implies further downside from current levels, and a 60% CDS-implied default probability is a rare condition for a company that is simultaneously deploying the newest GPU generation at production scale and expanding into India. The core analytical question is not whether AI infrastructure demand is real — the backlog and reported contract sizes suggest it is — but whether the current debt structure can be serviced if customer renewal rates slow, GPU rental prices compress under competitive supply from new entrants, or hyperscalers accelerate their own internal buildouts. Three signals are worth tracking: how much of the reported $104 billion backlog converts into contracted, cash-generating revenue as individual deals approach renewal; whether GPU spot pricing holds at the October levels signaled by Nebius or softens as Lambda and other entrants add supply; and whether any material change in CoreWeave's debt covenants or refinancing terms indicates how lenders are privately reading the same credit risk that the CDS market is currently pricing at 60%.