Every investment committee meeting this week opens with some version of the same question: is Nvidia’s financing of the AI buildout ordinary vendor support, or is it evidence that part of the demand it keeps selling into is propped up by Nvidia itself?
The numbers make the question unavoidable. As of its July quarterly filing, Nvidia disclosed $36 billion of AI cloud agreement commitments with certain AI cloud partners. Separately, it disclosed $20 billion of uncommenced data center leases it expects to reassign to third parties. Nvidia has also disclosed residual value guarantees, capped at $105 billion, tied to SB Energy’s PORTS-Pike Technology Campus in Ohio and 20-year leases to OpenAI, with obligations becoming effective in phases as conditions are met. That is not a rounding error on a chipmaker’s balance sheet.
Why the Circularity Argument Has Teeth Now
For most of this year, bears raised the circularity concern in the abstract: Nvidia finances customers who buy Nvidia chips, which inflates Nvidia’s own revenue. Bulls dismissed it as theoretical. It stopped being theoretical when CoreWeave’s credit began to deteriorate publicly and fast.
In mid-October, reporting citing S&P Global Market Intelligence data showed the spread on CoreWeave’s five-year credit default swaps trading above roughly 1,100 basis points, with prints reported above about 1,140 basis points. Market data commentary around the move noted that spreads above 1,000 basis points are commonly associated with credit distress. CoreWeave is also carrying heavy leverage, with widely circulated estimates putting total debt around $51.6 billion, and its longer-dated notes have recently been yielding in the mid-teens.
Shares of CoreWeave, Nebius, and IREN, all cash-burning companies in Nvidia’s orbit, sold off sharply during October’s risk-off stretch. Nvidia-backed Firmus abruptly withdrew its IPO this month. The public markets for Nvidia’s ecosystem are closing, not opening.
The Bull Case Is Not Stupid
For the vast majority of these deals, Nvidia will not have to pay unless project participants default. That is a meaningful distinction. Contingent liabilities are not the same as debt. Some analysts have argued CoreWeave’s marginal cost of capital may be less sensitive than its unsecured bond yields suggest if incremental funding is structured as asset-backed, project-level financing rather than traditional corporate unsecured issuance.
Other bulls argue Nvidia’s customers, including hyperscalers and neoclouds, are seeing solid unit economics on deployed capacity. Jensen Huang’s public position is that capital availability, not demand, is the bottleneck, and he has been explicit that working capital is what limits how quickly the ecosystem can scale.
What Investors Are Missing
The debate tends to frame this as binary: either the commitments never get called, or they do and Nvidia bleeds. The more important question is subtler. Around mid-October, market commentary and credit research notes described a sharp widening in AI-related high-yield spreads, with AI-linked high-yield spreads cited around 360 basis points over Treasuries after a roughly 25 basis point widening over a couple of days, while comparable non-AI spreads moved far less. The credit market is drawing a harder line around AI-specific borrowers.
That matters for Nvidia because its August financing initiative with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR is designed to mobilize third-party capital for exactly this ecosystem. Private capital is becoming critical because the next wave of AI demand includes labs, neoclouds, and infrastructure developers that lack the financing capacity of Microsoft, Alphabet, or Amazon. If institutional appetite for that paper dries up before those platforms are established, Nvidia’s role as lender of last resort becomes less contingent and more literal.
Nvidia wrapped up a third consecutive down week on Friday, and the stock has been down roughly mid-single digits month to date in October. The market is not yet pricing a credit event. It is pricing uncertainty about whether one is possible. That is a different kind of risk, and it will not resolve until CoreWeave’s bonds either stabilize or keep pushing toward the high-70s to low-80s cents-on-the-dollar zone that markets often associate with stressed credit.
Stocks to Watch
Nvidia (NVDA). The central name. Its commitments are contingent, but a single 4.4% decline on October 22 erased about $235 billion in market value. Sentiment moves faster than fundamentals here.
CoreWeave (CRWV). The bellwether. Its market cap is around $30 billion as of late October 2026. CDS spreads and bond yields are the data points that move the whole sector.
Nebius (NBIS) and IREN. Nebius and other AI infrastructure-adjacent names have traded in sympathy with CoreWeave, suggesting the market increasingly treats the neocloud group as a single credit story rather than individual bets. Any resolution in CoreWeave’s credit should lift both; any deterioration amplifies the contagion.
OpenAI (private). Nvidia’s Ohio residual value guarantee is capped at $105 billion and becomes effective in phases as project milestones are met. OpenAI’s ability to sustain financing access on favorable terms, with or without additional collateral support, may now be among the most consequential credit-adjacent variables in technology this year.
