Analysis
Nvidia is working on a fresh wave of AI infrastructure deals worth more than $750 billion combined, according to Bloomberg, a figure that folds together its more-than-$500-billion Korean initiative with SK Group and the roughly $250 billion backstop it is separately discussing for OpenAI's Ohio data center lease. The scale reignited a debate that has simmered around Nvidia all year: whether financing the very companies that buy its chips constitutes a form of circular, self-reinforcing demand that overstates the AI industry's real growth.
The market reacted immediately. The cost of protecting Nvidia's five-year debt against default jumped by the most on record Monday, rising as much as 0.14 percentage points to around 0.82 percentage points annually, according to ICE Data Services -- a sharp repricing of credit risk for a company that, until recently, traded more like a cash machine than a leveraged infrastructure financier.
Critics have raised the circularity concern for months: companies Nvidia finances or takes equity stakes in -- xAI, CoreWeave, and now OpenAI and SK Group -- overwhelmingly spend that capital buying or leasing Nvidia's own chips and data center capacity, meaning Nvidia is, in effect, underwriting demand for its own products rather than simply selling into organic demand. Nvidia CEO Jensen Huang has pushed back forcefully, telling Bloomberg the same day that the semiconductor industry needs to grow roughly tenfold over the next decade to serve a computing world increasingly dominated by autonomous AI agents rather than human users, and separately dismissing talk of an imminent chip-industry downturn as premature.
The $750 billion figure dwarfs even the eye-popping AI infrastructure commitments that defined the first half of 2026, and it lands the same week Alphabet posted its first-ever negative free cash flow quarter on AI capex, Intel dropped 11% on a capex raise despite beating earnings, and Meta, Microsoft and Amazon head into earnings under intense investor scrutiny over their own AI spending. Nvidia's willingness to keep extending its balance sheet into every corner of that buildout is now the single biggest variable in how the market prices the entire AI trade.
The bear case is straightforward: if any major counterparty's revenue growth fails to keep pace with the capacity Nvidia has financed on their behalf, the company's credit exposure -- not just its chip backlog -- becomes the story, and Monday's CDS spike suggests bond investors are already pricing that risk more seriously than equity investors are.
What to watch: whether Nvidia's five-year CDS spread keeps widening into next week, whether Huang's tenfold-growth framing holds up against skeptics ahead of Meta, Microsoft and Amazon's earnings, and whether any single deal in the $750 billion package gets renegotiated or scaled back as scrutiny intensifies.
