Nvidia's $500 billion AI financing push draws bond market scrutiny

Bond investors are questioning whether Nvidia's $500 billion AI financing partnership is building a $70 billion pile of off-balance-sheet liabilities that could surface exactly when the chipmaker can least afford them.

"It's like you're really gaming the system here; you're trying to get preferential treatment from rating agencies so that you get the best rating possible," Mariya Entina, a portfolio manager at DoubleLine, said. "We're coming into this era of financial engineering."

The structures, known as residual value guarantees, let chipmakers backstop debt tied to AI infrastructure without recording the exposure on their own balance sheets. Nvidia said it may provide residual-value support of up to 25 percent for some projects, while Broadcom already backstopped most of a $35 billion debt deal where Apollo Global Management and Blackstone financed custom AI chips leased to Anthropic.

The stakes are large. Goldman Sachs estimates AI-related financing now accounts for nearly one-quarter of all gross US investment-grade issuance, while AI investment approaches $600 billion this year. If rating agencies fold these contingent obligations into adjusted debt calculations, borrowing costs could rise for Nvidia, Broadcom, and the broader AI buildout.

How residual value guarantees work

The typical structure runs through a special-purpose vehicle that borrows money to buy chips, backed by cash flows from a contract with a company that will use the technology. If that firm stops paying, the assets are leased out again or sold to repay the remaining debt. If a shortfall remains, the chipmaker makes up the difference.

Meta created the template when it used the backstop as part of a roughly $27 billion debt package to finance its Hyperion data center in rural Louisiana, in a deal called Beignet, followed by a roughly $13 billion deal for its Sopaipilla project in Texas. Meta said in filings that "RVG payments are not probable, and therefore no liability has been recorded to date."

Broadcom extended the concept to chip financing with project Big Sky, where the residual value support helped senior debt tranches win investment-grade ratings at lower borrowing costs. The arrangement is part of a partnership called AI XPV that could amass $370 billion of senior debt by mid-2029, according to Bank of America strategists.

Unlike data-center deals spanning decades, chip financings amortize in roughly five years to match the rapid depreciation of the technology. That shorter horizon burns down the backstop value quickly, giving lenders a visible exit rather than a multi-decade bet.

Rating agencies are watching

Moody's warned that the primary risk lies in multiple such transactions occurring over a short period, saying a substantial increase in Broadcom's contingent obligations "could create an overhang on the company's credit profile." S&P Global Ratings said it considers the residual value support a "contingent debt-like obligation" that it will add to adjusted debt calculations.

CreditSights analysts compared Nvidia's residual value support to "writing a put," calling it pro-cyclical and warning it "exacerbates boom-bust potential." The guarantee is nearly costless in the boom phase but becomes most relevant in a severe downturn when customers are defaulting and hardware values are falling.

Not everyone sees a doomsday scenario. "You would have to have growth rates of token usage fall of a cliff, which we're just not seeing," said John Lloyd, global head of multi-sector and corporate credit at Janus Henderson Investors. "They're not trying to hide the contingent liability. They're trying to get it financed."

The financing push also marks a shift in who pays for the AI boom. A year ago, most of Big Tech financed AI from cash flows; now debt is taking over. Goldman Sachs estimates hyperscalers hold roughly $1.5 trillion in combined lease commitments, up from about $200 billion five years ago, with about $1 trillion classified as uncommenced. PIMCO's Lotfi Karoui said the AI capital expenditure cycle is on track to become the largest investment cycle since the nineteenth-century railway boom, with hyperscaler spending expected to exceed $1 trillion annually from 2027.

For investors, the question is whether the market has priced in the risk. Nvidia shares trade at a premium to the broader market, and its credit default swaps narrowed after the financing announcement but remain elevated relative to its rating. If the backstops are ever triggered, they would hit chipmakers' earnings precisely when AI demand is weakest — the same moment their own revenue is under pressure.

This article is for informational purposes only and does not constitute investment advice.