Bond Traders Worry Over $70 Billion in Shadow Credit for AI Firms

Sun Aug 16 2026
Ray Pierce (950 articles)
Bond Traders Worry Over $70 Billion in Shadow Credit for AI Firms

Even prior to Nvidia Corp.’s notable $500 billion financing partnership this week, investors were beginning to express concerns regarding the approximately $70 billion in latent liabilities that are absent from major AI companies’ balance sheets, yet could emerge at the most inopportune moment. With its latest move, Nvidia is positioned to offer potentially tens of billions of dollars in “residual value” support for debt deals associated with the artificial intelligence build-out — effectively enabling firms to leverage its robust credit rating in an effort to manage customer costs. With demand for computing power soaring, this backstop has become something akin to a free lunch for massive companies, as they can bolster sales to clients and AI firms without taking debt onto their own books. Meta Platforms Inc., which initially implemented this structure for its data centers, clearly stated its residual value guarantees in filings: “RVG payments are not probable, and therefore no liability has been recorded to date.” And “Not probable,” though, is increasingly not enough for some.  These backstops are anticipated to grow in response to the impending surge of AI chip debt. That has prompted investors to examine the limited number of previous transactions for insights into how Nvidia might arrange its contracts.

Under the structure of Meta’s agreements, the backstop would only be activated if Meta decides to exit its data centers prematurely. Broadcom has implemented the concept in the realm of chip financing, providing support to Anthropic PBC and thereby linking its own exposure to the actions of a customer. The arrangements typically involve a multi-step chain: a special-purpose vehicle borrows money to acquire the chips, secured by the cash flow from a contract with a company that will utilise the technology. If the firm ceases payments, the assets will either be re-leased or liquidated to settle the outstanding debt. If there remains a shortfall, the backstopper compensates for the discrepancy. In principle, that sequence represents a distant risk. Proponents contend that the demand for chips is poised to exceed supply for an extended period. The debt is designed to be amortised completely over time, indicating that the potential expense associated with any residual value support diminishes as the years progress. Arguably, the tech risk is appropriately situated on the balance sheets of companies possessing sufficient cash reserves to withstand potential setbacks should adverse events occur. Yet rating agencies are not dismissing the possibility of those backstops being utilised, and investors are advocating for caution regarding the risks that are building off balance sheet. Wall Street exhibited scepticism regarding the rapid acceleration of AI infrastructure expenditures and their potential to generate adequate returns.

In a period of industry contraction, these safeguards may compel semiconductor manufacturers to fulfilll commitments amounting to billions at a time when their earnings are facing significant strain. “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,” said Mariya Entina. “We’re coming into this era of financial engineering. And that’s one of my concerns: When you have financial engineering, you’re obscuring the financial reality.” Representatives for Nvidia and Broadcom did not provide a response to enquiries for comment. A representative for Meta declined to provide a comment. The question currently occupying investors is how to assess the likelihood that these off-balance-sheet contingencies will materialise into on-balance-sheet issues. With the residual value support, Nvidia is effectively “writing a put,” analysts noted in a report. “This is pro-cyclical and exacerbates boom-bust potential,” the analysts wrote. “The guarantee is nearly costless in the boom phase, but becomes most relevant in a severe, abrupt downturn, if / when customers are defaulting and market value for hardware is falling.” Nvidia Chief Executive Officer Jensen Huang stated in a post on X that his company might offer a residual-value support mechanism for as much as 25% of an opportunity, evaluated on a case-by-case basis. “Our role is to help unlock a very large pool of independent capital while maintaining disciplined risk exposure,” he wrote.

Nvidia provided limited information regarding the partnership, stating only that it aimed to address concerns related to “circular financing,” in which AI companies finance each other to acquire one another’s products, by introducing external capital. Meta established a framework for its AI expansion by utilising the residual value backstop structure within a debt package approximating $27 billion to fund the development of its Hyperion data center located in rural Louisiana, in a transaction referred to as Beignet. This essentially safeguards lenders in the event that Meta opts to terminate its 20-year lease prematurely. The company executed a comparable structure in a debt arrangement approximating $13 billion for its Sopaipilla data center initiative in Texas. Broadcom generated significant attention by implementing a comparable framework for chip financing thru project Big Sky, in which it underpinned the majority of a $35 billion debt arrangement. This deal involved investors such as Apollo Global Management Inc. and Blackstone Inc. financing the acquisition of specialised AI chips intended for leasing to Anthropic. This allowed the senior debt tranches to secure investment-grade ratings while benefiting from reduced borrowing costs. In contrast to the long-term agreements typically associated with data centers, financing arrangements for chips tend to function within more immediate timeframes. They generally amortise over a period of approximately five years to align with the swift depreciation of the technology. This shorter horizon rapidly diminishes the backstop value, instilling in lenders a sense of confidence derived from a clear, immediate outlook rather than a protracted, multi-decade wager. “This isn’t your garden-variety investment grade credit underwrite,” said Brian Gelfand. “It’s far more complex than that. The tail risks are elevated given the off-balance sheet nature.”

The Big Sky deal represented the initial phase of a partnership known as AI XPV, which Broadcom established in June. This platform has the potential to accumulate $370 billion in senior debt by mid-2029, as indicated by strategists. “The primary risk lies in multiple such transactions occurring over a short period,” analysts at Moody’s Ratings wrote in a note. “We would view a substantial increase in Broadcom’s contingent obligations, even if leverage associated with its outstanding debt remains low, as limiting Broadcom’s financial flexibility and could create an overhang on the company’s credit profile.” While the medium-term outlook for AI infrastructure remains robust, they noted, the risk of the backstop being activated still persists due to constrained long-term visibility. S&P Global Ratings stated that it views the residual value support provided by Broadcom as a “contingent debt-like obligation” that will be included in its adjusted debt calculation. Under US accounting rules, companies generally recognise contingent liabilities on their balance sheets when the likelihood of losses is high and can be estimated with reasonable accuracy. In certain instances, potential losses may be disclosed in the footnotes of financial statements. Some see a doomsday scenario as an overblown fear. For the residual value support to be triggered, “you would have to have growth rates of token usage fall of a cliff, which we’re just not seeing,” said John Lloyd. “They’re not trying to hide the contingent liability. They’re trying to get it financed.”

Ray Pierce

Ray Pierce

Ray Pierce is a Senior Market Analyst. He has been covering Asian stock markets for many years.

We use cookies to improve your experience.
Privacy Policy