Nvidia’s $500 billion AI financing plan gets a Wall Street reality check
Banks and investors are questioning how long Nvidia’s chips can generate revenue and demanding stronger guarantees for chip-backed loans.
Nvidia’s mammoth financing plan has opened a debate on Wall Street over how much its advanced chips and the infrastructure built around them are really worth. The answer from banks and investors appears to be: not as much as Nvidia thinks.
Some lenders want stronger guarantees than the company initially outlined, even for Nvidia’s industry-leading AI processing power, as they try to determine how long the revenue-generating life of its chips will last, banking sources and credit managers told Reuters. Several requested anonymity to discuss the financing structures.
The chips, which provide the critical processing power known as “compute,” are at the center of a potential mismatch between Nvidia and a more cautious Wall Street. That could create financing challenges for AI companies seeking to tap the deep pools of capital needed to fund the industry’s rapid expansion.
Bankers and asset managers said they have doubts about whether Nvidia’s chips can serve as long-term collateral on their own. As a result, they want greater protections around Nvidia’s $500 billion financing plan, which relies on chip-backed loans.
Deals currently in the pipeline are likely to offer investors greater certainty, including stronger guarantees, the sources said.
“Wall Street is much more conservative,” said Tony Trzcinka, a senior portfolio manager at Impax Asset Management, referring to Nvidia’s claim that its most specialized chips can generate revenue for a decade.
An Nvidia spokesperson said its “AI compute is a productive, durable and fungible asset that can support long-term financing. Our financing partners independently assess each opportunity, including customer commitments, expected cash flow and residual value.”
“Financing structures will vary as this market develops,” the spokesperson added.
Questions over the plan
The questions surrounding Nvidia’s financing strategy come as companies and investors pour hundreds of billions of dollars into data centers, chips and power capacity to support the broader AI boom, which is helping drive U.S. economic growth.
That spending has increased scrutiny of increasingly complex financing structures and the assumptions underpinning the long-term value of computing hardware.
Some analysts have questioned the use of private credit, vendor financing and circular financing arrangements. The latter two also played a role during the dot-com boom and bust more than 25 years ago, Morningstar analysts said in a recent note.
Nvidia’s plan, announced in August with financiers including Blackstone, Apollo and KKR, envisioned using its chips as collateral with limited guarantees, allowing AI developers to gain access to Nvidia’s compute. The strategy was seen as an expensive but potentially necessary way to create a financing market similar to the one used to finance aircraft.
Nvidia has said some deals could carry a residual-value guarantee of no more than 25%, arguing that the initiative is designed to address concerns over circular financing.
The relatively limited level of support compared with some other recent AI financing deals has led some on Wall Street to expect Nvidia to offer additional protections.
Three banking sources familiar with the matter but not part of the original financing group said Nvidia may ultimately need to provide guarantees on all of its deals, or have them backed by revenue streams from investment-grade customers such as major technology companies to cover debt payments.
For now, the sources said, the market is not yet ready to treat AI compute as an investment-grade asset comparable to aircraft.
More certainty for lenders
Nvidia is seeking to address some of these concerns, the sources said, providing details that had not previously been reported.
Tens of billions of dollars in loan deals currently in the pipeline are likely to include stronger guarantees and contractual protections, one source said. A second source said some structures under consideration could provide lenders with guarantees.
The initial deals will give AI developers access to Nvidia’s compute. They will be secured by Nvidia chips and backed by customer contracts and Nvidia’s underlying guarantee, the first source said.
Despite concerns over the guarantees, there remains strong demand to finance the deals, the sources said.
The Nvidia spokesperson did not directly address Reuters’ questions about potentially expanded guarantees.
Chief Executive Jensen Huang has said he wants to make Nvidia’s compute “an investable infrastructure asset,” rather than rely on the more traditional model in which customers purchase subscription contracts from AI companies to access compute.
In an August blog post, Huang said the $500 billion initiative was designed to address concerns over circular financing, in which a company helps finance purchases of its own products. The structure is intended to bring independent institutional capital into the AI infrastructure market, while Nvidia’s financial partners raise money from investors to support the loans, he said.
A third banking source said lenders and private capital are expected to play a role in the financing.
Five of Nvidia’s six financial partners declined to comment. Apollo did not respond to a request for comment.
How long will the chips last?
Huang has addressed some of the skepticism by arguing that Nvidia’s graphics processing units, or GPUs, have a useful life of up to a decade. The specialized chips provide the computing power needed to train and run AI models.
Some credit investors and bankers, however, believe the period during which the chips can reliably generate revenue is considerably shorter.
“Nvidia would imply that the GPUs work well north of five years, and that actually has been proven to be true thus far,” said Andrew Chang, a director at S&P Global Ratings. “Yet we take a conservative view of the value of those chips,” he said.
The first banking source said part of the divergence between Nvidia and lenders stems from the lack of historical data that would allow banks to confidently underwrite the long-term residual value of GPUs.
“Banks typically underwrite GPUs over a 3-4 year depreciation schedule,” Trzcinka of Impax said.
“That is different than Nvidia, which argues top-tier GPUs can earn revenue for a decade,” he said, adding that investors are likely to demand higher interest rates, larger financial cushions and stronger repayment protections before backing loans secured by AI chips.
“As investors, you’re going to be a lot pickier about the levels that you need to get compensated for to take incremental risk,” said Loren Moran, a fixed-income portfolio manager at Wellington Management, which has $1.3 trillion in assets under management.
Nvidia has pointed to several third-party studies showing that major cloud companies are extending the depreciation periods for servers to five or six years, from three or four.
The company has also cited a finding from Barkr, which provides valuations for AI collateral such as GPUs, that its latest GB300 NVL72 systems could have a useful life of nine to 10 years.
Chip-backed loans test the market
Several companies have already tested the market with chip-backed loans. Unlike Nvidia’s proposed structures, however, those transactions have been supported by revenue streams from technology companies that can be used to service the debt.
CoreWeave, in which Nvidia has a stake, closed an $8.5 billion facility earlier this year that was described as the first investment-grade GPU-backed loan. The facility received an A3 rating in large part because lenders rely on contractual payments from Meta, which are viewed as highly dependable.
Separately, Broadcom, an Nvidia rival helping finance AI computing capacity for Anthropic, backstopped more than 80% of a $35 billion financing structure, helping attract debt investors.
Nvidia itself previously provided a residual-value guarantee to support financing for SB Energy’s Ohio data center project, according to S&P and Moody’s.
“The precedent transactions so far would suggest that the creditor community does not subscribe to long average lives for these assets,” said Brian Gelfand, co-head of global credit at TCW, which manages more than $200 billion.
