NEW YORK: Nvidia’s ambitious plan to use its artificial intelligence chips to support a massive financing programme is facing growing scrutiny from Wall Street, with lenders questioning the long-term value of the hardware and seeking stronger safeguards.
Bankers and credit managers familiar with the discussions told Reuters that some lenders want greater guarantees than Nvidia initially proposed as they assess how long the company’s advanced chips can continue generating revenue.
The issue centres on the value of Nvidia’s GPUs, which provide the computing power required to train and operate AI models. While Nvidia views the hardware as durable and capable of supporting long-term financing, some banks and asset managers remain cautious about treating the chips as reliable collateral over extended periods.
The concerns could have wider implications for AI companies looking to raise billions of dollars to fund data centres, computing infrastructure and other investments tied to the rapid expansion of artificial intelligence.
“Wall Street is much more conservative,” said Tony Trzcinka, a senior portfolio manager at Impax Asset Management, referring to Nvidia’s expectation that its specialised chips could generate revenue for as long as a decade.
An Nvidia spokesperson defended the company’s approach, saying AI computing is a “productive, durable and fungible asset” that can support long-term financing.
The company added that financing partners assess each transaction independently, taking into account customer commitments, expected cash flows and the remaining value of the equipment.
LENDERS SEEK STRONGER PROTECTION
Nvidia announced its financing initiative in August alongside firms including Blackstone, Apollo and KKR. The proposed structure would allow AI developers to use Nvidia equipment as collateral, potentially opening access to additional funding for the costly computing infrastructure needed to expand AI operations.
Nvidia has indicated that some transactions could involve residual-value guarantees of no more than 25%, a structure intended in part to address concerns surrounding circular financing arrangements.
However, three banking sources familiar with the situation said the company could ultimately face pressure to provide guarantees covering a larger portion of its transactions.
They said lenders could also seek deals supported by predictable revenue from investment-grade customers, such as major technology companies, to provide greater certainty over debt repayment.
For now, the sources said, the market remains reluctant to assign Nvidia’s computing equipment the same financing status as established assets such as aircraft.
DEBATE OVER GPU LIFESPAN
The expected operating life of Nvidia’s GPUs has emerged as a key point of disagreement.
Nvidia CEO Jensen Huang has said the company’s GPUs can remain economically useful for up to 10 years. Some credit investors, however, believe the period during which the chips can reliably generate revenue could be considerably shorter.
Andrew Chang, a director at S&P Global Ratings, said Nvidia’s GPUs have so far demonstrated the ability to remain useful for more than five years, but added that the rating agency takes a conservative approach when valuing the equipment.
Trzcinka said banks commonly use a three- to four-year depreciation period when assessing GPUs, compared with Nvidia’s view that premium chips could continue producing revenue for a decade.
Investors could therefore demand higher borrowing costs, larger financial cushions and stronger protections before accepting AI chips as loan collateral, according to market participants.
Nvidia has cited studies showing that several major cloud companies have extended the depreciation period for servers to five or six years, compared with the previous three- to four-year range.
The company has also pointed to valuations from Barkr, which estimates that Nvidia’s latest GB300 NVL72 systems could have useful lives of between nine and 10 years.
PREVIOUS GPU-BACKED DEALS
Recent transactions involving AI computing equipment have generally included additional safeguards for lenders.
CoreWeave, in which Nvidia holds a stake, secured an $8.5 billion investment-grade GPU-backed facility earlier this year. The financing received an A3 rating in large part because debt repayment is supported by contractual payments from Meta.
In another transaction, Broadcom provided a guarantee covering more than 80% of a $35 billion financing structure connected to AI computing capacity for Anthropic.
Nvidia has also previously offered a residual-value guarantee for financing linked to SB Energy’s data-centre project in Ohio, according to S&P Global Ratings 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.
The debate highlights a broader challenge facing the fast-growing AI economy: determining how the rapidly evolving technology hardware should be valued when used as collateral for the enormous amounts of capital required to build the next generation of AI infrastructure.
