Nvidia’s ambitious plan to help finance the global artificial intelligence infrastructure boom using its advanced chips as collateral is facing increasing scrutiny from banks and institutional investors, with lenders seeking stronger guarantees and greater protection before treating AI computing hardware as long-term financing assets, according to people familiar with the discussions cited by Reuters.
Nvidia has been working with financial institutions including Blackstone, Apollo and KKR on a financing initiative that could facilitate as much as $500 billion in AI infrastructure investment. The strategy is intended to help AI companies and developers gain access to Nvidia’s computing capacity while creating a financing market around AI infrastructure similar in some respects to established asset-backed financing models.
At the centre of the debate is the expected useful life and residual value of Nvidia’s graphics processing units (GPUs), which provide the computing power required to train and operate AI models. Nvidia argues that advanced AI computing infrastructure can remain productive for many years and therefore support long-term financing. Some lenders and credit investors, however, are taking a more conservative view because the market does not yet have decades of historical data showing how long high-end AI hardware can reliably generate revenue.
Wall Street Seeks Greater Protection
According to banking sources cited by Reuters, some lenders involved in discussions around Nvidia-backed financing are seeking stronger guarantees than were originally contemplated. Potential structures could include guarantees covering the financing or arrangements in which debt is supported by revenue streams from highly creditworthy customers.
Some of the deals currently being developed could therefore provide lenders with additional contractual protections alongside the Nvidia chips used as collateral.
One source cited by Reuters said that tens of billions of dollars in loan transactions under development could feature stronger guarantees and customer contracts. Other structures being considered could similarly provide lenders with additional protection against the possibility that the underlying computing equipment loses value faster than expected.
The discussions highlight the difference between Nvidia’s view of AI compute as a durable infrastructure asset and the approach taken by traditional credit markets, which generally place greater emphasis on predictable cash flows and established depreciation patterns.
Nvidia said its AI computing infrastructure is a productive, durable and fungible asset capable of supporting long-term financing. The company also said financial partners independently assess each transaction, including customer commitments, expected cash flows and residual values.
Nvidia Wants AI Compute to Become an Investable Asset
Nvidia CEO Jensen Huang has described the company’s strategy as an attempt to make AI computing an “investable infrastructure asset” rather than relying solely on conventional arrangements in which customers purchase access to computing capacity.
Under the proposed model, institutional investors and lenders would provide capital to finance AI infrastructure, while Nvidia’s chips would form part of the underlying collateral. AI developers could consequently gain access to computing infrastructure without having to fund the entire cost of the hardware upfront.
Nvidia has said the initiative is also intended to address concerns around so-called circular financing, where companies indirectly help finance purchases of their own products.
The company has indicated that certain transactions could include residual-value guarantees of no more than 25%. The structure is intended to bring independent institutional capital into the AI infrastructure market while limiting Nvidia’s direct financial exposure.
However, the response from lenders suggests that some investors may seek stronger guarantees or additional sources of repayment before committing capital at scale.
Questions Over GPU Lifespans
A major point of disagreement is how long Nvidia’s advanced GPUs can continue generating meaningful revenue.
Huang has argued that Nvidia GPUs can have useful economic lives of up to a decade. Nvidia has also cited third-party research indicating that cloud companies are extending server depreciation periods to five or six years from the three-to-four-year periods commonly used previously.
The company has further pointed to valuations from Barkr, a firm specialising in AI collateral, suggesting that Nvidia’s GB300 NVL72 systems could potentially have useful lives of approximately nine to 10 years.
Credit investors remain more cautious.
Andrew Chang, a director at S&P Global Ratings, said Nvidia’s GPUs have demonstrated that they can remain operational for more than five years, but noted that credit investors continue to take a conservative approach when determining the value of the equipment.
Banks typically use shorter depreciation periods when underwriting GPU-backed financing. According to Impax Asset Management portfolio manager Tony Trzcinka, lenders commonly assess GPUs over a three-to-four-year depreciation schedule, substantially shorter than Nvidia’s projection of a decade of revenue generation.
The difference is important because the longer an asset can generate predictable cash flows, the greater the amount lenders may be willing to finance against it. If investors believe GPUs could become economically obsolete more quickly as newer generations of AI chips arrive, they may require higher interest rates, larger financial cushions or stronger guarantees.
Existing GPU Financing Offers a Comparison
Several recent transactions are providing investors with benchmarks for how AI infrastructure can be financed.
CoreWeave, in which Nvidia has an investment, completed an $8.5 billion GPU-backed financing facility earlier this year that received an investment-grade rating. The transaction was supported significantly by contractual payments from Meta, giving lenders a predictable revenue stream to service the debt.
That structure differs from Nvidia’s broader proposal because the financing does not rely solely on the residual value of GPUs. Instead, investors can assess the strength and reliability of the customer’s contractual payments.
Another example involves Broadcom, an Nvidia competitor that is helping finance computing capacity for Anthropic. Broadcom backed more than 80% of a $35 billion financing structure, providing additional support to attract debt investors.
Nvidia has itself previously provided a residual-value guarantee for financing connected to SB Energy’s Ohio data-centre project, according to ratings agencies S&P Global and Moody’s.
These transactions suggest that investors are currently placing significant emphasis on the revenue streams associated with AI infrastructure rather than relying exclusively on the future resale value of computing hardware.
AI Infrastructure Boom Creates New Financing Challenge
The debate comes as technology companies and investors commit hundreds of billions of dollars to data centres, AI chips, electricity generation and other infrastructure required to support rapidly expanding AI workloads.
The enormous capital requirements have encouraged financial institutions to develop new forms of private credit, vendor financing and asset-backed lending for the sector.
At the same time, the growing complexity of these transactions has prompted comparisons with financing structures used during earlier technology investment cycles. Morningstar analysts have raised questions about the use of private credit, vendor financing and circular financing arrangements as AI infrastructure investment accelerates.
For Nvidia, creating a deep financing market around its computing infrastructure could help expand the overall market for its products. More accessible financing could allow AI developers and data-centre operators to acquire large quantities of computing capacity without bearing the entire upfront capital expenditure.
For lenders, however, the central question is whether the underlying hardware will retain sufficient economic value throughout the life of the debt.
As the AI industry continues to evolve rapidly, newer generations of GPUs can offer substantial improvements in performance and efficiency. That creates uncertainty over how quickly older systems could become less attractive commercially, even if they remain technically operational.
The financing debate therefore extends beyond Nvidia’s own balance sheet. How banks and institutional investors ultimately value AI computing assets could influence the cost and availability of capital for the wider AI infrastructure industry.
For now, lenders appear to be seeking stronger safeguards while the market develops a longer track record for GPU-backed financing. Nvidia’s ability to demonstrate durable cash flows and establish financing structures acceptable to institutional investors will be an important factor in determining how its proposed $500 billion AI financing ecosystem develops.
Disclaimer: This report has been editorially prepared using publicly available information and official company statements. Readers are advised to refer to official announcements for further details.
Author
About the Author
