Nvidia struck agreements with six of the largest Wall Street asset managers this week to secure $500 billion in financing for AI data center construction, but analysts warn the entire structure depends on how long the company's chips hold their value in a market increasingly shaped by Chinese competition.
The deals with BlackRock, Blackstone, Apollo, KKR, Brookfield and Goldman Sachs aim to fund GPU clusters for companies that cannot afford to buy the hardware outright. In a CNBC segment flanked by executives from all six firms, Nvidia CEO Jensen Huang pitched the plan as a straightforward infrastructure financing play.
"Nvidia's AI factory platform is really an investable asset, an infrastructure asset," Huang said. "The reason for that is because it's productive, it's revenue generating, it is fungible, it's used by just about every cloud service provider, it runs every AI model."
Why chip lifespan is the core risk
Standard asset-backed finance works because lenders can repossess and sell collateral if borrowers default. Buildings, cargo ships, and toll roads have decades-long secondary markets. GPU chips have no such track record.
New Nvidia processors power the most demanding AI training work for three to five years, then shift to lower-margin inference tasks. That transition directly reduces their resale value and the collateral backing hundreds of billions of dollars in loans.
"Depreciation is the one key risk here," said Ben Emons, founder of FedWatch Advisors, who structured similar asset-backed loans for IndyMac before working as a portfolio manager at Pimco. Nvidia chips "could depreciate faster than expected," he said.
China could trigger a price war
Emons identified China as the single biggest threat to Nvidia's financing model. Chinese producers are rapidly expanding domestic compute capacity and could flood the global market with low-cost silicon in a price war.
If Chinese production drives hardware prices down, the collateral backing private loans could erode faster than debt terms. Emons estimates investors will demand high-yield returns of 11% to 17% to offset this risk, treating GPUs as high-depreciation equipment rather than real estate.
The borrowers themselves add another layer of risk. Most are likely to be non-investment-grade AI startups and neoclouds locked out of standard debt markets, according to a Bank of America Securities note. If those firms default, Wall Street fund managers must repossess and sell used chips into a potentially falling market.
Any immediate China threat is paused. Huawei, the dominant Chinese AI chip supplier, has been on the U.S. Commerce Department's Entity List since 2019. In May, the U.S. government said Huawei's Ascend AI chips violate export controls, barring their use by American companies.
Nvidia's counterargument: software extends chip life
For now, the economics are moving in Nvidia's favor. Rental rates for H100 chips rose from roughly $1.70 per GPU-hour in late 2025 to about $2.35 per GPU-hour this year, Huang said, driven by scarcity as hyperscalers race to build capacity.
Nvidia argues its CUDA software layer will get older chips more productive revenue after deployment, allowing them to generate yield longer than traditional depreciation models predict. The company says consistent software updates preserve chip value over time.
Nvidia still commands over a 75% market share in AI chips and remains by far the leading domestic supplier.
Why this matters for finance and government professionals
The $500 billion financing pipeline ties Nvidia's hardware directly to financial assumptions about depreciation and collateral recovery. Professionals working in asset-backed finance, infrastructure investing, or corporate credit should watch two thing: the spread on the loans (if it reaches 11% to 17%, the market is pricing real risk) and the resale market for used GPUs. For government officials, the structure raises a question: if collateral values erode, the government may face pressure to backstop infrastructure assets that were built for private sector production. That is not a theoretical outcome. It has happened with toll roads, warehouses, and other infrastructure assets that promised predictable returns.
Nvidia's situation is simpler: depreciation schedules and the secondary market price of used chips will settle the question, not press releases or TV segments.
AI for Finance professionals can track these trends directly: the collateral modeling in AI infrastructure loans is now a live case study in how quickly hardware assets can lose value and how markets price that unknown into yields.
Your membership also unlocks:
