Nvidia announced partnerships this week with six major asset managers—BlackRock, Blackstone, Apollo, KKR, Brookfield, and Goldman Sachs—to establish a $500 billion financing pipeline for data center and GPU cluster construction. The initiative, unveiled by founder Jensen Huang during a CNBC appearance, aims to extend credit to companies lacking the financial capacity or credit ratings to purchase millions of dollars in computing hardware directly.
The financing model rests on a critical assumption: that Nvidia’s graphics processing units will maintain their value over extended periods, functioning as stable infrastructure assets comparable to commercial real estate or toll roads rather than depreciating consumer electronics. Huang emphasized that Nvidia’s platform qualifies as investable infrastructure because it generates revenue, maintains fungibility across cloud providers, and supports all major AI models. In traditional asset-backed financing, lenders can repossess and resell collateral if borrowers default, providing protection through established secondary markets and long asset lifespans.
However, the productive lifespan of cutting-edge GPUs remains uncertain. While newer chips handle advanced model training, older generations transition to lower-margin inference tasks, directly affecting their resale value and utility as loan collateral. Ben Emons, founder of FedWatch Advisors and a former Pimco portfolio manager, identified depreciation as the primary risk, warning that chips could lose value faster than anticipated. China represents the single biggest threat in his assessment, as rapid expansion of domestic computing capacity could trigger a price war with low-cost silicon, potentially causing collateral underlying hundreds of billions in private loans to deteriorate faster than debt obligations mature.
Analysts estimate investors will demand high-yield returns ranging from 11% to 17% to compensate for GPU depreciation risks, treating the hardware as high-depreciation equipment rather than real estate. The financing will likely extend to non-investment grade borrowers, including AI startups and neoclouds typically excluded from traditional debt markets. Should these higher-risk borrowers default, Wall Street managers must repossess and resell used chips into a potentially declining market.
Currently, near-term Chinese threats appear constrained. Huawei, China’s primary AI chip provider, has remained on the U.S. Commerce Department’s Entity List since 2019, and U.S. export controls prevent American companies from using its Ascend AI chips. Nvidia maintains approximately 75% market share in American AI chips by most estimates. Meanwhile, rental rates for Nvidia’s H100 chips rose from approximately $1.70 per GPU-hour in late 2025 to about $2.35 per GPU-hour during the current period, reflecting strong near-term economics. Nvidia contends that its CUDA software layer continuously enhances hardware performance after deployment, enabling older chips to remain productive and generate returns longer than traditional depreciation models predict.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI