Nvidia announced a major financing initiative this week involving partnerships with six of the world’s largest asset management firms: BlackRock, Blackstone, Apollo, KKR, Brookfield, and Goldman Sachs. The arrangement aims to create a $500 billion pipeline for funding data center and GPU cluster construction for companies without sufficient credit ratings or available capital to purchase advanced silicon directly.
The financing model represents an unconventional application of traditional asset-backed lending principles. Huang positioned Nvidia’s graphics processing units as productive infrastructure assets comparable to commercial real estate or toll roads—investments that generate revenue and retain value over extended periods. Standard asset-backed financing typically relies on the ability to repossess and resell collateral if borrowers default. However, the long-term productive lifespan of cutting-edge GPUs remains uncertain, creating a fundamental challenge to the financing structure.
Industry analysts have identified depreciation as a critical risk factor. GPU chips typically transition from frontier model training applications to lower-margin inference work after several years, directly affecting their resale value and utility as loan collateral. According to market observers, Chinese competition presents the most significant threat to the model’s viability. China is expanding its domestic computing capacity and could potentially initiate a price war using lower-cost silicon, potentially destabilizing GPU market prices and eroding collateral values underlying hundreds of billions in outstanding loans.
To account for these risks, investors are expected to demand higher returns, potentially in the 11% to 17% range, treating GPUs as depreciating equipment rather than durable assets. Additionally, many borrowers in this financing arrangement are anticipated to be non-investment grade entities, including AI startups and alternative cloud providers, which introduces additional default risk into the loan portfolios.
Currently, U.S. export controls limit China’s access to advanced AI chip technology, with Chinese competitor Huawei facing long-standing restrictions. Nvidia maintains approximately 75% market share for AI chips domestically. Rental rates for Nvidia’s H100 chips have increased from approximately $1.70 per GPU-hour in late 2025 to around $2.35 per GPU-hour currently, supporting Huang’s position that the financing model remains economically viable for now.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI