In August, NVIDIA announced that it would collaborate with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to mobilize over $50 billion in third-party capital for AI infrastructure. NVIDIA believes that its computing power can operate like productive assets, generating long-term revenue while helping customers finance data centers and GPU clusters.

However, some banks and credit investors hold a more conservative view. Reuters reports that lenders are seeking stronger guarantees because they are not yet convinced that high-end GPU will maintain sufficient revenue-generating value over the next 10 years to support the debt structure envisioned by NVIDIA.

Wall Street has a more conservative valuation of GPU.

The core of the dispute lies in the lifespan of use.

NVIDIA has always argued that the productivity of top-tier GPU can last for up to 10 years, especially in cases where software continues to expand its use across various workloads. However, banks typically insure GPU with a depreciation cycle of three to four years.

This gap is very important because the longer the time it is expected that GPU will generate cash flow, the easier it will be to finance the assets at a lower cost.

This financing debate takes place against the broader backdrop of the shift in AI infrastructure towards debt financing. Cloud service providers and data center developers often borrow billions of dollars before new capacity begins to generate revenue.

Guarantees may determine whether this model can be scaled up.

NVIDIA initially envisioned financing supported by chips, along with limited residual value support, which in some cases amounted to about 25%. Today, lenders are pushing for stronger protective measures.

Some structures currently under consideration may include NVIDIA guarantees, or long-term revenue contracts from investment-grade clients. This is similar to existing transactions, such as the $8.5 billion GPU support loan for CoreWeave, where the contract payments from Meta helped to maintain the debt rating.

This issue is becoming increasingly important, as GPU has been used as collateral in the AI cloud market, and its residual value is directly incorporated into the credit model.

Another wave of guarantees worth $300 billion in AI also indicates that tech companies are increasingly taking on some of the financing risks behind the infrastructure boom.