NVIDIA Launches $500B AI Financing Platform; Circular Financing Concerns Send Stock Down 2.9%

On August 10, 2026, NVIDIA announced a partnership with six institutions — Apollo Global Management, Blackstone, BlackRock, Brookfield Asset Management, Goldman Sachs, and KKR — to establish a computing resource financing platform aiming to raise more than $500 billion in third-party capital to provide debt financing for AI data centers and GPUs.

Stock Decline Reflects Market Doubts About Demand Authenticity

Following the announcement, NVIDIA's stock closed down 2.9%. The Philadelphia Semiconductor Index fell nearly 3% for the day, with all constituent stocks closing lower. The market reaction was out of step with traditional positive news, with the core concern being "circular financing."

According to Reuters, the six institutions will issue bonds through special purpose entities, using computing power as collateral, and then lease the computing power to NVIDIA customers. The first batch of transactions is expected to launch in the coming months. Jensen Huang said in a CNBC interview that he approached only these six institutions, and none declined.

The Actual Mechanics of Debt Structure and Computing Power Collateral

The platform relies primarily on debt financing with attractive interest rates. Special purpose entities can issue tens of billions of dollars in debt at a time, with funds used to purchase or lease NVIDIA GPUs and related data center equipment. NVIDIA may provide support of up to 25% of project size for individual projects, but emphasizes that all funds come from third parties.

This structure transforms computing power into a tradable asset class. Institutional investors receive stable rental income, while AI labs and cloud service providers gain low-cost computing power. Huang said the move treats computing power as "investable infrastructure assets" that are productive and replaceable.

The Deep Roots of Circular Financing Concerns

The market worries that the financing chain could amplify apparent demand. Chipmakers provide the platform, financial institutions supply funding, customers use the funds to buy chips, which in turn generates new computing demand, forming a closed loop. Similar models have triggered asset valuation bubbles in other industries.

Huang responded that demand is real and investors will make independent financing decisions. The six institutions are not passively participating but are actively assessing project risks and returns.

These are now assets that can generate revenue — productive, long-lasting, replaceable, and flexible. — Jensen Huang

Impact on GPU Rental Prices and Utilization

Around the financing announcement, GPU rental prices rose rather than fell, with utilization and renewal rates for the previous two generations of hardware remaining high. This is inconsistent with the price declines typically seen in bubble assets. Real demand likely comes from frontier AI labs and enterprises' rigid procurement of computing power.

If the platform is successfully rolled out, it will increase market computing supply, potentially easing the supply shortage in the short term, but it may also pull down marginal returns for some projects.

Transmission Path Through the Supply Chain

Financing-accelerated data center construction directly drives demand for servers, power supplies, cooling, and network equipment. In the Taiwan supply chain, server system integrators and power management related manufacturers may benefit first. The pace at which demand materializes depends on the actual contracting speed of the first batch of transactions.

Compared with the 2008 subprime crisis, risk attribution in this transaction is clearer: computing power as collateral can directly generate cash flow, and institutional investors bear credit review responsibility, rather than packaging and transferring risk layer by layer.

Independent Assessment

The platform essentially transforms AI computing power from a procurement expense into a financeable asset, lowering customers' initial capital barrier. In the short term, it will accelerate global data center deployment. In the medium to long term, its success depends on whether end-user applications can generate sufficient commercial returns to cover rental and debt costs. If rental prices continue to rise and utilization remains stable, demand has a genuine foundation; if large-scale defaults occur or rental prices decline rapidly, valuation rationality will need to be reassessed.