Nvidia Teams Up with Six Wall Street Giants to Raise Over $500 Billion: Computing Power Is Turning Into a Bond

On August 10, 2026, Nvidia announced that it had signed a memorandum of understanding with six of the world's top asset management institutions — Apollo Global Management, BlackRock, Blackstone, Brookfield Asset Management, Goldman Sachs, and KKR — to jointly establish an independent computing power financing platform. The goal is to gradually mobilize more than $500 billion in third-party capital to support infrastructure construction for frontier AI laboratories, enterprises, and AI cloud service providers. Nvidia's official press release disclosed that the six institutions will establish dedicated capital pools for Nvidia customers, providing financing at attractive interest rates.

For a sense of scale: total enterprise AI infrastructure investment in 2026 is expected to approach $600 billion, and according to Goldman Sachs estimates, AI-related financing already accounts for approximately 25% of total U.S. investment-grade bond issuance. Nvidia's move is not merely a financing arrangement — it is an attempt to establish a financial pricing system for the entire AI computing power market, similar to real estate or infrastructure bonds.

Computing Power Becomes an "Investable Asset" for the First Time

Nvidia CEO Jensen Huang offered a new framework in the announcement: "In the AI era, compute is revenue." He described these AI infrastructure assets as "AI factories," emphasizing that GPUs have evolved from project procurement items into long-term infrastructure that is productive, replaceable, and upgradable. Huang said he approached only these six institutions, and "not one of them declined."

This redefinition is not mere rhetoric. Traditionally, GPUs are classified as IT equipment — they depreciate quickly, have poor liquidity, and are difficult to use as underlying collateral for bank loans. The core of Nvidia's strategy here is to transfer part of the credit risk of computing power assets back onto its own balance sheet by offering financiers a residual value guarantee of up to 25%, in exchange for institutional investors' willingness to participate. According to industry media MonitorDaily, 25% is the lowest residual value support ratio among comparable computing power financing arrangements currently on the market. Nvidia will evaluate on a project-by-project basis whether to provide the guarantee, rather than offering blanket backstop coverage.

From a business logic perspective, this structure offers direct benefits to Nvidia: customers gain access to lower-cost financing channels and can purchase more GPUs; Nvidia does not have to expand its own balance sheet, as risk is borne by the capital pools of the six institutions; and the six institutions gain entry into the high-growth field of AI infrastructure. On the surface, all three parties stand to gain.

The 25% Guarantee Is a Price Discount, Not a Vote of Confidence

However, technology strategist Ben Thompson, writing in Fortune, pointed out the real meaning of this structure: Nvidia's 25% residual value guarantee is, in essence, a disguised price cut — it uses corporate profits to subsidize customers' financing costs rather than directly lowering GPU prices. The significance of this difference is that the former allows the cost of the discount to be embedded in financial statements rather than directly hitting revenue, making it more "manageable" in accounting terms — but the economic substance is no different.

Deeper concerns stem from a mismatch in the nature of capital. A significant portion of the funds participating in this platform will come from insurance float and pension funds — capital designed to pursue safety and long-term stable returns, not to bear the risks of technology infrastructure obsolescence. Thompson warned that using retirement savings to finance AI computing power is "a completely nerve-racking new thing."

Michael Burry's Depreciation Warning

Investor Michael Burry directly challenged the underlying assumptions of this financing structure. He noted that Nvidia GPUs actually have a refresh cycle of roughly two to three years, yet computing power financing contracts typically calculate depreciation based on a useful life of five to seven years. This gap means GPU equipment depreciates in value far faster than the financing models assume. Based on this, Burry estimated that between 2026 and 2028 alone, the industry-wide understated depreciation could approach $176 billion.

This critique points to a structural contradiction: Nvidia's business model depends on rapid iteration (introducing more powerful chip architectures every two or three years), yet the financing system it is helping to build requires underlying assets to maintain long-term stable value. The tension between these two is something no residual value guarantee can fundamentally resolve. The Bank of England has also warned of financial stability risks posed by highly leveraged AI companies, noting that banks lack sufficient visibility into their related risk exposures.

The Circular Financing Debate: Genuine Demand or Self-Fulfilling Prophecy

Huang was explicit in the announcement that the $500 billion comes entirely from third-party capital, independent of Nvidia's investment decisions, in an effort to dispel market concerns about "AI circular financing." He even called the notion of "circular financing" "absurd."

Market concerns about circular financing are not unfounded: earlier reports revealed that chipmakers invest in AI startups, and after receiving funding, those startups then purchase products from the same chipmaker, creating a closed loop of capital within the supply chain that artificially inflates apparent demand. Morgan Stanley analyst Joseph Moore believes that bringing in six independent financial institutions this time "partly alleviates concerns about circular financing, because most of the capital will come from third-party investors."

But the key issue is not whether the capital comes from third parties, but rather whether these third parties' investment decisions are truly independent of Nvidia's ecosystem influence. When Nvidia simultaneously controls the supply of computing power, customer financing channels, and part of the residual value guarantees, the boundary between "genuine demand" and "demand created by the financing structure" becomes difficult to verify from the outside. This is the core transparency gap in the plan, and it is the direction regulators will focus on after the memorandum of understanding is converted into a formal agreement.

Independent Assessment

The strategic value of Nvidia's move is real: it redefines computing power as a productive asset available for institutional capital allocation. If successful, it will structurally resolve the financing bottleneck in AI infrastructure construction and provide a more stable source of demand for Nvidia's future hardware sales.

But this mechanism simultaneously embeds two unresolved risks. First, there is a hard mismatch between GPU depreciation rates and financing terms — no public data yet supports whether the 25% residual value guarantee can cover real value decay. Second, the participation of pension and insurance funds shifts the technical risk of AI computing power onto the social capital least capable of hedging it; in the event of large-scale writedowns, the consequences would extend beyond the AI industry itself.

The $500 billion memorandum of understanding is currently being converted into formal agreements, with the timetable for capital deployment and allocation details not yet disclosed. This financing framework has already established a new industry benchmark: computing power is transitioning from equipment on procurement lists to a line item on global asset allocation sheets.