Nvidia and Wall Street Just Committed $500 Billion to AI Infrastructure

On Monday, Nvidia announced it had signed memorandums of understanding with six of the most powerful financial institutions in the world — Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR — to assemble more than $500 billion in financing for AI infrastructure.

CEO Jensen Huang said he approached only those six firms, and not one turned him down. The deals are structured to provide Nvidia’s customers — AI labs, cloud providers, and enterprises — with access to capital at attractive rates to fund the data centres and compute clusters they need to run AI at scale.

This is not a government programme. It’s not a single company’s balance sheet. It’s the world’s largest chip company recruiting Wall Street’s biggest names to act as a financing arm for the AI buildout. And it’s one of the most significant financial structures ever assembled around a technology sector.

What the Money Is For

Nvidia announced it is partnering with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to assemble more than $500 billion in financing at “attractive rates” for “the buildout of AI infrastructure over time.”

The practical purpose is straightforward. Building an AI data centre capable of training frontier models costs billions of dollars. The GPUs alone — Nvidia’s Blackwell Ultra chips — cost tens of thousands of dollars each. A serious training cluster needs tens of thousands of them. Most companies, even large ones, can’t fund that from their own cash flows.

Such deals are set to start coming to market within months. The compute is also liquid, which means financing could be reallocated to different buyers of the compute, helping reduce the risk to debt investors. That liquidity feature is important — it makes AI compute more like a tradeable financial asset and less like a fixed piece of real estate.

How It Works: Chips as Collateral

The financing structure is genuinely novel. The money is to be raised through bonds issued by project-specific special purpose vehicles (SPVs). Because these bonds are then sliced into asset-backed securities (ABS) and resold to institutional investors and even some individuals, the risk gets distributed broadly across the financial system.

In plain English: Nvidia’s customers borrow money to buy Nvidia chips. Those loans are bundled into bonds. Those bonds are sold to pension funds, insurance companies, and other institutional investors. The chips themselves serve as the collateral.

This is familiar financial engineering applied to an unfamiliar asset class. “Chips have never been treated as a bankable, long-duration asset before, because chips depreciate fast and lose value the moment a newer generation arrives,” Nigel Green, CEO and founder of deVere Group, wrote in a statement. “Turning that into something institutions can lend against, the way they lend against a building or a highway, only works if the underlying asset actually holds its value over time.”

That’s the core risk. AI chips are not buildings. They depreciate. A cluster of Blackwell Ultra chips that cost $10 billion today may be worth a fraction of that in three years when the next generation arrives. Whether the financial structures being assembled around this deal adequately account for that depreciation risk will be one of the most important questions for AI markets in the coming years.

The Circular Financing Concern

There’s a legitimate concern embedded in this structure that analysts have begun raising loudly. The circular nature of many deals — where, essentially, one AI company invests in a second on the proviso that it will buy the first company’s products — has raised concerns that demand for AI may be artificially inflated.

The worry: Nvidia provides financing to customers who use it to buy Nvidia chips. Those customers build AI products they sell to other companies that also buy Nvidia chips. The revenue flows eventually circle back to Nvidia. If AI demand softens — if the enterprise AI adoption curve flattens, or if a major customer hits financial difficulty — the whole structure could experience simultaneous stress.

It’s not a prediction of collapse. It’s a structural feature worth understanding. The same pattern appeared in mortgage-backed securities before 2008 — not because housing was a bad asset, but because the financial structures layered on top of it amplified risk in ways that weren’t obvious until they weren’t.

What It Means for the AI Industry

For AI companies — labs, cloud providers, startups — the practical impact is positive: cheaper, more accessible capital to build the infrastructure they need. “These financing platforms will help customers access scarce compute at scale and build the AI factories that will power every industry and country in the age of AI,” Huang said.

Access to compute has been one of the biggest constraints on AI development outside the largest technology companies. A well-structured financing platform that lets mid-sized enterprises access Nvidia clusters without paying upfront for the full hardware could meaningfully accelerate AI adoption across industries that have been watching from the sidelines.

For investors, the deal represents Wall Street’s most explicit bet yet that AI infrastructure demand is real, durable, and financeable at scale. Six of the most sophisticated financial institutions in the world don’t sign $500 billion commitments on speculation. They sign them when their modelling tells them the cash flows will support the debt service. Their willingness to participate is itself a signal — though not a guarantee — that the AI buildout has fundamental demand behind it.

Whether the chips that underpin all of this hold their value, and whether AI revenues grow fast enough to service the debt being assembled around them, will be among the defining financial questions of the next five years. For more on the AI infrastructure race shaping all of this, see our coverage of AMD’s $5 billion Anthropic deal and the China $295 billion AI infrastructure plan, two other pillars of the global compute buildout.

Read more tech related articles here.