Amazon seeks to offload $8 billion in Nvidia AI chips to investors as Big Tech turns to Wall Street for AI compute financing

Amazon is seeking to move about $8 billion worth of Nvidia AI chips into an investment vehicle backed by outside capital, a financing structure that shows how far the AI infrastructure boom is pushing Big Tech into Wall Street-style financial engineering.

The proposed deal would place thousands of Nvidia Grace Blackwell chips installed across Amazon data centers into a special-purpose vehicle, or SPV, according to the Financial Times, which cited people familiar with the discussions. Amazon would then lease the chips back from the vehicle rather than keep the full hardware value on its balance sheet.

“Amazon.com Inc. is seeking to offload about US$8 billion of advanced Nvidia Corp. chips to external investors through a new vehicle aimed at strengthening its balance sheet,” the Financial Times reported, citing people familiar with the matter.

The SPV is expected to raise money from investors through debt issuance, with Amazon potentially taking an equity stake of up to 10%. Amazon either purchased or leased the chips, which are installed across more than a dozen U.S. data centers in five states, including Nevada and Virginia.

For Amazon, the appeal is straightforward. AI hardware has become enormously expensive, and shifting ownership of those assets to outside investors could let the company keep building compute capacity without carrying the entire cost directly on its balance sheet.

But the deal points to something much bigger.

AI chips are becoming a Wall Street asset class

The AI race is no longer financed mainly through corporate cash flow and traditional technology investment. Increasingly, the physical infrastructure behind AI is being financed through debt, leasing arrangements, special-purpose vehicles and supplier-backed credit.

Amazon’s proposed Nvidia deal arrives just one day after Anthropic disclosed that Broadcom had agreed to lend it up to $42 billion to finance computing infrastructure tied to a five-year chip leasing commitment. Reuters reported that the financing covers roughly one-third of Anthropic’s $125.2 billion commitment and could make the AI company one of Broadcom’s largest chip customers.

That arrangement drew attention partly for its circular structure. A company supplying the technology helps finance the customer buying access to it.

Amazon’s proposed structure is different, but the underlying pressure is similar: AI companies and hyperscalers need staggering amounts of capital to keep buying chips and building data centers.

The Financial Times reported that Amazon expects roughly $220 billion in capital spending this year, much of it tied to AWS, chips and AI infrastructure. The company has already tapped corporate bond markets, and the proposed SPV could give Amazon another funding channel without simply piling more assets onto its own books.

The approach resembles financing structures long used for aircraft, real estate and other expensive infrastructure. The difference is the asset at the center of the transaction: Nvidia GPUs.

That could mark a major shift in how investors think about AI hardware.

Advanced GPUs were once treated mainly as technology equipment purchased by cloud companies. They are increasingly being packaged as income-producing infrastructure, with value coming from the lease payments and compute demand attached to them.

The shift comes as financial markets are already absorbing enormous borrowing needs from hyperscalers building AI data centers. The Financial Times reported that borrowing by major technology companies could reach as much as $1 trillion by 2030 as infrastructure spending climbs.

That creates a new link between the AI boom and global credit markets.

If Amazon completes the transaction, investors would effectively be financing Nvidia hardware that Amazon uses to run its AI infrastructure. The chips generate compute capacity, the compute supports AWS customers, and lease payments flow back to the vehicle holding the hardware.

The model could become increasingly common if the cost of AI infrastructure keeps climbing.

The largest AI companies may still build the data centers and operate the systems. Wall Street may end up owning a growing share of the machines inside them.