Amazon is exploring a new way to pay for some of the expensive Nvidia chips powering its artificial intelligence (AI) infrastructure.
Instead of keeping about $8 billion worth of Grace Blackwell processors on its own balance sheet, the company is reportedly discussing a deal that would place the chips in a separate investment vehicle and lease them back for continued use by Amazon Web Services (AWS).
According to the Financial Times, Amazon has been speaking with investors in recent weeks about the plan, which would cover thousands of Nvidia chips already being installed across more than a dozen data centers in five U.S. states, including Nevada and Virginia.
How the $8 Billion Chip Deal Would Work
Under the proposed structure, the Nvidia chips would be moved into a special-purpose vehicle (SPV). This is a separate company created to hold specific assets and raise money around them.
The SPV would issue debt to outside investors, while Amazon could also offer investors an equity stake of up to 10%. Amazon would then lease the chips back from the vehicle, allowing AWS to keep using the same hardware for AI workloads even though the assets would no longer sit directly on Amazon’s balance sheet.
In turn, this arrangement would give Amazon a more asset-light way to finance hardware that costs billions of dollars upfront. Additionally, investors reportedly expect that the vehicle could receive an investment-grade rating linked to Amazon’s double-A credit profile, which could make the debt suitable for large investors such as insurers and pension funds.
Why Amazon Wants Outside Money for Its AI Chips
The proposal comes as Amazon is spending at a level rarely seen even among Big Tech companies.
In July, Amazon raised its expected 2026 capital spending to about $220 billion after AWS recorded its strongest cloud growth in more than four years. According to CEO Andy Jassy, customer demand was still exceeding Amazon’s available computing capacity, even as the company continued building data centers and buying more AI hardware.
As such, moving some Nvidia chips into an externally funded vehicle would allow Amazon to keep expanding that capacity without financing every piece of equipment directly from its own balance sheet. The company has already been using other funding routes, including corporate bonds, as the cost of AI infrastructure continues to rise.
AI Chips Are Becoming Financial Assets Too
Amazon’s talks also arrive as Nvidia is trying to bring far more outside capital into AI infrastructure. In August, Nvidia announced financing partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR aimed at mobilising more than $500 billion for AI computing infrastructure.
However, Wall Street is still debating how much long-term value advanced GPUs should carry, as banks reportedly value these chips using three to four-year depreciation schedules, while Nvidia argues that top GPUs can continue generating revenue for much longer. Hence, lenders have been asking for stronger guarantees and dependable customer contracts before treating chips like established infrastructure assets.
This explains why Amazon’s proposed structure matters, as the investors would be funding expensive Nvidia hardware, but the chips would continue serving one of the world’s largest cloud companies under a lease arrangement.
And if the deal goes ahead, it would show how the AI infrastructure boom is beginning to change how technology companies pay for the computing power needed to keep expanding.
