Amazon wants to sell billions of dollars worth of Nvidia AI chips without giving up access to them.
The company is in talks with investors over a roughly $8 billion financing deal that would place thousands of Nvidia Grace Blackwell chips into a special-purpose vehicle and lease them back to Amazon, according to the Financial Times.
The proposed deal would allow Amazon to transfer thousands of Nvidia Grace Blackwell chips into a special-purpose vehicle (SPV), then lease the hardware back for continued use in its data centers. The arrangement could help Amazon raise cash while shifting some of the financial burden of owning expensive AI infrastructure to outside investors.
Amazon Looks Beyond Bonds to Finance Its AI Buildout
The proposed transaction comes during an unusually capital-intensive year for Amazon. The company expects roughly $220 billion in capital expenditures in 2026, with spending focused heavily on AWS infrastructure, AI chips and data centers.
Amazon has also tapped bond markets to support that expansion. It increased a planned March bond offering from $37 billion to roughly $50 billion following strong demand, then returned with another $25 billion issuance in July.
According to the Financial Times, Amazon had to offer higher yields on some longer-dated debt during the later sale. A chip-backed financing structure would give the company another source of capital without giving up operational access to the GPUs.
The Nvidia chip financing business is growing
Amazon’s proposal reflects a growing shift in how technology companies finance AI infrastructure.
Rather than funding every data center and semiconductor purchase directly, companies are increasingly turning to external investors, asset-backed borrowing and leasing arrangements. Nvidia is also helping build this market. In August, the chipmaker proposed a $500 billion financing platform with major financial institutions, offering to backstop as much as $125 billion in related debt.
However, the approach introduces a concern that traditional infrastructure financing does not face to the same degree: technological obsolescence. Nvidia’s Grace Blackwell processors are among its most advanced chips, but the upcoming Vera Rubin generation is expected to replace them. Amazon estimates that each semiconductor generation will remain useful for at least five years, according to its regulatory filings.
If demand for older chips weakens faster than expected, investors could find themselves holding assets worth considerably less than anticipated.
What the Deal Could Mean for AWS
For AWS customers, the proposed transaction should have little immediate operational impact. Amazon would continue using the Nvidia hardware through lease agreements, meaning the ownership structure could change without the chips leaving its data centers.
That flexibility could help the cloud provider respond to growing demand for AI training, inference, and enterprise applications while keeping its infrastructure expansion on track.
The bigger significance is financial. If sale-leaseback arrangements become a regular part of hyperscaler spending, cloud companies could gain more flexibility to fund enormous AI infrastructure projects without putting every dollar of equipment directly on their own balance sheets.
For customers, the key question is therefore less about this particular transaction and more about whether alternative financing helps AWS expand capacity economically as demand for AI compute grows.
What's hot at TechRepublic
- Blackpoint Cyber vs. Arctic Wolf: Which MDR Solution is Right for You?
- Why AWS Sellers Choose Deepgram Over Other Voice AI Tools
- SS&C Intralinks DealCentre AI vs. Datasite: Which platform is built for the future of dealmaking?
- SS&C Intralinks FundCentre AI vs. Juniper Square: Which platform better supports modern private markets fund managers?
- Verito vs. Rightworks: Which IT Provider Is Best for Your Firm?
Amazon is betting on demand staying strong
Amazon’s reported plan shows how the AI boom is changing not just the technology inside data centers, but how that technology is financed.
GPUs that were once straightforward capital purchases are increasingly being treated as assets that can be sold, leased and financed by outside investors. That gives hyperscalers another way to keep expanding, but it also pushes more of the industry’s bet on sustained AI demand into financial markets.
If similar deals spread, the next phase of the AI infrastructure race may depend as much on access to capital as access to chips.
Other news: Tencent is reportedly turning to Southeast Asia for AI computing power, with a five-year Oracle deal worth about $7 billion that would give it access to roughly 100,000 advanced AI chips.