Amazon is exploring a plan to transfer about $8 billion worth of advanced Nvidia Nvidia Reduces Initial Financial Backing for OpenAI’s Ohio Data Center Projectartificial intelligence chips to outside investors through a newly created investment vehicle, according to a report by the Financial Times. The proposed arrangement would allow Amazon to raise capital from investors while continuing to use the high-value chips in its US data centres, reflecting the growing financial pressures and strategic choices surrounding the enormous cost of building artificial intelligence infrastructure.
The proposal comes as Amazon continues to expand its computing capacity to support the rapidly increasing demand for AI services. The company has been investing heavily in data centres, advanced processors and other infrastructure needed to operate increasingly sophisticated artificial intelligence systems. Nvidia’s high-performance chips are central to many of these systems, but acquiring and deploying them requires substantial amounts of capital.
Under the plan reported by the Financial Times, Amazon has held discussions with potential investors in recent weeks to determine their interest in a transaction involving thousands of Nvidia Grace Blackwell chips. The processors are being installed across Amazon’s data centre network in the United States and are expected to support the company’s expanding artificial intelligence and cloud computing operations.
Rather than selling the chips outright and giving up access to them, Amazon is considering transferring them to a special-purpose vehicle, commonly known as an SPV. The investment structure would allow outside investors to provide financing for the assets while Amazon would continue using the hardware.
The proposed arrangement would essentially separate ownership of the expensive computing equipment from its continued use by Amazon. The company would lease the Nvidia chips back from the investment vehicle, allowing it to retain access to the technology without carrying the entire cost of owning the equipment directly on its balance sheet.
This type of structure can provide companies with greater flexibility when they are investing in assets that require significant upfront spending. Instead of funding the entire purchase through its own capital, a company can bring in external investors and convert part of the value tied up in physical assets into financing.
For Amazon, the approach could represent a shift toward a more asset-light financial model at a time when artificial intelligence infrastructure is becoming one of the most expensive areas of technology investment. Modern AI data centres require not only processors but also high-capacity networking equipment, cooling systems, electricity infrastructure and large amounts of physical space.
The Nvidia Grace Blackwell chips involved in the proposed transaction are among the most advanced processors designed for demanding artificial intelligence workloads. Nvidia’s Blackwell platform has become an important part of the technology industry’s effort to develop and operate increasingly powerful AI models. These systems require substantial computing resources for both training and inference, making access to advanced chips an increasingly important consideration for cloud providers.
Amazon’s cloud computing division, Amazon Web Services, is competing in a market where major technology companies are spending heavily to expand AI capabilities. Companies are racing to provide customers with access to high-performance computing resources, AI models and specialized infrastructure. That competition has encouraged large-scale investment in data centres and advanced semiconductor equipment.
The proposed transaction would reportedly involve chips that Amazon has either purchased or leased. The processors are being installed across more than a dozen US data centres located in five states, including Nevada and Virginia. Their distribution across multiple facilities reflects the scale of Amazon’s ongoing infrastructure expansion and the geographical footprint required to operate large cloud computing systems.
According to the reported structure, the investment vehicle could raise money from outside investors through debt issuance. Investors would effectively finance the assets held by the vehicle, while Amazon would make lease payments for continued access to the chips. Such arrangements can allow companies to unlock capital from infrastructure while continuing to use the underlying equipment.
Amazon is also reportedly considering taking an equity position of as much as 10% in the special-purpose vehicle. Retaining an equity interest would allow the company to remain financially connected to the assets while transferring a significant portion of the ownership and financing burden to external investors.
The strategy highlights an important challenge facing technology companies in the current AI investment cycle. Demand for computing power has increased rapidly, but the infrastructure required to satisfy that demand is extremely expensive. Advanced AI processors can represent a substantial portion of the capital required to build new computing capacity, particularly when thousands of chips are deployed across large data centre networks.
For Amazon, bringing investors into the ownership structure could provide additional financial flexibility while allowing the company to continue expanding its AI infrastructure. The arrangement could also reduce the amount of capital tied up in physical semiconductor assets, potentially giving Amazon greater freedom to direct its resources toward other areas of its cloud and artificial intelligence business.
At the same time, leasing rather than owning infrastructure creates continuing financial obligations. Amazon would still need to make payments to use the chips after transferring them to the investment vehicle. The overall financial benefit would therefore depend on the terms of the transaction, the cost of financing and the long-term value and useful life of the processors.
The proposal also illustrates how the economics of artificial intelligence are evolving beyond software development. AI companies and cloud providers increasingly have to consider sophisticated financing strategies for the hardware needed to operate their services. As computing equipment becomes more expensive and demand continues to grow, traditional ownership models may increasingly be supplemented by leasing, infrastructure partnerships and other forms of asset-backed financing.
