Amazon Explores Offloading $8 Billion in Nvidia Chips via Special-Purpose Vehicle
Amazon is exploring a complex financing arrangement to offload approximately $8 billion worth of advanced Nvidia chips into a special-purpose vehicle, according to reporting by the Financial Times. The proposed deal forms part of a broader strategy by major cloud providers to adopt asset-light balance sheet approaches amid skyrocketing capital expenditure demands for artificial intelligence infrastructure.
In recent weeks, the e-commerce and cloud giant has held discussions with institutional investors to gauge market interest. The transaction would transfer thousands of Nvidia Grace Blackwell chips currently installed across more than a dozen United States data centers in five states, including Nevada and Virginia, into the newly established financial entity.
Special-Purpose Vehicle Structure and Debt Issuance
Under the proposed framework, the special-purpose vehicle would secure external financing through debt issuance, enabling institutional participants to fund the high-cost computing hardware. Amazon intends to offer an equity stake of up to 10 percent in the vehicle to investors while leasing the advanced AI semiconductors directly back from the entity.
By transferring the hardware ownership off its primary balance sheet, Amazon would continue operating the servers within its AWS infrastructure without maintaining the entire capital outlay on its corporate books. If the structure achieves an investment-grade credit rating, it is expected to attract widespread participation from institutional lenders such as pension funds and insurance firms.
Escalating Hyperscaler Capital Expenditure Pressures
The discussions highlight the immense capital strain facing major cloud operators as they race to construct AI data centers. Amazon anticipates capital expenditures exceeding $200 billion this year, with a significant majority directed toward Amazon Web Services to procure specialized graphics processing units and expand data center capacity.
Industry analysts note that combined capital investments by major hyperscalers could approach $9 trillion by 2031, while total debt issuance for AI infrastructure is projected to reach $500 billion by 2027. Traditional balance sheet financing has become increasingly prohibitive given the sheer scale of modern server clusters.
Hardware Obsolescence and Collateral Risks
While GPU-backed financing structures are gaining traction within the sector, with firms like CoreWeave previously utilizing similar loan arrangements, the Amazon proposal involves unprecedented scale and bleeding-edge silicon. The arrangement introduces notable risk factors for external lenders.
According to Dealroom, financial analysts cautioned that off-balance-sheet tech financing structures with residual-value guarantees risk obscuring retained liabilities from investors. According to Source, financing backed by graphics processing units (GPUs) as collateral has already gained popularity among companies like CoreWeave.
Advanced data center GPUs face rapid replacement cycles, with hardware often considered obsolete within approximately five years. Consequently, investors financing the special-purpose vehicle are exposed to technological obsolescence, where the long-term economic recovery of the debt depends entirely on sustained demand and the pricing power of leased computing capacity. Amazon and Nvidia declined to comment on the reports outside regular business hours.
When issuing USD 25 billion in bonds in July, investor demand for higher yields stemmed from weakening appetite for long-term debt. Nvidia proposed offering up to $125 billion in debt guarantees in August through a $500 billion financing platform developed alongside Wall Street institutions to decrease borrowing expenses for clients.
Amazon projected a minimum useful lifespan of five years for each semiconductor generation in regulatory filings. Market sources indicate Amazon discussed transferring Grace Blackwell chips to an SPV.