Amazon.com Inc. has engaged institutional investors to evaluate offloading approximately $8 billion worth of cutting-edge Nvidia artificial intelligence chips into a newly formed special-purpose vehicle (SPV), according to a report published by the Financial Times on October 2, 2026. Under the contemplated structure, Amazon Web Services (AWS) would transfer thousands of recently installed Nvidia Grace Blackwell processors to the external financing vehicle and immediately lease the computing capacity back to power its cloud infrastructure.
The proposed transaction marks a significant departure from the direct-ownership model that has characterized Big Tech’s multi-billion-dollar AI capital expenditure cycle over the past three years. By offloading physical semiconductor assets to outside debt and private-credit investors while retaining operational control of the compute capacity, Amazon aims to adopt an “asset-light” posture, easing balance-sheet pressure as hyperscalers face market scrutiny over escalating infrastructure commitments.
Key Takeaways
- $8 Billion Hardware Carve-Out: Amazon has held exploratory talks with institutional investors to move approximately $8 billion of advanced Nvidia Grace Blackwell AI chips into an independent special-purpose vehicle.
- Leaseback Mechanism: Under the structure, the SPV will fund the chip acquisition by issuing debt to external credit investors, while Amazon leases the processing hardware back to service enterprise cloud clients.
- Minority Equity Structure: Amazon plans to retain an equity stake of up to 10% in the vehicle, keeping the entity largely off its consolidated direct asset registry.
- Geographic Deployment: The semiconductors involved comprise thousands of high-performance units currently being installed across more than a dozen AWS data center facilities in five US states, including critical hubs in Virginia and Nevada.
- A Shift in Hyperscaler Financing: With Big Tech annual capital expenditure approaching record highs, the deal introduces financial engineering—mirroring commercial aviation and energy project finance—into the generative AI hardware supply chain.
Transaction Architecture: How the AI Chip Leaseback Works
The contemplated deal adapts traditional project-finance mechanisms—commonly used in aircraft fleets, shipping containers, and commercial real estate—to the rapid depreciation cycles of advanced semiconductor clusters.
THE PROPOSED AWS CHIP SPV
│
▼
[Institutional Debt & Equity Investors]
(Private Credit, Pension Funds, Insurers)
│
Funds Purchase (Debt Issuance)
│
▼
┌───────────────────────────────────────┐
│ Special-Purpose Vehicle │
│ • Owns $8B of Grace Blackwell Chips │
│ • Amazon holds <= 10% equity stake │
└───────────────────────────────────────┘
│
Long-Term Hardware Lease
│
▼
┌───────────────────────────────────────┐
│ Amazon Web Services (AWS) │
│ • Deploys chips across 5 US States │
│ • Operates compute for cloud clients │
│ • Pays recurring lease distributions │
└───────────────────────────────────────┘
Historically, hyperscale cloud providers—Amazon, Microsoft, Alphabet, and Meta—have purchased microprocessors directly, recording them as Property, Plant, and Equipment (PP&E) on corporate balance sheets and depreciating them over useful lifecycles typically modeled between four and six years.
Under the mechanism detailed by the Financial Times:
- Asset Ring-Fencing: Amazon bundles thousands of Nvidia Grace Blackwell superchips that are either newly acquired or already deployed across more than 12 data centers in states such as Virginia and Nevada.
- Transfer to SPV: Ownership of the physical hardware is sold or assigned to the newly created financing vehicle at a valuation near $8 billion.
- Debt Issuance to Third Parties: The SPV raises long-dated capital primarily by issuing fixed-income or private debt to institutional asset managers, private equity credit arms, and sovereign allocators seeking asset-backed yield.
- Off-Balance-Sheet Retainage: Amazon limits its direct equity exposure by retaining no more than 10% of the vehicle’s equity, allowing the underlying asset weight to sit outside its core corporate liabilities.
- Operational Continuity via Leaseback: AWS enters into a master service and leaseback contract with the SPV, paying fixed or usage-indexed leasing tariffs while selling compute runtime directly to enterprise clients, model builders, and AI startups.
Why Now? The Pressure Behind the $200 Billion CapEx Wall
The move arrives against a backdrop of escalating capital requirements among US cloud giants. Amazon previously signaled that its full-year capital investments could reach unprecedented highs, with the vast majority directed toward AWS to procure high-end accelerators, secure power grid allocations, and construct specialized liquid-cooled data halls.
| Company / Dimension | Traditional Hardware Model | Contemplated SPV Leaseback Model |
| Balance Sheet Treatment | Capitalized on balance sheet under PP&E | Offloaded to external SPV; treated as operating/lease obligation |
| Capital Source | Corporate cash flows, commercial paper, debt | External private credit and institutional debt markets |
| Depreciation Risk | Company absorbs full obsolescence risk if architecture shifts | Shared between equity syndicate and corporate tenant |
| Credit Rating Impact | Drives elevated leverage metrics and Free Cash Flow drag | Protects cash flow from operations, preserving corporate liquidity |
| Margin Profile | Higher gross margin; burdened by high upfront CapEx | Predictable OpEx lease schedule; reduced upfront capital outlays |
Managing the Accelerated Obsolescence Trap
While server chassis, fiber cabling, and physical real estate maintain utility over decades, frontier artificial intelligence processors depreciate along aggressive technological curves. Nvidia’s rapid cadence—transitioning from Hopper (H100/H200) to Blackwell (B200/GB200) within 18 months—creates severe impairment and replacement cycles.
If an enterprise amortizes billions in microchips that become cost-inefficient after 36 months due to next-generation silicon gains, balance sheets carry the risk of heavy writedowns. By shifting $8 billion in hardware assets into a vehicle funded by outside credit markets, Amazon effectively transfers residual asset risk to institutional capital, exchanging lumpy capital outlays for steady operating lease payments.
Market Repercussions: A New Era for AI Private Credit
If closed, Amazon’s transaction will set an operational benchmark for how the artificial intelligence buildout is financed globally.
1. The Emergence of “Compute-Backed Securities”
Wall Street investment banks and private credit mega-funds—including Blackstone, Ares Management, Apollo Global Management, and Blue Owl Capital—have accumulated hundreds of billions of dollars in dry powder aimed at digital infrastructure.
While private lenders previously underwrote the physical shells of data centers and long-term energy purchase agreements (PPAs), debt secured directly by high-end microchips has historically faced underwriting hurdles because GPUs were viewed as liquid, rapidly depreciating assets. A structured transaction with a tenant of Amazon’s creditworthiness validates “silicon financing” as a distinct institutional asset class.
2. Strategic Flexibility for AWS Custom Silicon
The deal also provides strategic flexibility for AWS’s internal chip division. While Nvidia’s Grace Blackwell platform remains the gold standard for training frontier models, Amazon has continued pouring capital into its in-house Trainium and Inferentia silicon to lower inference costs.
Offloading capital tied up in third-party merchant silicon allows AWS to preserve balance-sheet flexibility to invest in its own custom ASIC development, proprietary liquid-cooling infrastructure, and nuclear energy partnerships.
3. A Template for Competing Hyperscalers
Amazon is rarely alone in its balance-sheet strategies. Microsoft and Alphabet have faced parallel inquiries from institutional equity analysts regarding their free-cash-flow conversion ratios as AI infrastructure buildouts expand. If Amazon executes an $8 billion hardware transfer smoothly, peer cloud providers are likely to explore similar vehicle structures to fund secondary accelerator fleets without expanding gross corporate debt.
Unresolved Questions and Execution Risks
While the structure offers clear balance-sheet relief, several financial and operational questions remain subject to market negotiation:
- Debt Pricing and Coupon Yields: Because AI chips carry high technological obsolescence risks, institutional debt investors will demand an appropriate yield spread above benchmark sovereign debt. What coupon will the SPV need to offer to clear the debt issuance?
- Accounting Treatment: Auditors and credit rating agencies (Moody’s, S&P Global) will closely scrutinize whether the leaseback qualifies as an operating lease or must still be capitalized under ASC 842 rules, which require long-term lease liabilities to appear on corporate balance sheets.
- Secondary Liquidation Terms: In the event of a theoretical contract default, repossession and secondary liquidation of integrated, water-cooled server racks inside secure proprietary AWS data centers present complex physical and legal challenges.
Frequently Asked Questions (FAQs)
Why is Amazon looking to offload $8 billion of Nvidia chips?
Amazon aims to strengthen its corporate balance sheet and adopt an asset-light financing model. By selling the physical Nvidia Grace Blackwell chips to a special-purpose vehicle and leasing them back, Amazon converts heavy upfront capital expenditures into predictable operating leases while sharing obsolescence risks with institutional investors.
What chips are involved in the proposed deal?
The transaction involves thousands of high-performance Nvidia Grace Blackwell (GB200) chips that Amazon has purchased or leased, which are currently being installed across more than a dozen data center sites in five US states, including major cloud hubs in Virginia and Nevada.
Will Amazon stop using these Nvidia chips?
No. AWS will continue operating the chips seamlessly within its data center network. Under the sale-and-leaseback agreement, Amazon will rent the processing power back from the investment vehicle to provide high-performance compute capacity to its cloud customers.
How is the investment vehicle structured?
Amazon plans to hold a minority equity stake of up to 10% in the special-purpose vehicle (SPV). The remaining funding will be secured by issuing debt and equity to third-party institutional investors, such as private credit funds, pension funds, and asset managers.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
