Amazon is reportedly exploring a new way to finance its growing artificial intelligence business: selling billions of dollars’ worth of Nvidia chips to outside investors and leasing the hardware back for continued use in its data centres.
The proposed deal, valued at around $8 billion, would involve transferring thousands of Nvidia Grace Blackwell chips into a special-purpose vehicle funded by investors. Amazon would then lease the equipment from that entity, allowing its cloud computing division to keep using the chips without owning them directly.
The Financial Times first reported the proposal on 2 October, citing people familiar with the discussions. Reuters later reported that Amazon had spoken with potential investors to assess their interest. At the time of those reports, neither the proposed transaction nor its completion had been publicly confirmed.
The plan comes as technology companies face mounting costs from the AI boom. Building and operating AI infrastructure requires enormous investment in advanced processors, electricity, cooling systems, networking equipment and data centres. Even companies with strong cash flows are looking for ways to fund expansion without tying up too much capital in physical assets.
For Amazon Web Services (AWS), the proposed arrangement could provide additional financial flexibility while keeping its AI computing infrastructure in operation.
However, the potential benefits depend on the details. Lease payments, financing costs and the pace at which AI hardware loses value could all affect whether the arrangement makes financial sense.
HOW AMAZON’S PROPOSED $8 BILLION DEAL COULD WORK
At its core, the plan would separate ownership of the chips from their everyday use.
Under the reported structure, thousands of Nvidia Grace Blackwell chips would be transferred to a special-purpose vehicle, or SPV. This separate legal entity would raise money from outside investors, potentially through debt financing and an equity investment.
Amazon would then lease the hardware back and continue using it in its data centres. According to the Financial Times, the company could also offer investors an equity stake of up to 10% in the vehicle.
The proposed arrangement could follow five steps:
Transfer the hardware: Thousands of eligible Nvidia chips would be placed in a separate investment vehicle.
Secure investor funding: The vehicle would raise capital through debt and potentially equity.
Lease the chips back: Amazon would continue using the equipment under a lease agreement.
Release capital: Amazon could receive proceeds from the transaction, depending on the final terms.
Maintain operations: The hardware would remain available to support computing workloads in Amazon's data centres.
The important point is that this would be a change in financing and ownership, not necessarily a change in how Amazon uses Nvidia technology.
The financial outcome would depend on the final sale price, lease payments, financing costs and length of the agreement. Details such as maintenance responsibilities and the allocation of risks would also matter.
Until those terms are disclosed, it is too early to determine how much the arrangement would ultimately benefit Amazon.
WHY AMAZON IS LOOKING FOR MORE FINANCING OPTIONS
Amazon's AI ambitions come with a substantial price tag.
The company has indicated that its capital expenditure for 2026 could reach approximately $220 billion, with a significant share expected to support AWS, AI computing infrastructure, chips and data-centre expansion.
Against that backdrop, an $8 billion transaction would represent only a portion of Amazon's planned investment. Even so, it could offer another way to finance its expansion.
The costs go well beyond buying processors. AI data centres need reliable power supplies, sophisticated cooling systems, networking equipment, suitable sites and ongoing operational support. Companies must also ensure that the infrastructure generates enough revenue to justify the investment.
Selling selected equipment and leasing it back could allow Amazon to unlock some of the capital tied up in expensive hardware. That money could then be directed towards additional infrastructure, research, cloud services or other business priorities.
The structure could also attract investors who want exposure to the AI infrastructure market without investing directly in technology company shares.
There is, however, an important trade-off. Selling equipment brings in money upfront, but leasing it creates future payment obligations. The arrangement would only offer a clear financial advantage if its overall costs compare favourably with those of owning and financing the chips directly.
In short, Amazon could gain greater financial flexibility without necessarily reducing the total cost of running its AI infrastructure.
WHY NVIDIA’S GRACE BLACKWELL CHIPS MATTER
The proposed deal centres on Nvidia's Grace Blackwell platform, which combines the company's Grace CPU architecture with its Blackwell GPU technology.
These systems are designed for demanding computing workloads, including training and running large AI models. GPUs are particularly useful for AI because they can handle many calculations simultaneously.
Cloud providers rely on this kind of hardware to support generative AI applications, model development and enterprise computing services. As demand for AI grows, access to advanced processors has become a major part of the competition among cloud companies.
For AWS, these chips help provide the computing power businesses need to develop AI products, run models and process large amounts of data.
The reported leaseback arrangement would allow Amazon to continue using the hardware while changing how it finances and owns the equipment.
That distinction matters because advanced chips are expensive assets, and their long-term value depends on both customer demand and technological progress.
The challenge of keeping AI hardware valuable
A new generation of processors does not automatically make older chips obsolete. Previous-generation hardware can remain useful for AI inference, specialised applications and workloads that do not require the latest performance.
Still, newer processors may offer better computing performance, greater energy efficiency or improved capacity. Those advances can put pressure on the rental and resale value of older equipment.
This creates a challenge for investors financing AI hardware. The equipment must remain productive long enough to generate sufficient returns, even as the technology evolves.
Under the proposed structure, Amazon would take on lease payments while investors would own the assets through the investment vehicle. The precise division of responsibilities and financial risks would depend on the final agreements.
For investors, the central question would be whether the chips can generate enough value throughout the financing period to justify the capital committed.
AI INFRASTRUCTURE FINANCING IS ENTERING A NEW PHASE
Amazon's reported proposal reflects a broader shift in how technology companies are funding the AI boom.
Traditionally, large cloud providers have financed infrastructure through operating cash flow, corporate borrowing and direct capital expenditure. But the scale of current AI investment is encouraging companies and financial institutions to explore additional funding structures.
Sale-and-leaseback arrangements are already common in industries such as aviation, property and industrial equipment. An airline, for example, can sell an aircraft to an investor and lease it back, allowing it to keep operating the plane while changing how the asset is financed.
AI infrastructure offers a similar opportunity. Investors could provide capital against computing equipment, while cloud providers retain access to the hardware needed to deliver their services.
For technology companies, the attraction is another potential source of funding. For investors, it offers a way to participate in the expansion of AI computing through equipment-backed investments.
But there are important differences between AI hardware and some traditional infrastructure assets. Semiconductor technology advances quickly, making the long-term value of individual chip generations less predictable.
The economics will depend on whether customer demand, rental income and the useful life of the equipment are sufficient to cover financing costs and deliver acceptable returns.
If those conditions are not met, the arrangement could shift financial obligations without creating much additional value.
WHAT THE PROPOSAL COULD MEAN FOR AWS CUSTOMERS
For AWS customers, the reported deal would not automatically mean less access to AI computing power.
The proposal is based on Amazon leasing the chips back after transferring ownership to the investment vehicle. If the transaction proceeds on those terms, the ownership structure could change while the hardware remains in use.
Businesses relying on AWS for AI training, inference and other cloud computing workloads would not necessarily experience any immediate operational change as a result of the transaction.
The longer-term effects are harder to predict. Financing costs, demand for computing services, power availability and operating expenses will all influence how Amazon expands its infrastructure.
Customers should therefore distinguish between a change in equipment ownership and a change in service capacity.
The proposed transaction, by itself, does not show that Amazon is scaling back its AI investment or moving away from Nvidia hardware. Its immediate purpose appears to be financial rather than operational.
THE NUMBERS THAT WILL MATTER TO INVESTORS
The reported $8 billion valuation has attracted attention, but the headline figure alone does not establish whether the transaction would be financially attractive.
Investors and analysts will need to examine several factors:
Sale proceeds: How much cash would Amazon receive after transaction costs and any existing financing obligations?
Lease payments: Would the cost of leasing the chips compare favourably with the cost of owning or borrowing against them?
Equipment lifespan: How long could the processors remain commercially useful?
Utilisation rates: Would AWS generate enough computing revenue from the hardware to justify the lease?
Residual value: What would the chips be worth when the agreement ends?
Contractual obligations: Who would be responsible for maintenance, insurance, upgrades and other costs?
These details will help determine whether the transaction creates a genuine economic advantage or mainly changes the timing of cash flows.
A sale can release capital, but it does not automatically increase profits by the same amount. Future lease payments and other contractual obligations must also be considered.
For that reason, the proposed $8 billion deal should be viewed as a potential financing strategy rather than an immediate financial gain of equivalent value.
THE RISKS BEHIND THE PROPOSED TRANSACTION
The arrangement could offer Amazon more flexibility, but several risks remain.
AI demand could disappoint. If demand for computing capacity grows more slowly than expected, the revenue generated by the chips might not be enough to justify their operating and leasing costs.
Financing could prove expensive. Investors will expect a return for providing capital. If the required returns push up lease payments, the arrangement could cost Amazon more than direct ownership.
Hardware values could decline. New generations of processors may reduce the attractiveness of older systems, affecting the investment vehicle's assets and eventual resale proceeds.
The structure could add complexity. A separate investment vehicle introduces additional contractual and financial relationships. The final agreements will determine how risks are divided between Amazon, the vehicle and its investors.
There is also a wider industry question. As AI infrastructure spending rises, more companies may explore alternative financing arrangements to fund expansion.
That would not necessarily signal financial distress. Instead, it could reflect the growing importance of capital management in an industry where demand for computing power is rising alongside the cost of supplying it.
FACT CHECK METER
FACT METER: 94% — STRONGLY SUPPORTED REPORTING, BUT THE DEAL REMAINS UNCONFIRMED
Status: The proposed transaction was reported by the Financial Times and subsequently covered by Reuters and other business news outlets. The reporting describes discussions with potential investors, not a confirmed, completed sale.
What published reports indicate:
Amazon was reportedly exploring a transaction involving approximately $8 billion in Nvidia AI chips.
The hardware involved thousands of Grace Blackwell chips.
The proposed structure included a special-purpose vehicle and outside investor financing.
Amazon would lease the equipment back to continue using it.
The chips were reported to be deployed across more than a dozen US data centres in five states.
What still needs clarification:
The $8 billion figure represents the reported proposed transaction value, not confirmed proceeds from a completed sale.
Discussions with investors do not establish that financing has been secured.
The proposal does not mean Amazon will stop using Nvidia chips.
Final lease payments, ownership arrangements and the allocation of financial risks remain dependent on the transaction's terms.
AWX VERDICT: Amazon's reported plan illustrates how financing models are evolving alongside the AI infrastructure boom. The company could unlock capital while maintaining access to advanced Nvidia hardware, but the actual benefits will depend on the final agreements and the equipment's long-term commercial value.
WHAT TO WATCH NEXT
Official confirmation: Watch for announcements from Amazon or formal disclosures indicating whether the proposal has progressed beyond discussions with potential investors.
Financing terms: Details about debt, equity participation and lease payments will help establish the true cost of the arrangement.
Amazon's capital expenditure: Future spending guidance and AWS updates will show whether alternative financing becomes a larger part of its infrastructure strategy.
Nvidia's next generation of chips: New processor launches could influence the useful life and residual value of Grace Blackwell systems.
Investor appetite: Demand for equipment-backed AI financing will indicate whether investors are willing to fund more of the industry's infrastructure expansion.
Cloud industry competition: Other major cloud providers could explore similar arrangements if they offer attractive financing terms without restricting access to essential hardware.
THE AWX BOTTOM LINE
Amazon's reported $8 billion Nvidia chip proposal is about more than selling expensive hardware. It is an attempt to rethink how one of the world's largest technology companies finances its AI ambitions.
By transferring chips to an investor-backed vehicle and leasing them back, Amazon could free up capital while continuing to use the equipment in its data centres. The arrangement could give AWS another funding option as demand for AI computing grows.
But the proposal remains unconfirmed as a completed transaction in the available reporting. Its success will depend on the sale price, lease costs, investor returns and how long the hardware remains commercially productive.
The bigger lesson is that the AI race is no longer just about securing the most powerful chips. It is also about financing, deploying and operating that hardware profitably over time.
ATTLAS WIRE | AWX WORLD NEWS
Real Stories. Lasting Impact.
Comments
No comments yet.