Amazon is in talks to move roughly $8 billion worth of Nvidia Grace Blackwell chips off its own balance sheet and into a newly created financing vehicle, then lease the same hardware straight back for use in its U.S. data centers, according to a report from the Financial Times citing people familiar with the matter. Neither Amazon nor Nvidia has publicly confirmed the deal, and the proposal is still described as under discussion rather than signed. But the structure itself marks a notable shift in how the biggest cloud providers are paying for the AI buildout, and it puts a dollar figure on a financing trend that has been building quietly all year.
If it closes, the deal would see thousands of Grace Blackwell chips, already installed or being installed in U.S. data centers, placed inside a special-purpose vehicle (SPV). That SPV would raise money through debt issuance from outside investors, use the proceeds to effectively buy the chips, and then rent the same silicon back to Amazon Web Services. The hardware never leaves the building. What changes is who technically owns it and who carries the debt used to pay for it.
What’s Actually Being Reported Today
Strip away the financial engineering and the core claim is simple: Amazon wants to turn roughly $8 billion of Nvidia Grace Blackwell chips into a leaseback arrangement instead of carrying them as a straight capital purchase. The Financial Times, the outlet credited with first reporting the story, cited unnamed sources described as familiar with the discussions. That is a meaningful caveat. Nothing here has been confirmed on the record by Amazon or Nvidia, and several of the juicier details circulating online, including specific equity stakes or a final chip count, remain unverified.
What is solid, per the reporting, is the shape of the deal: an SPV holding Grace Blackwell chips, debt raised against that SPV, and Amazon leasing the hardware back for AWS workloads. The chips are already slated for, or already running in, U.S. data centers. Coverage of the story spread fast across financial and tech outlets within hours, including a write-up from Tech Startups, which framed it as part of a broader pattern of Big Tech turning to Wall Street to fund AI compute rather than paying cash up front.
Inside the Proposed Nvidia Grace Blackwell Leaseback
The mechanics of a chip leaseback SPV are not exotic in finance circles, even if they are new territory for AI infrastructure at this scale. Here’s a simplified look at how the reported flow of money and hardware would work:
1. Nvidia Grace Blackwell chips installed in Amazon/AWS data centers
2. Chips transferred into a Special-Purpose Vehicle (SPV)
3. SPV issues debt to outside investors to fund the transfer
4. Investors receive debt-backed returns from the SPV
5. Amazon leases the same chips back from the SPV
6. AWS continues operating the hardware without interruption
The Debt Side
The SPV reportedly plans to raise its funding through debt issuance rather than a straight equity sale. That detail matters because debt investors care about one thing above all: whether they get paid back. An SPV backed by a lease commitment from Amazon looks very different to a bond buyer than, say, a startup’s GPU cluster. Amazon’s credit profile effectively becomes the backbone of the structure, even though the debt technically sits outside Amazon’s own balance sheet.
The Lease Side
On the other side of the ledger, Amazon pays rent to the SPV to keep using the chips. That rent becomes the cash flow that services the SPV’s debt. It is functionally similar to a company selling its headquarters building to a landlord and then signing a 20-year lease to keep working out of the same offices. The building doesn’t move. The company just converts a big upfront asset into a recurring expense.
Why Grace Blackwell, Specifically
Nvidia’s Grace Blackwell platform is the chipmaker’s flagship AI training and inference hardware for the current generation, pairing Nvidia’s Grace CPU with Blackwell GPUs in a tightly coupled rack-scale design. It is expensive, it is in enormous demand, and it is exactly the kind of asset that depreciates fast on paper even when it keeps running useful workloads for years. That combination, high upfront cost plus fast on-paper depreciation, is precisely what makes a chip attractive for an off-balance-sheet financing structure.
There is also a succession question hanging over the hardware. Nvidia has already signaled its next-generation Vera Rubin platform as the follow-up act, a shift our earlier coverage of the Rubin Ultra memory constraints tied to the HBM shortage touched on directly. Every hyperscaler buying Grace Blackwell today is implicitly betting that the chips stay useful long enough to justify the spend before Vera Rubin capacity becomes widely available. Structuring the purchase as a lease, rather than owned capital equipment sitting on the books for years, gives Amazon more room to swap in newer silicon later without stranding an asset.
Off-Balance-Sheet Financing: Wall Street’s New AI Playbook
What makes this story resonate beyond Amazon is the pattern it fits into. AI compute has become so capital-intensive that hyperscalers are increasingly borrowing tools from private equity and structured finance rather than just writing checks from cash flow. An SPV lets a company keep buying hardware at the pace AI demand requires while keeping the related debt off its own primary balance sheet, which can matter for credit ratings, reported leverage ratios, and how Wall Street analysts score the company’s core financials each quarter.
It is not a new idea in corporate finance generally. Airlines have financed jet fleets this way for decades. What is new is applying it to GPUs that depreciate on a completely different timeline than aircraft, and doing it at a scale where a single reported transaction touches $8 billion in hardware. A rundown of the broader financing story was also picked up by AI Weekly, which placed the Amazon talks alongside a wider shift in how hyperscalers are structuring AI capital spending this year.
2026 AI Infrastructure Financing Moves at a Glance
Amazon’s reported SPV talks land in the middle of a busy year for chip-adjacent financing and infrastructure deals across the AI hardware stack. The table below lines up several of the headline moves shattered.io has tracked through 2026.
| Deal or Event | Companies Involved | Reported Figure | What It Signals |
|---|---|---|---|
| Chip leaseback SPV talks | Amazon, Nvidia | ~$8 billion | Off-balance-sheet financing for GPU capex |
| Silicon IP licensing deal | Synopsys and Amazon | $1 billion+ | Custom chip design investment for AWS silicon |
| Share buyback program | Nvidia | $150 billion | Capital return despite heavy AI capex demands |
| Data center campus acquisition | Blackstone, Nvidia | $99.6 million | Real estate changing hands to support AI operations |
| Custom CPU deployment | CoreWeave, Nvidia | 11,264 cores | Cloud providers leaning on Nvidia’s non-GPU silicon too |
| GPU cloud rental pricing | Nvidia B200 cloud rates | $8.01/hour, up 79% | Rising cost of renting the compute these deals finance |
Taken together, the pattern is consistent. Every major player touching Nvidia hardware this year, whether buying it, leasing it, housing it, or renting it out by the hour, is dealing with the same underlying pressure: AI compute costs more than any single balance sheet wants to absorb in one go.
Leaseback vs. Traditional Capex: How the Structures Compare
It helps to see the leaseback approach next to the more traditional way hyperscalers have funded GPU purchases, which is straight capital expenditure paid out of cash flow or ordinary corporate debt.
| Factor | Traditional Capex Purchase | SPV Leaseback Structure |
|---|---|---|
| Balance sheet impact | Asset and any related debt stay on the buyer’s books | Asset and debt sit in a separate off-balance-sheet entity |
| Who owns the chips | The cloud provider outright | The SPV, with investors holding a stake |
| Funding source | Operating cash flow or corporate bonds | Debt issued by the SPV to outside investors |
| Flexibility to upgrade hardware | Lower, since the asset is owned and depreciating on the books | Higher, since the lease can be structured around shorter terms |
| Credit rating dependency | Tied directly to the buyer’s own rating | Often still priced off the parent company’s credit strength |
The appeal of the leaseback column is obvious for a company spending at Amazon’s scale. The risk, which multiple finance commentators flagged after the Financial Times report broke, is that moving debt off the primary balance sheet doesn’t make the underlying obligation disappear. Investors and ratings agencies tend to notice these structures eventually, even when the headline balance sheet looks cleaner.
Why Amazon Needs the Chips Anyway
None of this changes the basic demand picture. AWS is still racing to add AI training and inference capacity for customers who want Nvidia-class GPUs on tap, and Amazon has made its own custom silicon push through Trainium alongside continuing to buy Nvidia hardware in bulk. A leaseback doesn’t reduce how much compute Amazon needs. It changes how Amazon pays for it. The chips keep running the same workloads in the same data centers, whether they sit on Amazon’s books or inside an SPV down the hall, legally speaking.
That is worth sitting with for a moment, because it is easy to read headlines about Amazon offloading chips and assume AWS capacity is shrinking. The opposite is closer to the truth. Companies only bother building SPVs for assets they plan to keep using heavily. You don’t build complex leaseback financing around hardware you’re trying to get rid of.
The Nvidia Angle: A Vote of Confidence, Not a Fire Sale
For Nvidia, a deal like this is a demand signal dressed up as a financing story. Nvidia isn’t the one selling the chips into the SPV in a way that reduces its revenue. Amazon already bought or is buying the hardware either way. What the SPV structure says to Nvidia is that its biggest customers are willing to get creative with financing specifically to keep the Grace Blackwell orders flowing, rather than slowing purchases because of balance sheet strain. That is a far cry from the kind of order cancellation or inventory glut that would actually worry Nvidia’s own investors.
It also reinforces something evident across Nvidia’s own 2026 financial moves. A company funding a $150 billion buyback, as Nvidia has done this year, is not a company short on confidence about continued AI hardware demand, leaseback structures or not.
The Credit Question: Who Buys This Debt
The hardest part of any SPV leaseback is finding investors willing to hold the debt. For a structure like the one reportedly under discussion, the SPV’s bonds would likely need a credit profile strong enough to attract large, conservative buyers such as insurers or pension funds, the kind of institutions that typically demand investment-grade paper before committing serious capital. That almost certainly means the deal’s pricing and structure would lean heavily on Amazon’s own corporate credit strength, even if the debt itself sits inside a separate legal entity.
This is the part of the story regulators and credit analysts will watch closest in the months ahead, assuming the deal moves forward. A wave of similar SPVs across the AI hardware industry, all leaning on a handful of hyperscaler balance sheets to backstop debt tied to fast-depreciating chips, is exactly the kind of concentration risk that ratings agencies get paid to flag early.
Historical Context: How We Got Here
From Cloud Capacity Deals to Chip-Backed Financing
Cloud providers have always juggled huge infrastructure bills, but for most of the last decade that meant data centers, networking gear, and server racks that could run workloads for five to seven years without losing much relevance. GPUs broke that pattern. Nvidia’s generational cadence has accelerated so much that a chip bought at the top of its game can look dated within two to three years, which upends the normal math companies use to justify owning hardware outright.
2026: The Year Financing Caught Up to Demand
This year has produced a steady drumbeat of financing experiments across the AI hardware world, from custom chip licensing deals to real estate acquisitions built specifically around GPU housing. Amazon’s reported SPV talks aren’t an isolated event so much as the latest, largest example of hyperscalers treating AI compute financing as its own discipline, separate from how they’ve historically funded ordinary corporate infrastructure.
Competitive Landscape: How Rivals Finance Their Own Build-Outs
Amazon isn’t alone in facing this math. Every major cloud provider spending heavily on Nvidia hardware this year has had to answer the same question: buy outright, finance through debt, or find a structure that spreads the cost differently. Microsoft, Google, and Meta have all disclosed enormous AI capital spending commitments through 2026, and each has leaned on a mix of corporate cash, debt issuance, and, increasingly, creative financing vehicles for specific chip orders.
What sets Amazon’s reported deal apart is the specificity of the number and the directness of the leaseback mechanism tied to one named chip family. Most competitor disclosures bundle AI infrastructure spending into broader capex guidance without breaking out a standalone financing vehicle built around a single Nvidia product line. If the Amazon structure closes as reported, expect analysts to start asking Microsoft, Google, and Meta directly on upcoming earnings calls whether they’re considering, or already running, similar arrangements of their own.
Market and Analyst Reaction
Because the deal is still reported rather than confirmed, the immediate market reaction has centered more on what the story reveals about industry financing trends than on any single stock price movement. Coverage aggregated by Ground News tracked how quickly the Financial Times report spread across financial media within a single news cycle, a sign of how closely Wall Street is watching any shift in how hyperscalers fund AI infrastructure. Analysts who cover both Amazon and Nvidia have generally read the story as a financing optimization rather than a fundamental change in either company’s AI strategy, though several have noted it as the clearest example yet of an SPV leaseback applied specifically to GPU hardware at this scale.
Amazon’s own newsroom, where the company publishes AWS updates and infrastructure announcements, had not addressed the reported talks directly as of this writing. Readers tracking official statements can follow updates through Amazon’s AWS news page as the story develops.
Risks Hiding in the Fine Print
Every financing innovation carries a flip side, and SPV leasebacks are no exception. Three risks stand out in a structure like the one reportedly under discussion. First, chips depreciate faster than almost any other asset class financed this way, so a lease built around a 2026 flagship chip could face renegotiation pressure the moment a faster, cheaper successor like Vera Rubin ships in volume. Second, concentrating this much debt in a single SPV tied to one chip generation creates a narrow point of failure if AI demand cools faster than expected. Third, regulators have a long memory for off-balance-sheet structures that quietly shift risk around without eliminating it, a lesson carried over from past financial crises involving similar accounting treatments in other industries.
None of that means the deal is reckless. It means the structure deserves the same scrutiny analysts apply to any large, complex financing vehicle, especially one built around hardware that the industry itself expects to replace within a few short years.
What Happens Next: Five Predictions
- The deal closes in a modified form. Expect the final structure, if it happens, to differ somewhat from the earliest reported terms as investors negotiate protections around chip depreciation and lease duration.
- Other hyperscalers test similar structures. Microsoft, Google, and Meta are likely to explore comparable SPV arrangements for their own Nvidia orders within the next few quarters, even if none go public immediately.
- Ratings agencies start publishing guidance specifically on chip-backed SPVs. Given the size and novelty of applying this model to GPUs, expect credit agencies to formalize how they treat these vehicles in hyperscaler risk assessments.
- Nvidia stays financially unaffected either way. Whether Amazon owns the chips outright or leases them through an SPV, Nvidia’s order book and revenue recognition for the hardware sale itself shouldn’t change materially.
- Scrutiny grows once Vera Rubin ships widely. The real test of this leaseback structure arrives when Amazon has to decide whether to renew, renegotiate, or walk away from Grace Blackwell leases as a newer Nvidia platform becomes the obvious upgrade path.
Frequently Asked Questions
Has Amazon confirmed the $8 billion Nvidia chip leaseback deal?
No. The deal was reported by the Financial Times, citing people familiar with the matter, and both Amazon and Nvidia had not publicly confirmed the transaction as of this report. The talks are described as proposed or under discussion, not finalized.
What is a special-purpose vehicle (SPV) in this context?
An SPV is a separate legal entity created to hold specific assets, in this case Nvidia Grace Blackwell chips, and raise financing against them independently of the parent company’s main balance sheet. Amazon would reportedly lease the chips back from the SPV for continued use in AWS data centers.
Which Nvidia chips are involved in the reported deal?
The chips identified in the reporting are Nvidia’s Grace Blackwell chips, the company’s current flagship AI training and inference platform, combining Nvidia’s Grace CPU with Blackwell-generation GPUs.
Does this mean Amazon is reducing its AI chip spending?
Not based on what has been reported. The chips stay installed in U.S. data centers and continue running AWS workloads either way. The leaseback changes how the hardware is financed and who holds the related debt, not how much compute Amazon is deploying.
Who would fund the debt issued by the SPV?
The exact investor group has not been confirmed. Structures like this typically target large institutional debt buyers who want exposure to investment-grade-rated paper, though no specific lenders or investors have been named in the available reporting.
Why would Amazon lease back chips it already essentially owns?
Moving the chips and associated debt into an SPV can keep the obligation off Amazon’s primary balance sheet, which can affect reported leverage and how the company’s core financials are evaluated by credit rating agencies and investors, without disrupting the physical operation of the hardware.
Is this the first time a hyperscaler has used this kind of chip financing?
Off-balance-sheet leaseback financing itself isn’t new to corporate finance broadly, but applying it specifically to a named Nvidia chip generation at this reported scale, roughly $8 billion, is among the most prominent examples disclosed so far in the current AI infrastructure buildout.
What happens to the leased chips once Nvidia’s Vera Rubin platform ships widely?
That hasn’t been addressed in current reporting. It is one of the open questions analysts are expected to raise, since lease terms and renewal options would determine how Amazon handles a transition away from Grace Blackwell once a newer Nvidia platform becomes the standard choice for new AI workloads.
