| Quick Answer Amazon is in talks to move about $8 billion worth of Nvidia AI chips off its own books and into a separate company, called a special-purpose vehicle (SPV). Outside investors would pay for the chips, and Amazon would lease them back to keep using them in its data centres. This lets Amazon keep buying huge amounts of AI hardware without the full cost sitting on its own balance sheet.It does not mean Amazon is slowing down on AI. It means Amazon is finding a new way to pay for it. |
Amazon buys a considerable number of computer chips each year. But the current story about Amazon’s chip-buying tendencies is going viral for a different reason. Amazon is considering a proposal to finance around $8 billion of Nvidia’s AI chips in order to free up its balance sheet. Instead of Amazon buying the chips itself, a separate firm would be able to buy the chips from Nvidia and then lease them back to Amazon. Sounds complicated, but it’s actually very simple once you start unpacking the details. Let’s take a look at what’s actually going on and what it means for how we measure the true cost of building AI.
What Is Actually Going On Here?
Amazon is said to be in talks with external investors about a financing scheme based on thousands of Nvidia Grace Blackwell chips, which represent the company’s largest and most expensive chips to date. The chips were already installed in over a dozen Amazon data centers across the US.
Rather than retaining full ownership of the chips, Amazon would put them in a new company that would be financed by outside investors who would own the chips. Amazon would then lease the chips back and continue to use them in its data centers in the same way it has been doing, for its own AI and cloud computing purposes.
The usage of the chips would remain the same, but the financial risk of depreciation would be transferred to the outside investors who would own the chips.
What Is a Special-Purpose Vehicle, in Plain English?
The financial term for such a separate entity is a special-purpose vehicle (SPV). Think about an SPV as a company that is focused on a single goal and objective. In this particular case, the SPV’s sole purpose would be to own Nvidia chips and receive lease payments from Amazon.
These kinds of entities are not uncommon for financing purposes. For example, airlines use SPVs to finance their aircraft, and real estate developers use SPVs to finance entire buildings. The same reasoning applies to AI-chips: these are expensive and highly demanded, and distributing the risk between multiple private investors might be beneficial for the companies involved.
According to the described scenario, the SPV could also issue its own debt and sell up to 10 percent of the vehicle for investors, thus turning the chips into a financial instrument.
Why Would Amazon Want to Do This?
The short answer is money. Since building AI infrastructure is incredibly costly, and the expenditures keep piling up. Amazon expects to spend around $220 billion on capital assets this year, and the bulk of that spending will be allocated to AWS, data centers, and AI chips.
Such massive expenditures have a significant impact on the company’s financials, especially when billions of dollars in hardware purchases are expensed on the balance sheet. The company’s free cash flow turned negative over the past year, with a considerable amount of cash spent on PPE, which has been driven by AI investments.
By moving $8 billion in chips to an SPV, Amazon will be able to finance its AI ambitions without showing the expenditures on the balance sheet of the parent company. This way, Amazon will be able to grow its presence in the AI space while reassuring shareholders that costs are not ballooning out of control.
Why This Matters Beyond Just Amazon
This isn’t an isolated case. Across the AI industry, companies are finding ways to raise capital to help pay for the huge costs of chips, servers, and electricity. Just one day before news of Amazon’s deal broke, Anthropic announced that Broadcom had agreed to loan it up to $42 billion to fund its computing infrastructure in exchange for a five-year chip lease.
Nvidia itself is pushing for a roughly $500 billion in financing support across the AI industry, and recently approved a $150 billion share buyback, its largest ever. Debt financing for AI infrastructure could balloon to as much as $500 billion by 2027, according to some estimates.
Taken together, these figures suggest something bigger than just one company’s accounting choice. AI chips are set to become a financial asset class unto themselves, akin to aircraft or real estate, complete with leasing deals, debt markets, and private equity vehicles designed to own them.
What Are the Risks?
There are two risks worth understanding here.
- Chips age fast. Unlike buildings or aircraft, AI chips can become outdated within about five years as newer, faster models arrive. Lenders tend to assume a shorter useful life than Nvidia itself claims, which creates disagreement over how these assets should really be valued.
- Debt-funded growth carries risk. If an SPV raises money through debt to buy chips, and AI demand ever slows down, the value backing that debt could drop quickly. That’s a new kind of financial risk tied directly to how fast AI technology changes.
None of this means the plan is a bad idea. It simply means that as AI spending grows, the financial tools being used to support it are getting more complex, and more exposed to how AI technology evolves.
What This Means for the Future of AI Infrastructure
This single deal, if it goes through, won’t change the AI industry on its own. But it’s a clear sign of where things are heading. Big tech companies are spending unprecedented sums on AI hardware, and traditional ways of paying for that hardware are starting to show strain.
Expect to see more deals like this one. Leasing arrangements, special-purpose vehicles, and chip-backed debt are likely to become standard tools for any company serious about competing in AI, not just Amazon. The chips themselves aren’t going anywhere. What’s changing is who technically owns them, and how the bill gets paid.
Frequently Asked Questions
What does it mean to move chips off the balance sheet?
It means the chips are legally owned by a separate company rather than Amazon itself. Amazon still uses them day to day, but they don’t count as Amazon’s own assets or debt on paper.
Is Amazon in financial trouble?
No. This is a financing choice, not a sign of trouble. Amazon’s cloud business, AWS, is growing fast. The company is simply looking for smarter ways to pay for very expensive AI hardware.
What is a special-purpose vehicle (SPV)?
An SPV is a separate legal company set up to own one specific thing, in this case, Nvidia chips. Investors put money into the SPV, the SPV buys the chips, and the main company leases them back. It’s a common tool in industries like aircraft leasing and real estate.
Why does Nvidia care about this deal?
Nvidia benefits either way, since the chips still get bought and used. But deals like this also help Nvidia’s bigger goal of lining up around $500 billion in financing support across the AI industry.
Will this affect regular people or just big investors?
Directly, it mostly affects investors and the stock market. Indirectly, it affects everyone, because it shows how fast AI spending is growing and how deeply it’s becoming tied to the wider financial system.
Does this slow down Amazon’s AI plans?
No. Reports suggest Amazon is still buying Nvidia chips aggressively. This deal is about how the chips are financed, not whether Amazon keeps expanding its AI infrastructure.
