Amazon has disclosed a secured loan facility of up to about $80 billion in its third-quarter filing, a new way to pay for the technology hardware behind its artificial intelligence buildout. The facility, revealed in the company’s quarterly report for the period ended September 30, 2026, lets Amazon borrow at its own discretion against purchases of technology infrastructure assets. It is capacity for future borrowing, not cash already raised.
What the filing actually says
The disclosure sits in the debt note of Amazon’s Form 10-Q, filed on October 30, 2026. Amazon says it enters into supplier financing arrangements for certain purchases of technology infrastructure assets, and those arrangements include a secured loan facility. The ceiling is approximately $80 billion, but the amount available will vary with purchase orders and pricing terms.
Borrowings under the facility can be drawn from the first quarter of 2027 through the third quarter of 2028. Each borrowing carries a variable interest rate and matures five years after it is drawn. Amazon says it will grant a security interest in certain purchased assets, and amounts outstanding will be classified as long-term debt on its balance sheet.
The filing also puts the facility in the context of a rapidly growing debt load. The face value of Amazon’s long-term debt stood at $180.9 billion on September 30, 2026, up from $68.8 billion at the end of 2025. The same debt note records a $17.5 billion unsecured term loan drawn in September 2026, maturing in September 2029.
A secured break with past practice
What makes this facility different is the security. Amazon’s bonds, its $17.5 billion term loan and its revolving credit lines have been unsecured, meaning lenders relied on the company’s general credit rather than on specific assets. A secured facility ties borrowing to the hardware itself, giving lenders collateral if Amazon cannot repay.
Amazon has paid for data centre construction and chip purchases from cash generated by its cloud and retail operations, plus its biggest-ever bond sales. It had avoided creative funding arrangements used by some rivals, such as joint ventures to build data centres or deals to lease chips bought by other parties. The new facility moves Amazon closer to that playbook.
What the filing does not say
For all its size, the disclosure leaves several questions unanswered. It does not name the lenders or suppliers involved, it does not state an interest margin, and it does not list which assets will be pledged. It also describes purchases only as technology infrastructure assets, and it does not use the word artificial intelligence.
Bloomberg’s reporting noted that many rival financing deals have focused on Nvidia chips, often a data centre’s most expensive component. That context helps explain why the facility matters to the AI race, but it does not turn the filing into a chip order. Amazon has not said which suppliers will benefit or how much it expects to use.
Rivals are reaching for similar tools
Amazon is not alone in testing ways to fund the buildout. Reports have said Nvidia could offer credit guarantees worth billions to support financing for a data centre planned for Anthropic, although no binding commitment has emerged, as covered in Nvidia’s reported credit backing for an Anthropic data centre. Microsoft and Meta have also turned to bond and stock sales.
That shift is why a single filing can move the wider debate. When the largest cloud providers start pledging hardware or guaranteeing other companies’ debt, AI financing becomes part of the story, not a footnote to earnings. Lenders are watching how quickly borrowing turns into revenue, and a secured loan must still be repaid from cash flows not yet arrived.
Our read: flexibility, with a leash attached
The most important distinction here is between capacity and borrowing. Amazon has not borrowed $80 billion; it has arranged the ability to borrow up to that amount, at its discretion, over eighteen months, subject to purchase orders and pricing. If orders slow, the available amount falls, and treating the headline as money already spent would overstate what happened.
The secured form cuts both ways. Collateral can make lenders more willing to lend at scale, because they hold a claim on tangible equipment rather than only on Amazon’s promise to pay. Pledging assets reduces flexibility later, since equipment tied to one facility cannot be pledged again, and a secured lender ranks ahead of unsecured creditors if anything goes wrong.
Why It Matters
This facility matters: it shows how the AI infrastructure race is being financed, not just built. Amazon’s spending commitments, including debt, interest, unstarted leases and equipment obligations, reached $883 billion by September’s end. That is up 36 percent in three months and more than double a year earlier, as analysts warned that AI’s trillion-dollar buildout carries a growing debt burden.
For customers and competitors, the signal is that Amazon intends to keep building at full speed and change its funding habits to do so. For investors, the signal is more mixed: the company has secured headroom for the next phase, but added another layer of debt to a balance sheet already grown sharply this year.
Background
Amazon reported third-quarter results on October 29, 2026, disclosing a larger-than-expected rise in capital spending, before the filing revealed the facility. The disclosure was first reported by The Information on October 30, followed by Bloomberg, as every major AI builder faces the same question: how to pay for data centres and chips whose costs keep rising faster than revenues.
Source: Bloomberg report on Amazon’s facility, filing terms in detail
0 Comments