Amazon’s balance sheet is now part of its AI strategy. In the space of 48 hours this month, the company arranged roughly $31.5 billion in new financing: a $17.5 billion term loan from a syndicate of banks and a record C$14 billion bond sale in Canada, the largest corporate bond offering ever in that currency.
The loan, signed June 8 and disclosed in a Securities and Exchange Commission filing, is a senior unsecured delayed-draw term facility with Citibank as administrative agent. The lender group includes JPMorgan Chase, Bank of America Securities, HSBC and Wells Fargo, along with more than a dozen additional banks. Reuters first reported the transaction, describing it as part of Amazon’s push to fund artificial intelligence infrastructure.
The structure gives Amazon unusual flexibility. The company can draw the money in tranches as needed before the facility’s September 30 commitment expiration, and each borrowing matures three years from the date it is drawn. Interest is set at Amazon’s option: either a floating base rate with zero margin or term SOFR plus 62.5 to 87.5 basis points, depending on the company’s credit rating at the time of borrowing. The loan carries no financial covenants and allows flexible prepayment, terms that reflect Amazon’s investment-grade standing and its negotiating power in the credit markets.
The Canadian bond sale, completed days earlier, added another roughly $10 billion in dollar terms. Together with the loan, the two transactions gave Amazon more than $27 billion in fresh borrowing capacity, and the company said the proceeds will be used for general corporate purposes.
The scale of the borrowing tracks the scale of the spending. Amazon has told investors it plans roughly $200 billion in capital expenditures for 2026, most of it aimed at data centers, chips and the power and networking systems that support AI workloads. First-quarter capital spending came in above $43 billion, up from $25 billion a year earlier. The strain is visible in the cash flow statement: trailing twelve-month free cash flow fell to $1.2 billion from $25.9 billion a year earlier, reflecting a $59.3 billion year-over-year increase in purchases of property and equipment. Long-term debt stood at $65.6 billion at the end of 2025, up from $52.6 billion a year before, and the new facilities add to that load.
The loan is the second major debt deal in Amazon’s AI financing sequence this year. In March, the company sold about $37 billion of bonds, its largest single borrowing, to fund the same buildout, and it has said repeatedly that it will use every tool available: corporate bonds, bank facilities, leases and, if needed, stock. The March deal and the June facilities together mean Amazon has raised roughly $65 billion in debt this year alone, a pace that would have been unthinkable for the company a decade ago, when its balance sheet was built on the principle of negative working capital.
The spending philosophy comes from the top. Andy Jassy, Amazon’s chief executive, has said repeatedly that the biggest constraint on the company’s AI ambitions is electricity, not demand or engineering talent. A year of $200 billion capital outlays turns that constraint into a line item on bank statements, and the recent financing shows how the numbers are being squared.
Amazon is not alone in the markets. Two weeks before the loan, Alphabet announced an equity raise that grew from $80 billion to roughly $85 billion within 48 hours, the largest stock offering in the company’s history. Berkshire Hathaway agreed to take $10 billion of it. The coincidence of timing is no coincidence: the four largest American technology companies are collectively planning close to $700 billion in AI capital spending this year, and each is now deciding how much of that bill to fund with debt, how much with stock and how much with cash flow.
For Amazon, the choice so far has favored borrowed money. The delayed-draw structure lets the company time its draws to construction schedules, borrowing more as data center projects firm up. CreditSights, a research firm, said the pace of spending could eventually push Amazon toward an equity raise if cash generation does not catch up, a scenario that would dilute shareholders after years of debt-funded growth.
The bond market’s appetite suggests lenders see the story the same way Amazon does. The Canadian issue, split into five parts, was the largest ever in that market, and the term loan was oversubscribed at terms that favor the borrower. Banks are competing to lend to the companies building AI infrastructure, a dynamic that has kept spreads tight across the sector.
What the financing buys, in concrete terms, is time: time for the data centers under construction to come online, for power contracts to be signed, and for the AI services that justify the spending to start generating the revenue that pays the bills. Jassy has framed the buildout as the foundation of Amazon’s next decade, on par with the early investment in AWS. The balance sheet is now committed to that view, one draw at a time.
