Amazon has secured a $17.5 billion loan from a group of major banks: Citigroup, JPMorgan Chase, Wells Fargo, HSBC, and Bank of America. The structure of the deal is worth noting: Amazon does not have to take the money all at once. It can draw down funds as needed through September 30, 2026, and repay each amount over three years from whenever it borrows. That flexibility matters when you are building at the scale Amazon is building.
This came just two days after Amazon filed to raise roughly $14 billion in Canadian bonds, bringing its new financing total to over $31 billion in 48 hours. To put that in context: $31 billion is larger than the entire annual revenue of many Fortune 500 companies.
Amazon plans to spend $200 billion on capital expenditure in 2026. Most of that goes toward data centers and the computing hardware needed to run AI services. In the first quarter of 2026 alone, the company spent $44.2 billion, up from $25 billion in the same period a year earlier. Free cash flow, the money left over after all that building, has dropped sharply as a result: from around $26 billion on a trailing twelve-month basis to just $1.2 billion.
Amazon CEO Andy Jassy has pushed back on the concern. He draws a direct parallel to how Amazon built its cloud computing business, Amazon Web Services, which required years of heavy spending before it became the profit engine it is today. AWS is now growing at 24% per year and posted $35.6 billion in revenue in the last quarter of 2025 alone. Amazon's AI-specific business has separately crossed $15 billion in annualized revenue. The argument is that the same playbook works at a larger scale.
Credit agency S&P has warned that Amazon's debt load will increase substantially and that the company is likely to report negative free operating cash flow over the next two years. That is a known trade-off, not a surprise. Amazon's credit rating remains AA, which is why banks are comfortable lending at this scale.
What makes this bigger than an Amazon story is that every major tech company is doing the same thing at the same time. Meta announced a $30 billion bond sale. Alphabet disclosed plans for $85 billion in fundraising. Combined, the big technology companies are on track to borrow close to $570 billion this year to fund AI infrastructure, according to Morgan Stanley. That is more than four times what was raised over the same period in 2025.
The practical consequence for everyone else is twofold. First, this level of borrowing by a small group of very large companies absorbs a significant portion of available capital in credit markets, which can push up borrowing costs for other companies looking to raise debt, including non-tech businesses. Second, the infrastructure being built is what AI services run on: the data centers, the computing power, the storage. Whoever controls that infrastructure will set the terms for anyone who wants to use AI commercially, whether that is a hospital, a logistics company, or an insurance firm.
The spending is enormous, the debt is real, and the returns are not guaranteed. But the companies doing the borrowing are among the most creditworthy in the world, and the demand for AI computing capacity is, by all current signals, growing faster than it can be built.