Amazon is rapidly ramping up its borrowing to fuel an aggressive push into artificial intelligence, raising fresh questions over how much debt Big Tech can absorb before AI returns materialize.

Early June: A borrowing blitz

In the span of roughly two days, Amazon lined up about $31.5 billion in new financing, combining a record C$14 billion (about $10 billion) Canadian bond sale with a new $17.5 billion bank loan.¹² The Canadian deal was billed as the largest corporate bond sale ever in that currency, underscoring investor appetite for Amazon’s paper.¹

Soon after, the company agreed to a $17.5 billion “delayed-draw term loan” led by Citigroup, with JPMorgan Chase, Bank of America, HSBC and Wells Fargo among more than a dozen banks participating.¹² The structure allows Amazon to pull funds as needed through the end of September, with each draw carrying a three‑year repayment window.¹

Mounting leverage for AI buildout

By March 31, Amazon’s total short‑ and long‑term debt, including leases, had surged more than 50% in a year to exceed $225 billion.¹ The company has signaled plans for about $200 billion in 2026 capital expenditures, largely for new data centers and AI chips, after first‑quarter spending alone hit $43.2 billion — the highest among Big Tech.¹

Amazon formally describes the new loan as for “general corporate purposes,” potentially including business investments, capex and debt repayment, while credit analysts say it could also support equity stakes in AI firms such as OpenAI and Anthropic.¹ The unsecured loan carries a relatively low spread of 0.625 to 0.875 percentage points above SOFR, reflecting strong market confidence in Amazon’s ability to service its obligations despite the rapid buildup.¹

Wider AI arms race — and investor doubts

Analysts note Amazon is far from alone: Alphabet is raising tens of billions via equity, and Meta is planning its largest bond sale to date as companies “burn through exorbitant sums of money to keep pace in the AI arms race.”¹² Across Wall Street and Silicon Valley, the core question is converging: can future AI revenues grow fast enough to justify today’s historic borrowing levels?¹²