AI's Debt Surge Is Unsustainable — Hidden Borrowing Reaches $1.65 Trillion

As AI hyperscalers continue their massive investment plans, bond issuance and hidden borrowing by related companies are rapidly increasing. Amazon recently raised its capital expenditure outlook, and borrowing by related companies including Nvidia is growing larger.

According to S&P Global, hyperscalers and related companies have issued $225 billion in bonds so far in 2026. This is a 973.7% increase compared to the same period last year, and if the current trend continues, annual issuance is projected to reach $400 billion.

While the market is still absorbing this volume, leverage is rapidly rising among companies previously regarded as having strong cash flows, and bond yield premiums are also increasing. S&P Global noted that market participants have begun to grow wary of the rapid increase in leverage.

At the same time, the U.S. government must issue large amounts of Treasury bonds to cover a fiscal deficit expected to reach approximately $2 trillion this fiscal year. With the Federal Reserve no longer the major buyer of Treasuries that it once was, the burden on private investors is growing.

Capital Economics assessed that if the trend seen in the first half of this year continues into the second half, corporate and government bond issuance relative to GDP could reach an all-time high, excluding the COVID-19 period. While bond demand is currently strong, RSM warned that this demand will not last forever.

The hidden borrowing of five major U.S. tech companies is reported to have swelled to $1.65 trillion, raising the possibility that the actual cost of the AI boom is greater than what has appeared in public bond markets.