AI Is Reaching Into the Bond Market—Is a Prelude to 2008 Unfolding?

Over the past two years, the most prominent area for AI has been the stock market, with the valuations of leading AI-related companies being driven higher and higher.

The more important change now is: AI is turning massively to debt. The five major hyperscalers—Amazon, Alphabet, Meta, Microsoft, Oracle—have issued $121 billion in US corporate bonds in 2025, while the annual average from 2020 to 2024 was only $28 billion. The forecast for new debt issuance by the Big Five in 2026 has been raised to $175 billion. Global AI-related debt issuance may approach $570 billion in 2026.

This is because training and running AI is not just writing a few lines of code; it requires extremely expensive real-world infrastructure: chips, servers, data centers, land, electricity, cooling systems, fiber optics, and long-term leases. In other words, AI has long since shifted from a “light asset, software story” to a “heavy asset, infrastructure story.”

This is where the bond market comes in to foot the bill, as capital from the equities market alone is no longer sufficient.

According to the latest reports, Amazon is planning to issue at least $25 billion in bonds for AI-related capital expenditures and debt repayment.

Despite so many AI-related bonds being issued now, corporate bond spreads have not skyrocketed immediately.

The reason is simple: many of the biggest issuers are “super blue chips”—Microsoft, Alphabet, Amazon, Meta. These companies have strong cash flows, high profits, and excellent credit ratings. Bond investors think: even if some AI investments are wasteful, it’s unlikely that these companies couldn’t pay back their debts.

This is also why the market looks very calm now. Spreads for many investment-grade bonds remain very low, meaning investors have not yet asked these companies to pay significantly higher extra interest.

But this is precisely one of the hidden risks. Financial crises often do not start with assets that “everyone knows are hazardous,” but with those that “everyone assumes are perfectly safe.”