Fed report says AI and advanced computing now drive 31% of corporate bond issuance

Nearly a third of the high-quality corporate debt sold by US non-financial companies is now tied to one thing: building the machinery behind artificial intelligence.

A Federal Reserve report finds that borrowing linked to advanced computing makes up 31% of nonfinancial investment-grade corporate debt issuance.

Big Tech discovers the bond market

The report comes from the Federal Reserve Bank of St. Louis and was released on November 10, 2026. It looks at how the AI investment boom is being paid for, with a focus on the companies building data centers.

The central finding is that these builders are leaning more heavily on debt. The firms in question are the hyperscalers, meaning giant cloud and tech operators such as Microsoft and Amazon. They are spending at a pace their cash reserves were not designed to cover alone.

The numbers on their balance sheets tell the story. According to the St. Louis Fed, the hyperscalers’ cash-to-assets ratio fell from 35% in 2018 to 16% in 2026.

Over the same stretch, the St. Louis Fed found that the hyperscalers’ investment-to-assets ratio nearly doubled.

Goldman Sachs reached a similar conclusion from a different angle. In a late-2025 analysis, the bank estimated that AI-related borrowing accounted for about 30% of net investment-grade corporate bond issuance. That lines up closely with the Fed’s 31% figure for nonfinancial debt.

A record-setting supply of paper

The AI buildout has helped push investment-grade bond issuance to historic levels in both 2025 and 2026. The money is flowing into three main buckets: data centers, semiconductor chips, and the power infrastructure needed to keep all of it running.

Goldman Sachs reported that US investment-grade issuance topped $1.7 trillion in 2025. That came close to the record set in 2020, a year when companies borrowed heavily to ride out the pandemic.

The pace is not slowing. Goldman Sachs forecasts nearly $500 billion in AI-related debt issuance in 2026 alone, with Microsoft, Amazon, and Alphabet leading the way.

Looking further ahead, projections put additional hyperscaler bond issuance at around $440 billion in 2027.

Why this matters beyond Silicon Valley

The shift from self-funding to debt-funding changes who carries the risk of the AI bet. When a company spends its own cash, shareholders absorb the outcome. When it borrows, bondholders become part of the wager, and the broader credit market takes on more exposure to a single theme.

Concentration is one of the clearest concerns. If close to a third of new high-grade nonfinancial debt is linked to advanced computing, a big slice of the investment-grade market now rises and falls with one industry’s spending plans.

The research points to rising bond yields and wider spread premiums on bonds tied to AI projects. A spread premium is the extra interest investors demand over safer benchmarks.

The supply question looms largest heading into 2027. With roughly $440 billion in further hyperscaler issuance projected for that year, the market will be tested on how much AI-linked paper it can digest without demanding meaningfully higher compensation.

Microsoft, Amazon, and Alphabet remain among the most creditworthy companies in the world, which is exactly why they can borrow so much at investment-grade rates.