Nvidia issues $20 billion of bonds to fuel its AI boom

Nvidia is issuing at least $20 billion of bonds, its first since 2021, Bloomberg reported on June 15, citing sources familiar with the matter. The bonds are broken up into seven parts with varying maturity dates between 2–30 years, and the final bond has a spread of approximately 0.9 percentage points over similar U.S. Treasury securities. JPMorgan Chase, Morgan Stanley and Goldman Sachs are handling the issuance, which Nvidia says will go towards “general corporate purposes, including refinancing debt.”

But the important figure here is not $20 billion. Instead, focus on the timeline. Nvidia last issued bonds in June 2021, for a total of $5 billion. At the time, it was a rapidly growing semiconductor company. Today, it’s one of the world’s largest firms by market cap, awash in cash after selling AI accelerator chips that data center operators can’t get their hands on quickly enough. So why does it need to borrow?

1Why the richest company in AI is borrowing

Because debt is cheap. When you can deduct interest payments and reinvest the rest of your capital, borrowing money makes sense. And for a firm with Nvidia’s credit rating, paying less than 1% above Treasury yields is practically free. It allows them to refinance old debt, preserve their cash for share buybacks and M&A, and secure long-term financing while the going is good. There’s nothing wrong with any of these actions in isolation.

However, it’s interesting because Nvidia is part of a trend, according to the article. Along with Alphabet and Amazon, these firms have borrowed hundreds of billions of dollars over the past year to expand their AI computing infrastructure. This is how data centers get built — and it’s becoming clearer that they’re being built with debt, not cash flow. Even the firm with the highest profit margin in the industry is turning to bonds, indicating that the entire sector agrees that AI will be paid for with borrowed funds.

My thoughts: this is the side of the AI hype cycle that you don’t see in a demo, but it’s also the part that I find most interesting. Borrowing money to fund AI compute demand is rational — but the real danger lies in the opposite direction. Bonds have to be paid back regardless of how well (or poorly) the AI market performs, and much of the expenditure financed by bonds relies on future growth. So even if AI demand only shrinks instead of disappearing entirely, those payments will still be made on time.

This won’t affect your daily use of AI apps tomorrow, but it does give insight into the realities of the situation: the infrastructure powering AI is being constructed with leveraged capital and a belief in continued demand. Keep this in mind next time someone tells you that the AI boom is all profit-driven.