NVIDIA issues corporate bonds for the first time in 5 years, borrowing $25 billion. Why still seek financing with so much cash on hand?

Nvidia returns to the bond market after 5 years, raising the issuance size to $25 billion. The market reaction was beyond expectations: up to $85 billion in funds chased these bonds, oversubscribed by 3.4 times.
(Background: Nvidia’s Q1 earnings are off the charts! Revenue hits $81.6 billion, a record high; Jensen Huang enthusiastically declares, “The Age of Agentic AI is Here,” and dividends jump 24 times)
(Additional context: Nvidia plans to invest $2 billion in xAI; Jensen Huang says he wants to participate in all of Musk’s projects)

On Monday, Nvidia filed a financing plan with the U.S. Securities and Exchange Commission (SEC). With $48.6 billion in free cash flow, the company issued corporate bonds again after five years. The final size was increased from the initial target of $20 billion to $25 billion, drawing up to $85 billion in funds that chased the deal, which was oversubscribed 3.4 times.

With so much cash on hand, why borrow?

To understand this bond issuance, let’s first look at a set of figures side by side.

Nvidia’s free cash flow as of April 2026 reached $48.6 billion, and quarterly revenue was $81.6 billion—up 85% year over year. But in the same period, its strategic investment roster was as follows: injecting $5 billion into Intel, investing $10 billion in Anthropic, and participating in OpenAI’s $30 billion funding round. Just these three alone exceed the scale of this bond issuance.

In other words, the speed at which Nvidia is spending money today is faster than the speed at which cash can naturally replenish itself. It’s not borrowing because its pockets are empty—it’s borrowing because it doesn’t want to wait.

Bloomberg Intelligence analyst Robert Schiffman put it even more directly:

“Relatively inexpensive long-term debt financing helps reduce Nvidia’s average cost of capital, and without impairing its AA credit rating, provides funding support for its strategic investments in the AI space.”

Average cost of capital, put simply, is what a company pays on average for each dollar it borrows and each share it issues. Long-term AA-rated debt has lower interest rates than equity financing and does not dilute existing shareholders’ ownership. This is the most standard capital-structure optimization play found in finance textbooks.

This financing is divided into 7 term issuances, ranging from 2 years to 30 years. The yield on the longest-maturity tranche narrowed by 0.25 percentage points versus the initial price guidance, and the final premium was set at a level 0.65 percentage points above U.S. Treasuries.

What can $25 billion buy?

The official stated use of funds is “general corporate purposes, including repaying and refinancing existing debt.” This is the standard wording for corporate bonds—practically speaking, it can cover almost anything.

But based on Nvidia’s recent moves, the direction is already fairly clear: the $5 billion for Intel, the $10 billion for Anthropic, Nvidia’s participating share in OpenAI’s funding round, plus the data center construction that is expected to keep expanding over the next few quarters. The destination of the funds is already quite obvious—betting that its own ecosystem can drive and lead the next wave of AI infrastructure buildout.

Compute power, models, and capital—these three lines are forming a closed loop at Nvidia.