The AI buildout is being paid for with an extraordinary amount of borrowed money, and the bill is getting harder to hide. Even as Nvidia (NVDA) sets fresh records, a growing debt pile is making some investors ask how long the party can last. Here is the scale of it and why it matters.
How Big Is the Borrowing?
Hyperscalers and related names like Nvidia have issued about $225 billion in bonds so far in 2026, a jump of nearly 974% through midyear, and are on pace for roughly $400 billion for the full year. Morgan Stanley estimates AI related debt issuance will reach about $570 billion by the end of 2026, more than double the prior year's pace. And that is just the visible part: hidden debt at US tech giants has reportedly ballooned eightfold in four years to around $1.65 trillion, more than the debt that actually appears on their balance sheets.
The Circular Financing Worry
Part of what unsettles critics is how intertwined the money has become. Nvidia has taken equity stakes in the very companies that buy its chips, including a reported $30 billion investment in OpenAI's latest round, plus stakes in CoreWeave, Nebius and Anthropic. In August it lined up partnerships with firms like Apollo, BlackRock, Blackstone and Brookfield to mobilize more than $500 billion in outside capital for AI infrastructure. Bulls call it smart ecosystem building. Skeptics call it circular, with Nvidia effectively funding its own customers' demand.
Why Nvidia Still Wins While Oracle Sinks
Here is the divergence of the moment. The credit scare has hammered Oracle (ORCL) and Broadcom (AVGO), whose debt is funding data centers, while Nvidia keeps hitting records. The reason is simple: Nvidia sells the picks and shovels and collects the cash up front, while the borrowers carry the balance sheet risk. As long as demand for computing power outstrips supply, Nvidia gets paid first, and the risk sits downstream with the companies going into debt to buy its chips.
Could the Debt Trigger a Jump, or a Fall?
This is the trillion dollar question. Bulls argue demand is real and still exceeds supply, which keeps the buildout resilient for now. Bears argue that $1.65 trillion in hidden debt, plus trillions more in off balance sheet commitments, is exactly how bubbles end: with a credit event that forces a sudden repricing. The honest answer is that nobody knows the timing, but the bond market, through signals like Oracle's record CDS, is already demanding a higher price for the risk. Other giants carrying heavy AI spend, from Meta (META) to Microsoft (MSFT), are being watched closely too.
Track Nvidia and the AI leaders on Stock Market ROIFrequently Asked Questions
How much debt is funding the AI boom?
Visible bond issuance by hyperscalers and related firms is on pace for about $400 billion in 2026, and Morgan Stanley sees AI related issuance near $570 billion. Estimates of hidden, off balance sheet debt at US tech giants reach around $1.65 trillion.
What is circular financing in AI?
It refers to chipmakers like Nvidia investing in the same customers that buy their chips. That boosts demand but ties the companies' fortunes tightly together, which worries critics.
Why is Nvidia rising while Oracle falls?
Nvidia sells chips and collects cash, while Oracle borrows heavily to build data centers. The credit scare hits the borrowers first, so Nvidia keeps setting records while Oracle's stock and credit rating slide.
What To Watch
The AI boom is now as much a debt story as a technology story. For now, demand is real and Nvidia is riding it to records. But the growing pile of borrowing, and the bond market's rising unease, mean the next chapter may be written in credit markets, not chip sales. Watch the CDS spreads and October 28, when the Fed's rate decision will reset the cost of all that debt. For the market wide view, see our global markets snapshot.
This article is for informational purposes only and is not financial advice. Always do your own research before investing. Related: Three scenarios for the AI trade: the melt-up, the cooldown, and the crash.
