From Cash Machines to Credit Risks: AI Spending Puts Big Tech Under the Debt Spotlight

With input from Axios and Reuters.

The AI boom just ran into bond math.

Oracle’s latest earnings miss didn’t just knock its stock price — it sent the cost of its borrowing sharply higher. The company’s bond yields jumped to their highest level of the year, a clear sign that lenders now want more compensation to hold Oracle’s debt.

That wobble is about more than one company. It highlights a bigger shift in how investors see the tech giants: less like bottomless cash fountains, more like companies making huge, leveraged bets on AI.

“This is the 1998 moment,” says Anuj Kapur, CEO of software firm CloudBees and a veteran of the dot-com era.

Or, as Mark Malek, chief investment officer at Siebert, put it in a note:

“The promise of AI is meeting the inconvenient math of debt and capex.”

Big Tech isn’t just talking about AI — it’s throwing staggering amounts of money at it.

• The biggest tech firms are expected to spend over $700 billion next year on AI infrastructure, according to Bloomberg data and company earnings reports.
• For years, their “fortress” balance sheets meant they could fund most of that with cash.
• Now, more of that spending is being financed with debt.

Companies like Meta, Google, CoreWeave and Oracle have together issued about $121 billion in investment-grade bonds this year — up from just $17 billion last year, Bank of America data shows.

That kind of step-change gets the bond market’s attention.

Kapur argues that, for AI now, the credit market is more important than the stock market.

Equity investors can fantasize about 10x returns. Bond investors can’t — they just care if they’ll get their money back with interest.

So when lenders start to worry:

• They demand higher yields to own that debt.
• Those higher yields are a direct signal that investors see more risk in a company’s balance sheet and business model.
• Every time sentiment sours on the AI trade, that nervousness shows up first in the bond and credit markets.

It’s not just bonds. Derivatives that insure against default, called credit default swaps (CDS), are also starting to move.

Malek has a particularly vivid line about Oracle’s CDS pricing:

“Oracle’s credit default swaps are priced like the company is jogging barefoot across a field of Legos.”

Some investors use CDS not just as insurance but as a way to make more speculative bets that a company’s credit risk is rising.

To be fair, not all data-center and AI infrastructure giants are maxing out their corporate credit cards.

Many are still funding most of their build-outs with cash, says Joseph Briggs, an economist at Goldman Sachs.

He notes that:

• We did see a significant rise in corporate debt in the late 1990s as tech ramped up.
• Something similar could happen again as AI spending ramps.
• But, in his view, that’s “certainly not the bulk of what’s happening today.”

Goldman’s take for now: we’re not in an AI bubble yet.

For all the hand-wringing over balance sheets, everything still works as long as AI demand keeps exploding.

Kapur says investors are watching one thing above all: enterprise demand.

If there’s even a hint that companies are starting to slow their AI spending:

• Lenders could get jumpy.
• Debt for heavily leveraged AI players would likely get more expensive.
• Credit markets could separate “overextended” from “safe” very quickly.

The era of “AI anything goes” is fading. Investors are no longer impressed just because a company says “we’re spending billions on AI.”

They’re starting to ask:

• How much debt are you taking on?
• What are your financing costs?
• How fast are you burning cash — and when does that turn into real, profitable revenue?

In other words, the market is beginning to rank AI players not just by their growth story, but by the strength of their balance sheets. Who can fund this arms race sustainably — and who’s betting the house?

For the AI trade, that’s the new reality: hype still matters, but credit math matters again too.