Behind the Yield

When Spreads Forget Risk: Lessons from the Tech Bubble for Today’s AI Debt Boom

By Charles Urquhart, CFA · September 30, 2025

When Oracle sold $26 billion of new bonds in September, it was pitched as an ordinary deal to finance most of the company’s purchase of a cloud computing firm. On paper, just another mega deal. In fact, it was a reminder of just how bubbly credit markets become when a hot narrative takes hold. A borrower with a BBB+ rating was effectively treated like crown jewels, with $88 billion in orders at spreads hardly wider than Microsoft’s. That’s not price discovery. That’s a stampede.

Deals like this do not occur in a vacuum. They occur because investors have collectively talked themselves into believing that A.I. is the next great wave of corporate growth. Anything associated with it sustainable or sensible be damned is money well spent. Not the first time credit markets have suspended disbelief. If history is precedent, it won’t end well. Spreads are supposed to be the reward for risk. When they compress into nothingness, they cease to function and instead blink complacency.

Back to the Tech Bubble

If this sounds familiar, it should. I was there when the dot com bubble showed us how pernicious it can be to let balance sheets take a back seat to story. At the time, spreads collapsed under the belief among investors that “new economy” companies did not have old economy risks. Debt had piled up, covenants were flouted, credit ratings brushed aside. It was all working until it stopped. When it flipped, the same spreads that had looked tight and efficient turned into the trap door.

Telecoms were the easiest example. WorldCom, Global Crossing and Qwest all had built empires on cheap debt and optimistic assumptions. Their bonds traded so tight due to the story, and it was a good one. But when the story broke, it was credit investors who took the pain. Defaults mounted, recoveries were paltry and spreads blew out overnight. It was a brutal but simple lesson: credit cycles do not bend to hype.

It wasn’t just telecoms. Big tech names borrowed billions to fuel expansion, and when expectations reset, equity and credit investors suffered. Loans had been issued at tight spreads that left no room for deterioration.

Three Lessons Credit Markets Keep Forgetting

1. It’s when markets lose the ability to differentiate risk … until they can’t.

And when the gap in yield between AAA bonds and BBB contracts to the thickness of a dime, it’s not because of some magic achievement at a BBB company. It’s because investors quit asking tough questions. That satisfaction came to a screeching halt in 2000. It will again. That Oracle could raise capital at spreads near indistinguishable from Microsoft is a measure of how far the markets have travelled down that road already.

2. There is leverage in “story stocks” both ways.

Debt feels inexpensive and endless when spreads are tight. “Growth” companies issue all the equity they can because capital is practically free. But leverage is like a spring. When the fundamentals change, it snaps back hard, and losses get magnified. Investors tell themselves additive debt to fuel transformative technology is special. It’s not. Although leverage is about multiplying bets in both directions.

3. Credit markets usually follow several weeks behind stocks until they don’t.

The equity markets hog the limelight, but it’s credit that silently determines how long the party goes on. When spreads ultimately do adjust, they tend to do so quite brutally. We saw it with the telecoms in 2002. We saw it, once again, in structured credit in 2008. The AI build out isn’t exempt. The longer risk is ignored, the more violent the eventual correction.

AI Isn’t a Free Pass

Maybe AI will be transformational. Maybe it will “justify” investment booms and big ticket financings. But none of that changes the mechanics of credit cycles. Saying that doesn’t change the fact that BBB balance sheets are still awful, or that leverage is still plenty dangerous. Sure, Oracle got the red carpet treatment this time but that was more about the desperate desire for yield than Oracle’s own safety. And it’s not just Oracle, either. Smaller AI linked companies have been issuing outright garbage, with some marching up to the line between investment grade and high yield. And the market’s telling them, quite clearly, that they can borrow as much as they want. They’re willing to ignore their liabilities until it all blows up in their faces. Just ask the ghost of WorldCom.

The Psychology of Complacency

Credit investors like to think they’re the adults in the room. But each cycle shows bondholders are just as vulnerable to narrative. Once a theme like AI takes hold, AAA and BBB blur, and risk becomes what you convince yourself it isn’t.

The irony: those seeking stability are hit hardest when spreads compress to nothing. One downgrade or policy shift can flip the story overnight, as the tech crash’s fallen angels proved.

Parallels to Today’s Market

There are clear echoes of the tech bubble. In each round, they had convinced themselves of a story: technology had changed the game, and the old risks no longer applied. In both, spreads flattened out, leverage soared and investors talked themselves into believing that growth trumped all. The added wrinkle today is the colossal size of debt markets. Outstanding corporate debt has more than doubled since 2000, and institutional demand for yield is as high as it’s ever been. When spreads crack, the ripple effects will move quicker and further than two decades earlier.

Private credit is another echo. Like telecoms tapping bonds in the ’90s, today’s AI firms lean on private lenders with fewer safeguards. If spreads widen, those protections may prove weak.

The Takeaway

Spreads are not mood rings. They’re supposed to tell us how risky something is. When they fall to the levels where BBB trades like AAA, that’s not because the world got safer, but rather it’s because investors stopped asking questions.” What history teaches us is that this will not last. Dot coms, telecoms and structured credit all demonstrated the same thing: The cycle inevitably turns. A.I. may change the world, but it won’t change the rules of credit. And the message it delivers most often isn’t very kind.