According to Business Insider, Mark Zandi, the chief economist at Moody’s Analytics, is sounding a major alarm on the financial risks of the AI boom. He warns that AI “hyperscalers” are taking on record levels of debt through bond issuance in 2025, a level that already eclipses the borrowing done by tech companies in the years leading up to the dot-com bubble burst. Zandi emphasized on X that this isn’t just refinancing old debt, but new borrowing to fund the fight for AI market share. He stated that, even after adjusting for inflation, tech-sector borrowing is at all-time highs. His core argument is that this soaring debt should be seen as a “mounting potential threat to the financial system and broader economy,” especially if the AI payoff doesn’t meet sky-high expectations.
The Dot-Com Parallel, But With More Debt
Here’s the thing: the dot-com comparison is everywhere, but Zandi is adding a crucial, scarier layer. Back in the late 90s, the frenzy was largely equity-driven. Crazy valuations, yes, but funded by stock market optimism. The crash mostly burned investors holding those stocks. Now? The AI infrastructure arms race—those massive data centers and chip purchases—is being fueled by enormous debt. Companies are issuing bonds to pay for it all. So if the AI revenue doesn’t materialize as hoped, it’s not just stock prices that tumble. It’s a wave of potential corporate defaults. That’s a credit market problem, and as 2008 taught us, those have a nasty habit of spreading far beyond a single sector. The systemic risk is simply higher when debt is the accelerant.
The Infrastructure Gold Rush
Zandi nails the core issue: “massive (over) investments” in AI infrastructure. Everyone from cloud giants to chipmakers is in a capex war, building data centers at a breakneck pace. But what’s the actual demand for all this AI compute? It’s the big, multi-billion-dollar question. The optimism is priced into stocks like Nvidia, but the debt is financing the physical build-out. It’s a classic bubble signal—huge capital expenditures based on projected future demand that may or may not pan out. And let’s be real, the economics of generative AI for many businesses are still shaky. If the use cases don’t become profitable fast enough, who’s left holding the bag for all those data centers? The bondholders. And that’s where the financial system gets a stomach ache.
Why This Warning Actually Matters
Look, we hear bearish takes all the time. But Zandi isn’t some fringe commentator; he’s the chief economist at a major ratings agency that lives and breathes credit risk. When he says this should be on the “radar screen” for systemic risk, regulators and big banks probably perk up. His point about the borrowing “dwarfing” the Y2K era is stark. It cuts through the hype and asks a very simple, uncomfortable question: What if we’re building way too much, way too fast, with borrowed money? The potential domino effect—AI stock crash, credit stress, loss of confidence in tech—isn’t hard to imagine. It’s a reminder that technological revolutions and financial stability don’t always go hand-in-hand. Sometimes, they collide.
