According to Fortune, Moody’s Analytics Chief Economist Mark Zandi is sounding the alarm on LinkedIn, calling borrowing by AI companies a “mounting potential threat to the financial system.” He notes that big tech firms, including Meta, Amazon, Nvidia, and Alphabet, are issuing more debt now than before the dot-com crash, even after adjusting for inflation. The 10 largest AI companies are set to issue over $120 billion in bonds this year alone. Crucially, they’re not just refinancing old debt—they’re taking on massive new debt to fund the AI infrastructure boom. Zandi warns that if these companies fall short of sky-high investor expectations and their stock prices tumble, that mountain of debt could become a serious problem very quickly.
Why This Time Is Different
Here’s the thing: this isn’t a repeat of the 1990s. Back then, internet companies were fueled by stock mania and venture capital. They didn’t have much debt. Now? The AI giants are going straight to the bond market. And the market is treating them like utilities, lending them money for 10 to 40 years at super low rates. As strategist Shay Boloor told Fortune, it’s the “cheapest and cleanest” way to fund a buildout that will last a decade and cost trillions. They have the profits to pay cash, but why would they? Debt is cheap when everyone believes you’re building the future.
The Hidden Risk In All That Hardware
But there’s a massive catch that makes this different from building railroads or cell towers. A lot of this debt is paying for physical stuff—data centers packed with expensive chips. George Calhoun from Stevens Institute of Technology points out the brutal innovation cycle. The hardware in an AI data center today could be obsolete way faster than a fiber optic cable. What if a company borrows for 30 years to build a $10 billion data center, and the tech inside is uncompetitive in just 5? You’re stuck with a stranded asset and decades of payments. That’s a real risk the dot-com era didn’t have.
And let’s talk about the players. Nvidia can point to blowout earnings as “proof in the pudding.” But what about OpenAI? They don’t have massive profits to cushion their bets. If a central player like that stumbles, the ripple effect could be substantial for the ecosystem that depends on it. It’s not just about the tech giants.
The Real Bottleneck Isn’t Cash
So, is the bubble gonna pop? Maybe not from debt alone. Boloor’s biggest worry is actually the U.S. energy grid. Think about it. You can issue all the bonds you want, but if you can’t plug your trillion-dollar AI factory into the wall, what then? The buildout could literally outpace the grid’s capacity. That’s a physical constraint no amount of creative financing from economists like Zandi can fix.
Look, I get it. This is how big infrastructure gets built. But when the financing is this aggressive and the technology is moving this fast, you have to ask: are we building a stable foundation, or just the world’s most expensive house of cards? The market is betting it’s the former. Let’s hope they’re right.
