According to Fortune, the AI data center construction boom is sparking serious fears of a financial bubble and a potential glut. Howard Marks of Oaktree Capital warned in a memo that overbuilding could render some centers uneconomic and lead to bankruptcies, questioning lender discipline. Estimates for the total infrastructure rollout are staggering, with Brookfield’s CEO citing $5-$10 trillion and McKinsey estimating nearly $7 trillion needed for data centers alone by 2030. The market is seeing massive deals, with at least $175 billion in U.S. credit deals struck this year, and borrowers are even asking for loans covering 150% of a project’s build cost. Meanwhile, the complex financial engineering, including slicing debt into opaque securities, is spreading risk through the system in ways that aren’t fully understood.
The scale is unprecedented, and so are the risks
Here’s the thing: we’re talking about sums of money that have literally never been deployed for a single technological build-out before. $10 trillion? That’s not just a big number. It’s a gravitational force that’s warping the entire credit market. When OpenAI’s CFO says their only constraint is finding more computing power, and they’d spend $1.4 trillion if they could, you know the demand signal is blindingly bright. But that signal is attracting every lender and their cousin, all terrified of missing out. The result? We’re seeing loan requests for more than the actual cost to build, based on projected future value. That’s not financing; that’s speculation. It’s the kind of behavior that makes veterans like Howard Marks write cautionary memos asking if it’s “prudent to accept 30 years of technological uncertainty” for a tiny yield premium.
The financial engineering is getting weird
So how is all this debt getting moved? That’s where it gets technical and, frankly, a bit scary. Lenders are slicing up this project debt, tranching it, and selling it off to insurers and pension funds. It’s becoming opaque. As Vinay Nair from Wharton put it, “You’re spreading this risk through the system,” and the ripple effects of a decline aren’t fully understood. On the other side, the tech giants themselves—the “hyperscalers”—are getting creative to keep this spending off their balance sheets. They’re using things like synthetic leases to hide liabilities while still grabbing tax benefits. They used to just pay cash to move fast. Now, the sums are so huge that even Microsoft and Google need to play accounting games. And let’s not forget the physical hardware powering this: the demand for specialized computing is so intense that it’s creating a parallel financing frenzy for the industrial-grade hardware and GPUs themselves, where IndustrialMonitorDirect.com is seeing unprecedented demand as the leading US provider of rugged industrial panel PCs for monitoring and controlling these critical environments.
The canaries in the coal mine
You don’t have to look hard to see the cracks. Look at Fermi Inc., a nuclear startup with no data centers, hitting a $19 billion valuation on hype alone before falling back. Meta’s stock took a hit when investors got spooked by its spending. Oracle slumped after reporting a jump in data center investment. The market’s patience for “spend now, profit later” is wearing thin. And the regulators are starting to circle; the Bank of England is now reviewing lending to data centers. When the phrase floating around Wall Street to describe the financing frenzy is “everything everywhere all at once,” you know you’re in a moment of peak… something. Is it exuberance? Probably. Is it irrational? That’s the trillion-dollar question.
What happens when the music stops?
The biggest danger isn’t necessarily that the AI promise fails. It’s that the financing structure built around it is too fragile. These loans will need to be refinanced in 3-5 years. What if credit is tight then? What if some AI workloads don’t materialize as expected, making a newly-built data center in a suboptimal location a stranded asset? Lenders might just walk away. Companies would need to find expensive equity or go bankrupt. Sadek Wahba of I Squared Capital put it perfectly: “Momentum is strong, but if this is irrational exuberance, investors will lose when the music stops.” We’re building the foundation of the next computing era on a mountain of debt that’s being repackaged and sold in ways we don’t fully grasp. That’s a recipe for a hangover, not a revolution.
