8 hours ago
13
Some experts are warning of a looming “debt bomb” crisis because big datacenter builders such as Meta, Oracle, xAI and CoreWeave are not only raising billions to construct these facilities but are also not recognizing these long-term debt obligations on their balance sheets.
Should we be concerned? No. I know because I’ve seen this movie before.
It works like this: Meta wants to build a datacenter to accommodate its growing AI and cloud computing needs. It forms a separate entity that’s not consolidated in its financials, which builds the datacenter. To fund this, the entity raises money from investors, banks and other financial firms (including some from Meta) that own the majority of the entity. There is a contract by which Meta has exclusive and full use of the datacenter once built. This way Meta gets its datacenter but most of the debt incurred to build it is not shown as a liability on its books.
The worry comes from the enormous sums of money being plowed into these entities. The Financial Times reported in December 2025 that tech companies had shifted more than $120bn of AI datacenter spending off their balance sheets through special-purpose vehicles and similar financing structures. Goldman Sachs estimates that hyperscalers could spend $5.3tn on AI and datacenters through 2030 and expects private markets to play an increasingly important role in financing that buildout.
Skeptics are worried that these tech guys are pulling the wool over the public’s eyes and not properly disclosing the long-term impact of all this debt. They point to Enron, the energy firm that spectacularly failed in 2001, causing tens of billions of dollars of losses to its shareholders and contributing to a historic market meltdown.
Fine. Scrutinize them. Read the footnotes. Argue about consolidation. But don’t call them Enron. Enron caused the Enron crash. It’s highly unlikely that fraud at that level is being perpetuated now by these companies.
When I first started out in accounting back in the mid-1980s, my biggest client was a publicly held biotechnology firm called Centocor, which, at the time, was developing drugs using a technology based on monoclonal-antibody treatment for various purposes including sepsis. To raise money, Centocor also did off-balance-sheet financing. The company formed limited partnerships in which it held a minority interest but the entity issued debt and took in investments from limited partners in order to fund the development of various drugs, which Centocor then had the exclusive right to use.
In the 1980s and early 1990s, hundreds of millions of dollars were plowed into these partnerships. This was common practice in the biotechnology during that time.
Did some of the drugs fail in clinical testing? Yes. But did this cause a stock market panic? No. Like with today’s datacenters, the risks were spread. During that time – before the Internet, SEC disclosures and 24/7 media scrutiny – the public little knew of these vehicles. The good news is that the accounting has evolved, although the economic idea has stayed the same – no debt was shown on Centocor’s balance sheet. Ultimately, this financing vehicle became more expensive and less popular because it was easier and more affordable to raise money other ways.
The off-balance-sheet financing strategies today that are being used by big tech firms have different risks than what the biotech companies were doing when I was younger. Today, the disclosures required by companies doing this are significant. The scrutiny is intense. The investing public is smarter. Today’s risks are different and arguably more recoverable.
In the 90s Centocor, Genentech, Amgen, Biogen and other companies like them were raising millions in partnerships that were developing products with a high probability of failure during clinical testing.
Today’s investors are financing land, buildings, electrical infrastructure and computing equipment. A datacenter can disappoint financially, but it doesn’t disappear because a clinical trial fails. It’s why Jeff Bezos calls AI an “industrial bubble”. Industrial bubbles leave things behind. Railways. Fiber-optic cable. Factories. And this time, datacenters.
If this is a glut, it’s a strange one: developers increased North American capacity by 36% last year and vacancy still fell to a record 1.4%. According to CBRE’s North America Data Center Trends H2 2025 report, demand is outpacing supply in nearly every major market. Unlike the drug companies hoping for success, there’s a legitimate market need today for datacenters. AI isn’t going away. Microsoft estimates that only 17.8% of the world’s working-age population currently uses generative AI. If that’s anywhere near correct, we’re still much closer to the beginning of adoption than the end.
Some of these investments will fail. Some lenders will lose money. Some datacenters will be worth less than their owners paid for them. But that’s exactly why these financing structures exist: to spread enormous capital requirements and risk among investors willing to take it. The obligations are disclosed, the assets are real and demand for computing capacity remains strong. That’s why I’m not losing sleep over big tech’s off-balance-sheet debt. I see financial engineering, yes. I don’t see a debt bomb.
