Nikkei: Big Tech Hides $1.65 Trillion in AI Investment Debt

Five American tech giants are financing the AI boom in a way that largely bypasses their official balance sheets. An analysis by the Japanese newspaper Nikkei found that Alphabet, Amazon, Meta, Microsoft and Oracle have amassed a combined $1.65 trillion in liabilities tied to AI data centers that do not appear as debt in standard financial statements.

How the mechanism works

The core of the issue lies in a legal structure used by all five companies. Instead of taking out a loan itself to build another data center, a company sets up a separate legal entity, a special purpose vehicle, with outside investors participating. That entity takes on the debt needed to build the facility, while the sole tenant of the facility is Google, Microsoft, or another giant.

Formally, the tech company itself has not borrowed anything, so it doesn't have to show that liability as debt on its own balance sheet. In practice, however, multi-year lease agreements and guarantees to purchase computing capacity mean the financial risk remains real, just shifted out of view of standard debt metrics.

Echoes of Enron

The comparison to Enron, the energy giant that collapsed in 2001 after its accounting manipulations came to light, is not accidental in analysts' commentary. Enron also used special purpose vehicles to shift liabilities off its main balance sheet and inflate the picture of its financial health for investors.

Experts point out an important legal difference between today's situation and the scandal of twenty-five years ago. The structures used by Alphabet, Amazon, Meta, Microsoft and Oracle comply with US GAAP accounting standards as well as international IFRS rules, and the liabilities are formally disclosed, albeit in footnotes and notes that almost no one reads in full.

Enron's crime wasn't having special purpose entities. Enron's crime was hiding them. - Gil Luria, analyst

The accounting approach itself is in vogue right now. But what if one of these companies turns out to be a house of cards, propped up precisely by these kinds of accounting arrangements? - Tom Selling, accounting consultant

Company by company

The starkest example is Meta, whose off-balance-sheet liabilities reach around $420 billion, compared with just $140 billion in officially reported debt. One concrete project is the Hyperion data center in Louisiana, where $27 billion in debt was placed in a separate legal entity outside Meta's balance sheet.

Oracle, meanwhile, is set to report $260 billion in lease obligations in the future, and its off-balance-sheet debt has grown roughly thirtyfold over four years. Nvidia, while not one of the five main companies covered by the analysis, looms in the background of the whole picture with $119 billion in purchase commitments to suppliers and customers in the AI sector.

What it means for investors

The scale of the phenomenon raises questions about how reliably the market can assess the risk tied to the enormous spending on AI infrastructure. If demand for AI computing power turns out to be lower than assumed in multi-year lease contracts, the costs will eventually fall on the same tech companies anyway, only investors will find out about it later.

Analysts stress that the problem isn't illegality but transparency. The growing gap between the market value of AI companies and their actual profits had already raised concerns about the sector overheating, and the disclosure of the scale of off-balance-sheet debt adds a concrete, quantifiable financial dimension to that.

For Polish investors and pension funds, who have indirect exposure to Big Tech stocks through stock indices and passive funds, the conclusions of the Nikkei report matter regardless of geography. The valuation of these companies largely rests on the assumption that AI investments will pay off, and the hidden debt means that the risk in the event of a demand collapse would be greater than official balance sheets suggest.

White papers and analyst reports have for months warned about the gap between Big Tech's AI spending and the cash flows it generates. Nikkei's findings don't change that narrative, but they supply it with hard numbers and a concrete historical precedent that market commentators are now invoking.