Look at what Alphabet, Microsoft, Amazon, Meta, and Oracle officially report as debt, and the numbers appear large but manageable. Yet a deeper investigation by Nikkei has uncovered a second, far larger pile of obligations lurking off the books. That hidden figure now stands at an estimated $1.65 trillion—roughly eight times what the five companies collectively reported just four years ago. The sum exceeds the $1.35 trillion in debt these firms openly acknowledge on their balance sheets.
The financing mechanism behind this staggering imbalance is both familiar and controversial. The companies are using off-balance-sheet vehicles, known as special purpose entities (SPEs) or special purpose vehicles (SPVs), to keep massive borrowing out of their reported financial statements. This is the same structural trick that allowed Enron to hide billions in debt before its spectacular collapse a quarter-century ago. Back then, the practice was fraudulent because the debt was intentionally concealed from investors and regulators. Today, thanks to tightened accounting rules and fuller disclosure requirements, the same structures are legal—provided the obligations are disclosed in footnotes. The difference is transparency, not architecture.
As analyst Gil Luria told Bloomberg Law, Enron's crime wasn't having special purpose vehicles. Enron's crime was hiding them. The modern tech giants are not hiding their off-balance-sheet exposure completely; they tuck the details into dense pages of annual reports, often buried under technical jargon. Yet the sheer scale of the hidden debt raises questions about whether investors are getting a true picture of leverage in the AI arms race.
How the off‑balance‑sheet engine works
The mechanics are straightforward. A company packages its capital expenditures for AI infrastructure—chips, servers, cooling systems, and data centre construction—into a separate legal entity. This entity is typically a joint venture with a financial partner, such as an investment firm or a real estate trust. The venture borrows money directly, and that debt never appears on the tech company's own balance sheet. The company signs a long-term lease or purchase agreement with the venture, arguing that it is not the primary obligor and therefore need not record the debt as its own.
Take Meta's Hyperion data centre in Louisiana as a concrete example. Meta partnered with Blue Owl Capital to create a separate structure that took on $27 billion in debt to finance the facility. Meta is the sole tenant and has committed to using the data centre for years. Yet Meta argues—and its auditors agree—that because it is not the entity responsible for finding replacement tenants if it vacates, the debt does not need to appear on Meta's books. This reasoning is consistent with current accounting standards, but critics say it paints an incomplete picture of risk.
Oracle provides another stark illustration. The company has disclosed roughly $260 billion in future lease commitments that will eventually land on its balance sheet once the data centres become operational. Nvidia, though not one of the five companies in the initial study, carries $119 billion in purchase obligations, most of which are related to AI chip supply agreements. Alphabet, Microsoft, and Amazon all employ similar structures, collectively keeping trillions of dollars in potential liabilities out of their headline debt figures.
The numbers behind the AI boom
The scale of this hidden leverage is the real story. Meta's off-balance-sheet debt alone is estimated at about $420 billion—nearly three times its reported total debt. Oracle's off-balance-sheet obligations have grown roughly thirtyfold in just four years, from under $10 billion to more than $300 billion. For all five companies combined, the hidden debt has increased at a compound annual growth rate of over 70% since 2020.
This explosion is driven by an unprecedented spending spree. The industry is expected to invest more than $3 trillion between 2024 and 2028 to build and equip AI data centres. Much of that spending is financed against the very chips and servers housed inside those facilities. The financing is structured to keep the debt off the sponsors' books, allowing them to preserve their credit ratings and reported leverage ratios. In effect, these companies are using the assets themselves as collateral for borrowing that never appears on their balance sheets.
Why it matters now
The timing of this revelation is particularly awkward. Four of the five companies—Alphabet, Microsoft, Amazon, and Meta—are scheduled to report quarterly earnings in the next two weeks. Their reported debt figures will appear tidy and well under control. The $1.65 trillion sitting in the footnotes will not make the headlines. Investors reading the earnings releases will see less than half of the total leverage.
The catch is what happens when these off-balance-sheet obligations eventually come due. When a data centre becomes operational, its lease typically rolls onto the company's balance sheet in full. If AI demand falls short of the rosy forecasts, the facility may be marked down in value. The resulting losses would land not on the tech giants themselves—at least initially—but on the lenders and insurers who funded the construction. Banks, pension funds, and insurance companies have been eager to provide capital for AI infrastructure, drawn by the high yields and the apparent creditworthiness of tenants like Meta and Oracle. Yet if a downturn occurs, these financial institutions could face significant write-downs, potentially triggering a cascade of credit tightening.
Some market observers are already sounding alarms. S&P Global Ratings cut Oracle's credit rating in early 2025, citing stretched leverage from its data centre commitments. Both Morgan Stanley and Moody's have issued reports flagging the broader risk. Accounting consultant Tom Selling posed the uncomfortable question: What if one of these companies was a house of cards and was propping itself up with this accounting treatment? The concern is that the off-balance-sheet structures, while legal, obscure the true financial health of some of the world's largest technology firms.
Historical context: echoes of Enron and beyond
The use of special purpose vehicles is not new. Enron famously employed thousands of SPEs to hide losses and inflate earnings, leading to the largest bankruptcy in American history at that time. After the scandal, the Financial Accounting Standards Board (FASB) introduced new rules—most notably FIN 46 (now codified in ASC 810)—requiring companies to consolidate SPEs if they bear the majority of risks and rewards. However, the rules still allow for exceptions. If a company is not the primary beneficiary and does not have control, the SPE can remain off the books, provided its obligations are disclosed.
Tech companies have taken full advantage of these exceptions. Their joint venture structures are carefully designed so that no single party controls the entity. The financial partner shares risk and often takes a first-loss position, making the tech firm technically not the primary beneficiary. This is why auditors sign off on the treatment. But as the scale grows, the line between legitimate off-balance-sheet financing and economic substance begins to blur.
The honest version: legally disclosed, but effectively hidden
None of this is illegal, and the companies insist that future earnings will more than cover the bills. The disclosures do exist—buried in the footnotes of annual reports, often under headings like Lease Commitments or Unconditional Purchase Obligations. For the analyst willing to dig, the information is there. Yet the average investor, focused on earnings headlines and reported debt ratios, is seeing only half the picture.
In a year already characterized by loud talk of an AI bubble, the quiet number sitting off the page may be the one that matters most. The $1.65 trillion in hidden debt represents a systemic risk that is growing faster than the industry's ability to generate cash flows. If the AI demand surge continues, the off-balance-sheet structures will pay off handsomely, and the lenders will be repaid. If it stalls, the losses could reverberate through the financial system in ways that regulators are only beginning to assess. For now, the five tech giants are betting trillions on a future that is still being built—and financed in the shadows of their own balance sheets.
Source: TNW | Amazon News