The race to dominate artificial intelligence is generating a financial reality that does not appear on the balance sheets of America's five largest technology companies. A study by Nikkei Asia has found that Alphabet, Microsoft, Amazon, Meta, and Oracle have accumulated an estimated $1.65 trillion in off-balance-sheet AI-related obligations — a figure that has grown roughly eightfold in four years and now exceeds the $1.35 trillion in debt these companies formally report.
The obligations are not secret. They are disclosed in the footnotes of quarterly financial statements, following accepted accounting rules. But they sit outside the headline debt figures that most investors and analysts focus on, making it harder to assess the true financial exposure of some of the world's most valuable companies.
What is hiding in the footnotes
The commitments take several forms. Long-term contracts for Nvidia graphics processing units (GPUs, the chips that power AI systems), future server deliveries, and data centre leases that have not yet commenced typically remain off the balance sheet until accounting rules require them to be recognised as formal liabilities. Under both US Generally Accepted Accounting Principles (GAAP) and International Financial Reporting Standards (IFRS), the structures are entirely compliant.
The numbers at individual companies are striking. Meta carries approximately $420 billion in off-balance-sheet obligations against only $140 billion of recorded debt, according to Nikkei Asia. Oracle's hidden liabilities have reached $273 billion, a rise of roughly 2,900 percent since 2022. S&P Global has already responded, downgrading Oracle's credit rating citing its stretched leverage from AI commitments, including around $260 billion in future lease obligations.
“"What if one of these companies was a house of cards, and was propping itself up with this accounting treatment?" — Tom Selling, accounting consultant”
The cloud businesses of Microsoft, Alphabet, and Amazon add another layer. Their combined remaining performance obligations — services already contracted by customers but not yet delivered — reached $1.45 trillion at the end of March, Nikkei reports. These represent revenue that is committed on paper but has not yet flowed through to earnings.
Shadow borrowing and the BIS warning
Economists at the Bank for International Settlements, the Basel-based institution that acts as a central bank for the world's central banks, described the broader financing approach as 'shadow borrowing' in a March report. The term refers to capital raised through institutional investors via joint ventures and infrastructure partnerships that avoid immediate recognition as debt. The BIS warned that any slowdown in AI demand could expose vulnerabilities across the entire data centre investment cycle.
Moody's, in a separate assessment, put off-balance-sheet commitments at $1.2 trillion across the sector, with more than $820 billion attributed to data centres still under construction. The credit ratings agency described these as debt-equivalent liabilities that will leave companies committed to significant future rent payments. Moody's also noted, however, that the five hyperscalers still hold some of the most robust balance sheets in the corporate world and that their investment-grade ratings are not facing imminent risk.
“"Previously, these companies relied on asset-light structures centered on software, intellectual property, and scalable cloud services that required modest capital investment." — Moody's”
The Enron comparison — and why it matters but has limits
Comparisons to Enron, the US energy trading company whose 2001 collapse remains one of the largest corporate bankruptcies in history, have surfaced repeatedly in coverage of this issue. Enron used off-balance-sheet partnerships to hide hundreds of millions of dollars in debt from investors and lenders. The mechanics of that concealment and the structures Big Tech is using today share a family resemblance, but the legal and regulatory context is fundamentally different.
The distinction was put plainly by analyst Gil Luria, whose comments were reported by Bloomberg Law: "Enron's crime wasn't having special purpose vehicles. Enron's crime was hiding them." Today's structures comply with accounting standards and are disclosed, at least in principle. The concern is not fraud but visibility: the disclosures exist in fine print that most investors do not read closely, and that gap between technical disclosure and practical transparency is where the risk concentrates.
The moment of reckoning approaches
The financial pressure is already showing in bond markets. In 2025, the five companies issued approximately $121 billion in new bonds, up from roughly $40 billion in 2020. AI-related bond issuance globally reached nearly $236 billion by the end of May 2026, running at roughly four times the pace of the previous year. When AI spending exceeds operating earnings, companies must turn to external financing to cover the gap.
The moment of reckoning for much of this hidden debt will come as physical infrastructure opens. When a data centre goes live, its lease rolls onto the balance sheet in full. If AI demand growth fails to match the pace of capacity being built, utilisation rates fall, asset values may decline, and the losses land on whoever holds the financing. Not showing interest payments on credit funding AI infrastructure also boosts profit metrics such as EBITDA (earnings before interest, taxes, depreciation, and amortisation), a figure widely used by analysts to value technology firms.
“"$1.65 trillion in extremely opaque circular finance is too low. Double or triple that in total exposures, and that may be closer." — Peter Feierstein, investment strategist, speaking to Newsmax”
Amazon Web Services chief executive Matt Garman pushed back on the characterisation of risk, telling Nikkei that the company's investments are 'not speculative'. All five companies declined to comment further to Nikkei on their rising off-balance-sheet commitments. For international investors holding these stocks or bonds, the message from analysts, auditors, and rating agencies is consistent: the headline debt figures alone are no longer sufficient to understand the full picture.
This article is free to read. It always will be — no paywall, no account, no tracking.
