Big Tech Off-Balance-Sheet Debt: $3 Trillion AI Data Center Risk for Amazon and Co.

Key Takeaways

• Five major technology companies (Alphabet, Microsoft, Amazon, Meta, Oracle) had off-balance-sheet commitments of $1.65 trillion for AI infrastructure as of August 8, 2026, exceeding their combined balance sheet debt of approximately $1.35 trillion.
• Meta reports off-balance-sheet commitments of approximately $420 billion, 2.8 times its official debt, while Oracle reported $273.3 billion as of end of May 2026 — a thirty-fold increase over four years.
• The four largest hyperscalers (Alphabet, Amazon, Meta, Microsoft) spent a combined $381 billion on capital expenditures in 2025; a more than 60 percent increase to $700 billion is expected for 2026.
• More than $120 billion in AI data center spending was moved outside balance sheets through special purpose vehicles and similar structures by December 2025 within less than two years.
• AI revenues of approximately $60 billion in 2025 are significantly below capital expenditures of $400 billion, creating a substantial funding gap.

Off-Balance-Sheet Commitments Reach Record Levels

By mid-August 2026, major U.S. technology companies had accumulated between $2.13 and $3 trillion in off-balance-sheet commitments for building AI data centers. An analysis by Nikkei Asia from August 8, 2026 calculated the combined off-balance-sheet commitments of five companies — Alphabet, Microsoft, Amazon, Meta and Oracle — at $1.65 trillion. This amount exceeds the combined official debt of these companies, which stands at approximately $1.35 trillion.

The $1.65 trillion represents an eightfold increase compared to four years ago. Separate analyses from mid- to end-August 2026 covering nine technology companies arrived at a range of $2.13 to $3 trillion in off-balance-sheet commitments.

Individual companies show different levels of burden: Meta reports off-balance-sheet commitments of approximately $420 billion as of August 2026, which equals 2.8 times its official debt. Oracle reported off-balance-sheet commitments of $273.3 billion as of end of May 2026 — a thirty-fold increase over four years. For Amazon, the company is expected to generate negative free cash flow of up to $28 billion in 2026.

Financing Structures Through Special Purpose Vehicles

Off-balance-sheet commitments arise through several interlocking financing mechanisms. The four largest hyperscalers — Alphabet, Amazon, Meta and Microsoft — spent a combined $381 billion on capital expenditures in 2025, a historic high. For 2026, an increase of more than 60 percent to $700 billion is projected. According to Goldman Sachs estimates, hyperscalers could spend a total of $5.3 trillion on AI and data center infrastructure by 2030.

The commitments include long-term purchase agreements for GPUs and servers, lease agreements for data centers that have not yet become operational or been delivered, as well as complex securitization structures and GPU-backed credit facilities. Added to these are corporate bonds and private credit agreements.

Special purpose vehicles (SPVs) and legally independent subsidiaries play a central role. These structures build and own the data centers, while the parent companies retain exclusive and complete usage rights through contractual arrangements. Financing comes from investors, banks and financial firms, sometimes also from the parent company itself. The debt does not appear as accounting liabilities in consolidated balance sheets because the assets have not yet been delivered or due to legal isolation through the special purpose vehicles.

According to a litigation-risk report published in March 2026, technology companies had moved more than $120 billion in AI data center spending outside their balance sheets through SPVs and similar structures by December 2025 within less than two years.

Accounting and Disclosure Practices

Off-balance-sheet commitments are disclosed in footnotes to financial reports, not as primary balance sheet liabilities. This treatment makes it more difficult for investors to grasp the complete financial picture. An accountant quoted in August 2026 explained that this accounting practice was currently "in vogue," but warned: "What if one of these companies turns out to be a house of cards, kept running only by this accounting?"

Multiple sources draw parallels to Enron Corporation, which before its collapse in 2001 used similar accounting instruments — debt was bundled in off-balance-sheet entities to relieve the parent company's balance sheet. However, an analysis published in the Guardian on August 23, 2026 argues that the situation is not "Enron 2.0" because the SPV structures are legal and disclosed in footnotes. This distinguishes them from Enron's fraudulent concealment. Additionally, major financial institutions and rating agencies monitor the structures, and historical precedents showed that mature companies could successfully manage similar structures.

The same analysis, however, acknowledges that given the scale of capital deployment, careful review of disclosures, footnote analysis and consolidation treatment remains warranted.

Risk Areas and Liquidity Concerns

The risk structure encompasses multiple layers. Should demand for AI computing services not grow as expected, data center utilization rates could decline, resulting in asset impairments and substantial losses. AI revenues of approximately $60 billion in 2025 are significantly below capital expenditures of $400 billion, creating a substantial funding gap.

A litigation-risk report published in March 2026 identified several potential areas of conflict: payment defaults and insolvency cascades across interconnected capital structures, securities fraud lawsuits due to off-balance-sheet opacity, credit rating disputes similar to RMBS disputes after 2008, valuation and maintenance margin conflicts due to rapidly depreciating GPU collateral, as well as construction and power contract disputes related to aggressive timelines. The report pointed to a possible deferred write-down of more than $520 billion over three years, which was noted on August 22, 2026.

The growing reliance on off-balance-sheet financing creates interconnection between banks, private credit and capital markets. Goldman Sachs assumes that private markets will play an increasingly important role in financing the $5.3 trillion expected for AI and data centers by 2030.

Market Context and Regulatory Perspectives

Reporting and analysis focused between July and August 2026. Earlier reference points include a Financial Times report from December 2025 on the $120 billion moved off balance sheets, as well as the litigation-risk analysis published in March 2026 that forecast trends for 2026.

While the legal permissibility of the structures is undisputed, the question of transparency and risk assessment remains open. Disclosure currently occurs through footnotes rather than primary balance sheet items. This complies with current accounting standards, but makes complete capture of liabilities more difficult for investors.

The historical parallel to Enron remains contested: while the accounting instruments are structurally similar, the degree of disclosure and legal framework differ. Whether the current structures carry systemic risks or merely represent a legitimate financing strategy will be answered in coming quarters by the actual development of AI demand and the utilization of the financed data centers.