Five of America’s largest technology companies — Alphabet, Microsoft, Amazon, Meta, and Oracle — have accumulated approximately $1.65 trillion in off-balance-sheet debt over the past four years, according to a Nikkei Asia study. This hidden borrowing, which the Bank for International Settlements (BIS) calls “shadow borrowing,” now exceeds the companies’ combined on-balance-sheet debt of roughly $1.35 trillion, meaning investors analyzing standard debt-to-equity ratios are missing more than half of the actual liabilities tied to the AI infrastructure buildout.

The scale is staggering. In just four years, off-balance-sheet obligations at these five hyperscalers grew roughly eightfold, driven almost entirely by the race to build AI data centers, secure GPU compute capacity, and finance the physical infrastructure required to power large language models and generative AI services. What makes this trend particularly concerning is that these liabilities are structured in ways that keep them off traditional balance sheets — through long-term leases, joint ventures with private credit funds, project-finance vehicles, and equity derivatives that commit companies to future payments without recording them as debt today.

What Is Shadow Borrowing? BIS and the New Language of AI Finance

The Bank for International Settlements, often called the central bank for central banks, has coined the term “shadow borrowing” to describe this phenomenon. Shadow borrowing refers to capital raised through structures that create binding financial obligations without appearing as formal debt on corporate balance sheets. For the five US tech giants, these structures include:

• Long-term data center leases structured as operating leases rather than capital leases
• GPU capacity commitments that lock in years of future payments
• Joint ventures with private credit funds where the tech company guarantees returns or commits to purchasing capacity
• Project-finance vehicles that build infrastructure on behalf of the company without consolidating the debt
• Equity derivatives that function as financing arrangements but are classified as derivatives

The BIS warning is significant because it signals that global financial regulators are paying attention. When the institution responsible for monitoring systemic financial risk flags a $1.65 trillion shadow borrowing phenomenon at five of the world’s most valuable companies, it suggests that the scale of off-balance-sheet AI financing has reached a level that could pose broader financial stability risks.

Traditional credit analysis relies heavily on debt-to-equity ratios, interest coverage ratios, and leverage metrics derived from balance sheet data. Shadow borrowing renders these metrics incomplete. An investor looking at a company’s reported debt figures might conclude leverage is manageable, when in reality the company has committed to hundreds of billions in future payments that function economically like debt.

Meta’s $420 Billion Off-Balance-Sheet Bet — Inside the Blue Owl JV and Hyperion Campus

Meta Platforms leads the five companies in off-balance-sheet debt with an estimated $420 billion in hidden liabilities — nearly three times its transparent on-balance-sheet debt. The centerpiece of Meta’s shadow borrowing strategy is a $27 billion private-credit joint venture with Blue Owl Capital to finance the Hyperion data center campus in Louisiana.

The Hyperion campus is one of the largest AI data center projects in the world, designed to support Meta’s ambitious AI and machine learning workloads, including its Llama large language model development and AI-powered recommendation systems. By structuring the financing through a private-credit JV rather than direct corporate borrowing, Meta keeps the associated debt off its balance sheet while still securing the capital needed to build.

This structure is emblematic of a broader trend. Private credit funds — which have grown from roughly $500 billion in assets under management a decade ago to over $1.7 trillion today — have become the primary financiers of AI infrastructure. These funds provide capital that banks, constrained by post-2008 regulations, cannot easily offer. In exchange, they receive steady returns backed by long-term contracts with some of the world’s most creditworthy technology companies.

The risk, however, is that these off-balance-sheet structures create obligations that are every bit as binding as formal debt. If Meta’s AI revenue fails to materialize at the expected scale, the company is still on the hook for its commitments to the Blue Owl JV. Moody’s has specifically flagged this “pre-operational lease risk” — the danger that companies commit to paying for AI infrastructure capacity before that capacity generates any revenue.

Oracle’s 30x Debt Surge — From Project Finance to S&P Downgrade

Oracle presents the most dramatic case of hidden debt growth among the five companies. Its off-balance-sheet obligations have surged to approximately $273 billion as of May 2026 — more than 30 times what they were just four years ago. This explosion in shadow borrowing is directly tied to Oracle’s aggressive push into AI cloud infrastructure, including major data center projects in Texas and Wisconsin financed through project-finance structures.

The consequences of this hidden leverage are already visible. Oracle’s total debt burden — including off-balance-sheet obligations — now stands at roughly 2.5 times its annual sales. This elevated leverage ratio prompted S&P Global Ratings to downgrade Oracle’s credit rating to the lowest tier of investment grade, a move that signals heightened risk for bondholders and could increase Oracle’s future borrowing costs.

The Oracle case study is instructive for understanding how hidden debt can become visible. When rating agencies and analysts begin to factor off-balance-sheet obligations into their assessments, the true leverage picture emerges. Oracle’s reported debt figures looked manageable, but once the $273 billion in shadow borrowing was accounted for, the company’s credit profile deteriorated significantly.

Note: Amazon and Microsoft figures are not fully broken out in the Nikkei study but are included in the aggregate $1.65T total.

Alphabet’s Quiet Shift — $40.7 Billion in Off-Balance-Sheet Commitments

Alphabet, Google’s parent company, has disclosed $40.7 billion in future funding commitments to off-balance-sheet vehicles — a figure that includes approximately $30 billion in equity derivatives structured as AI infrastructure financing and a new joint venture with Blackstone, the world’s largest alternative asset manager.

While Alphabet’s hidden debt is smaller relative to its visible debt compared to Meta and Oracle, the trend is still concerning. Alphabet has long been considered the most conservatively financed of the five hyperscalers, with a strong balance sheet and significant cash reserves. The fact that even Alphabet is moving toward shadow borrowing structures suggests that off-balance-sheet AI financing has become the industry standard rather than an exception.

The Alphabet-Blackstone JV is particularly noteworthy because it brings together a technology company and a private equity giant in a structure that blurs the lines between corporate financing and asset management. Blackstone provides capital; Alphabet provides the technical expertise and long-term demand for AI compute capacity. The debt incurred to build the infrastructure sits in the JV, not on Alphabet’s balance sheet.

Amazon and Microsoft — The Rest of the Hyperscaler Picture

Amazon and Microsoft, the two largest cloud infrastructure providers, are also significant participants in off-balance-sheet AI financing, though the Nikkei study does not provide company-specific breakdowns for their hidden debt. Both companies have been aggressively expanding their AI data center footprints through operating leases, GPU capacity commitments, and infrastructure joint ventures.

Amazon Web Services (AWS) has announced plans for massive data center expansions across Virginia, Ohio, and international locations, many financed through structures that keep debt off the balance sheet. Microsoft has similarly committed tens of billions to AI infrastructure, including its partnership with OpenAI, which itself involves complex financing arrangements.

The aggregate $1.65 trillion figure includes Amazon and Microsoft’s contributions, but the lack of transparency around individual company figures is itself a concern. Investors in Amazon and Microsoft cannot fully assess the scale of off-balance-sheet obligations at these companies, making it difficult to compare risk profiles across the five hyperscalers.

Why This Matters for Investors — Debt Ratios That Don’t Tell the Full Story

For equity investors, the hidden debt phenomenon creates a fundamental information asymmetry. Standard financial metrics that investors rely on — debt-to-equity ratio, net debt to EBITDA, interest coverage ratio — are calculated using balance sheet data that excludes off-balance-sheet obligations. When $1.65 trillion in liabilities is invisible to these metrics, the risk assessment is incomplete by definition.

Consider a hypothetical investor comparing two technology companies. Company A has a reported debt-to-equity ratio of 0.5, suggesting conservative leverage. Company B has a ratio of 1.5, suggesting higher risk. But if Company A has $420 billion in off-balance-sheet debt while Company B has none, the apparent risk differential is reversed. This is precisely the situation facing investors in the five US tech giants today.

Credit rating agencies are beginning to adjust their methodologies. S&P’s downgrade of Oracle reflects a willingness to look beyond reported debt figures. Moody’s warning about pre-operational lease risk signals that rating agencies are increasingly focused on the gap between reported and actual leverage. However, the pace of adjustment may be slow relative to the speed at which hidden debt is accumulating.

Moody’s Warning — Pre-Operational Leases and the Revenue Gap

Moody’s has specifically flagged pre-operational lease risk as a growing concern in the AI infrastructure space. The issue is straightforward: hyperscalers are signing long-term leases and capacity commitments for data centers and GPUs that will take months or years to become operational. During this pre-operational period, the companies are paying for capacity that generates no revenue.

This creates a cash flow mismatch that traditional credit analysis may not fully capture. A company might appear to have healthy cash flow from operations, but a significant portion of that cash is already committed to pre-operational leases that will only begin generating returns in future periods. If AI revenue growth slows or fails to meet expectations, the gap between lease commitments and revenue could widen rapidly.

The Moody’s warning is particularly relevant for Meta and Oracle, where hidden debt is largest relative to visible debt. Both companies have made enormous bets on AI infrastructure that assume continued strong demand for AI compute capacity. If demand softens — due to economic downturn, technological shifts, or competitive pressure — the pre-operational lease commitments become a significant financial burden.

The Private Credit Connection — How Wall Street Is Financing the AI Buildout

The rise of private credit has been one of the most significant developments in finance over the past decade, and AI infrastructure financing is accelerating this trend. Private credit funds — including Blue Owl, Blackstone, Apollo, and KKR — have become the primary source of capital for AI data center construction, filling a gap left by traditional bank lending.

The appeal for tech companies is clear. Private credit offers flexible terms, faster execution, and structures that can be designed to keep debt off the balance sheet. For private credit funds, the appeal is equally clear: long-term contracts with investment-grade technology companies offering stable, predictable returns in an environment where traditional fixed-income yields remain relatively low.

However, the rapid growth of private credit AI financing also raises systemic concerns. The private credit market is less regulated than traditional banking, with less transparency around risk concentration and leverage. If multiple AI infrastructure projects financed through private credit encounter difficulties simultaneously, the interconnected risks could amplify through the financial system.

What Comes Next — Regulatory Scrutiny, Rating Agency Actions, and Market Risks

The $1.65 trillion hidden debt phenomenon is unlikely to remain invisible indefinitely. Several forces are converging that could bring these liabilities into the spotlight:

Regulatory scrutiny. The BIS warning about shadow borrowing is a signal that global financial regulators are monitoring the situation. The Financial Stability Board (FSB) and national regulators could push for enhanced disclosure requirements around off-balance-sheet AI financing commitments. The U.S. Securities and Exchange Commission (SEC) has already shown increased interest in climate and ESG disclosures; AI financing disclosures could be next.

Rating agency actions. S&P’s downgrade of Oracle may be the first of several rating actions as agencies incorporate off-balance-sheet obligations into their credit assessments. If Moody’s and Fitch follow suit, the cost of borrowing for these companies could increase, potentially creating a feedback loop where higher borrowing costs lead to more off-balance-sheet financing.

Market correction risk. The most significant risk is that a market event — an economic downturn, a technology shift that reduces AI demand, or a specific company disappointment — triggers a reassessment of hidden debt. In such a scenario, the gap between market valuations and underlying leverage could close rapidly, with significant implications for equity and bond investors.

Investor activism. Institutional investors and activist shareholders may begin demanding greater transparency around off-balance-sheet AI financing. If companies cannot or will not provide clear disclosure, investors may adjust their risk premiums or reduce exposure.

For now, the five US tech giants continue to build AI infrastructure at an unprecedented pace, financed through structures that keep trillions in debt out of sight. The question is not whether this hidden debt will become visible — it is whether it will become visible through gradual regulatory and rating agency action, or through a sudden market event that forces the issue.

Frequently Asked Questions

What is hidden debt at US tech giants? Hidden debt refers to off-balance-sheet liabilities — long-term leases, joint venture commitments, project-finance obligations, and equity derivatives — that create binding financial obligations without appearing as formal debt on corporate balance sheets. The five largest US tech companies have accumulated approximately $1.65 trillion in such hidden debt.

How does hidden debt compare to visible debt at these companies? The $1.65 trillion in hidden debt now exceeds the combined on-balance-sheet debt of approximately $1.35 trillion at Alphabet, Microsoft, Amazon, Meta, and Oracle. This means standard debt-to-equity ratios capture less than half of the companies’ actual leverage.

Which company has the most hidden debt? Meta leads with approximately $420 billion in off-balance-sheet debt, nearly three times its visible debt. Oracle follows with about $273 billion, a 30-fold increase over four years. Alphabet has disclosed $40.7 billion in off-balance-sheet commitments.

What is shadow borrowing according to the BIS? The Bank for International Settlements uses “shadow borrowing” to describe capital raised through structures that create binding financial obligations without appearing as formal debt. The BIS has flagged this as a potential financial stability risk given the $1.65 trillion scale at the five US tech giants.

Why should investors care about off-balance-sheet AI financing? Standard financial metrics like debt-to-equity ratios and interest coverage ratios miss off-balance-sheet obligations, giving investors an incomplete picture of company leverage. This information asymmetry can lead to mispriced risk, as demonstrated by S&P’s downgrade of Oracle after factoring in its hidden debt.