Big Tech AI 'Hidden Debt' Over $1 Trillion: Key Findings

Reports have circulated that some of the largest tech companies, specifically the "Big Five AI hyperscalers" (Alphabet/Google, Amazon, Meta, Microsoft, and Oracle), have accumulated over $1 trillion in "hidden debt." This debt is reportedly kept off their balance sheets through complex ownership structures. While the AI spending spree has faced skepticism, with OpenAI alone committing over $600 billion by 2030, the concern now is the extent of these off-balance sheet borrowings.

Despite the term "hidden debt," these amounts are not entirely concealed. Each company's financial filings include explicit sections on commitments or contractual obligations that detail these figures. The goal of this analysis is to compile these off-balance sheet items to provide a clearer understanding of the companies' financial positions.

On-Balance Sheet Debt

Historically, with the exception of Oracle, major US tech companies have maintained strong balance sheets with minimal borrowing and substantial free cash flow. While AI spending has led to an increase in debt, their balance sheets remain relatively healthy.

As of their latest reports (June 30th for most, May 31st for Oracle), and with minor adjustments for recent announcements (e.g., Alphabet's $25 billion bond issuance), the situation is as follows:

• Alphabet and Microsoft: Both companies have negative net debt balances, meaning their cash on hand exceeds their borrowings (including recent bond issuances and leases).
• Other Companies: For those with positive net debt, their net debt-to-income ratios are generally small. Oracle is an outlier with more substantial debt, reflected in higher debt-to-equity ratios compared to the others.
• Sector Averages: For the other four companies, debt-to-equity ratios are at or below the US information technology sector average of 0.51x. Net debt-to-income is very low across various profit measures.
• Interest Coverage: These companies can easily cover their debt costs, with interest expenses representing a small fraction of total revenues. While interest expenses are accelerating, two companies even have negative net interest expenses, earning more interest on cash than they pay on debt.

It's important to note that while these companies have significant cash, much of it is earmarked for future investments, limiting their ability to pay off debt immediately. Nevertheless, their on-balance sheet debt appears manageable when compared to profitability.

Off-Balance Sheet Obligations

The "hidden debt" primarily falls into three or four categories:

1. Non-Cancellable Contractual Commitments

These are legal obligations to purchase goods or services in the future. They typically include:

• Components for Data Center Infrastructure: Such as semiconductors, secured through purchase agreements to ensure supply.
• Power Purchase Agreements: Long-term "take or pay" arrangements for data centers, where companies pay for electricity whether used or not, incentivizing power providers to invest.
• Third-Party Cloud Capacity: Renting compute power for AI models from other companies to ensure necessary supply.

These agreements are legally binding but do not appear as liabilities on the balance sheet until the goods or services are delivered, adhering to standard accounting rules.

Commitment Amounts: The five companies have collectively committed over $1.5 trillion in spending, predominantly for AI data center buildouts. This is a significant increase compared to five years ago, indicating a strong link to AI expansion.

Company Breakdown: * Alphabet: Has the largest commitment at over $800 billion. * Meta: Follows with approximately $350 billion. * Percentage of Assets: Alphabet and Meta stand out, committing a majority of their total assets to these future obligations.

2. Funding and Construction Commitments

This category involves investment commitments, such as those related to joint ventures or funding obligations (e.g., Microsoft's investment in OpenAI). While smaller than contractual commitments, totaling around $100 billion cumulatively, they are still material. These can sometimes be contingent on meeting certain benchmarks.

3. Uncommenced Leases

These are lease agreements that have been entered into but have not yet commenced, so they do not appear on the balance sheet. This is standard accounting practice.

• Types of Leases: Most are operational leases (renting an asset over time), though some may be financial leases (like a mortgage).
• Debt-like Liability: Many analysts view operational leases as debt-like because they are legally binding, with consequences for missed payments similar to debt defaults.
• Substantial Increase: Companies plan to significantly increase their lease agreements. Across the five companies, there are over $1 trillion in uncommenced lease obligations, which would more than double their current total debt (including existing leases) of around $700 billion.
• Company Breakdown: Meta and Microsoft have the largest dollar amounts, each exceeding $300 billion in future leases.

These commitments are concerning given the substantial spending on AI infrastructure, an area whose long-term financial viability is still unproven.

4. Financial Guarantees and Backstops

This is arguably the most controversial category, where debt is actively being "hidden." Some companies establish Variable Interest Entities (VIEs) or Special Purpose Vehicles (SPVs).

• VIE/SPV Structure: Companies set up these entities, owning less than 50% to avoid consolidating their assets and liabilities onto their own balance sheets. Instead, the VIE's value is summarized as a single line item.
• Meta's Example (Hyperion): Meta co-developed a $50 billion data center project in Louisiana called Hyperion, owned by an SPV called Benet Investor (20% owned by Meta). This SPV raised $27.3 billion in corporate bonds. While Meta will be the sole lessee of this data center, it provided a residual value guarantee for the SPV's debt. This means Meta is ultimately responsible if the SPV defaults, effectively making the SPV's debt Meta's responsibility, despite not appearing on Meta's balance sheet.
• Meta's Example (El Paso): Meta has a similar joint venture with BlackRock for an SPV called Soapia Investor in El Paso, Texas, which has issued $12.3 billion in secured notes, also backed by Meta.

While these details are in financial filings, such structures appear to be an elaborate way to move debt off the main balance sheet.

Guaranteed Debt Amounts: * Meta and Alphabet: Have guaranteed tens of billions in debt. Meta's guarantees include a $14.7 billion contingent purchase commitment. * Oracle: Has $3.3 billion in similar guarantees.

While these amounts are smaller than other off-balance sheet items, they are material.

Contextualizing the Obligations

The critical question is whether these companies can handle these obligations.

• "Hidden Debt" Accusations: The accusations of over $1 trillion in hidden debt are somewhat exaggerated. While Meta is actively transferring debt off its balance sheet, it's primarily two companies (Meta and Alphabet) with guarantees under $100 billion. This amount, while material, could theoretically be covered by less than one year of earnings if the company had to take the full amount.
• Not Direct Debt Equivalents: Many off-balance sheet items, like contractual commitments, are not direct equivalents to debt. They represent commitments for goods or services that the company expects to need, similar to personal subscriptions or rent.
• Present Value: The reported amounts are nominal sums of all future payments, not present values. This means they don't account for the time value of money. For example, Meta's $350 billion in uncommenced leases represents payments for agreements starting through 2036 with terms up to 30 years. The timing of these payments significantly impacts the actual burden.

Impact on Credit Ratios

If these off-balance sheet items (excluding financial guarantees, which are often related to leases) are added to total debt, the debt ratios would significantly worsen compared to historical standards.

However, it's important to consider that higher debt balances are typical for traditional data center companies. For instance, Equinix, a pure-play data center company, had comparable debt ratios five years ago when including lease and spending commitments. This suggests that as tech companies move into the hyperscaler space, their balance sheets might naturally evolve to reflect this capital-intensive model. Oracle remains an outlier with a weaker financial position.

Risks and Future Outlook

Despite the potential normalization of higher debt levels for data center operations, this shift represents a real risk:

• Capital-Intensive Model: These companies are moving from a capital-expenditure-light model to a very capital-expenditure and asset-heavy business model, which is less attractive from a profit standpoint.
• Increased Risk Profile: Any fixed payment obligation increases risk, especially when revenues can fluctuate. The risk profile for these big tech companies is deteriorating.
• Loss of Nimbleness: In the past, companies like Meta could pivot from unsuccessful projects. With these massive commitments, abandoning projects is no longer a viable option. This signifies a "point of no return" for AI investments, making companies less nimble if market conditions change.
• Funding Sources: Much of this buildout has been funded by the traditional, highly profitable businesses of these companies, rather than proven AI revenues. This means long-term liabilities are being locked in with short-term, unproven revenues to service them.

While these companies are taking a calculated gamble based on high expectations for future AI demand, the long-term implications are uncertain. It's unclear if these spending commitments are additional to existing capital expenditure trends or if revenue will grow fast enough to justify them.

This analysis is not an investment recommendation but aims to highlight a significant risk for investors in these companies. The situation could continue to evolve, and how companies manage these obligations will be crucial.