AI boom hides a massive debt bill

As a wave of optimism surrounding artificial intelligence drives the shares of major technology companies to record highs, contrasting signals are emerging in credit markets, where lenders are reassessing the risks of financing AI infrastructure amid concerns over massive financial obligations that are not fully reflected on balance sheets. A report published by the Financial Times indicates that data center lease contracts represent one of the most prominent sources of concern, as they may conceal long-term financial burdens as significant as traditional debt, while technology companies race to expand their computing capacities to meet the growing demand for AI applications.

While equity investors focus on earnings growth and future expectations, lenders look at companies’ ability to meet their obligations and repay debts. The divergence between stock performance and bond pricing reflects a widening gap between optimism about AI opportunities and concerns regarding the cost of financing it. These risks are gaining importance as major technology companies shift from relying on available cash flow to borrowing to fund their capital investments.

While the impact of traditional bonds is easily calculated through their face value, interest rates, and maturity dates, obligations arising from data center lease contracts appear more complex, particularly when projects are still under construction.

**Hidden Debts**

The report stated that credit analyses treat long-term lease contracts as financial obligations that should be accounted for when assessing corporate indebtedness, even if they are not classified in the same way as loans and bonds. Under the methodology used in adjusted leverage analysis, operating lease obligations are added to debt, and the adjusted debt is then measured against earnings before interest, taxes, depreciation, and amortization (EBITDA), after adding lease expenses to this metric. The significance of this treatment lies in the fact that, like interest payments on bonds, leases represent ongoing contractual commitments. They can become a heavy burden if revenues or cash flows decline, even if reported debt levels appear relatively modest.

Meta provides a clear example of the scale of potential obligations. According to the report, it disclosed future lease commitments of approximately $347 billion for projects that have not yet entered operation, while its debt stood at around $84 billion by the end of 2025. These figures do not mean that lease obligations will convert entirely into debt at the same nominal value; the final accounting value depends on the timing of contract commencement and the discount rates used to calculate their present value. However, they reveal the scale of future commitments that investors and lenders must consider when evaluating the company’s financial position.

The picture becomes more complex with the “Hyperrion” project, a massive data center in Louisiana built by Meta in partnership with private investment firm Blue Owl. Financing infrastructure through lease arrangements and joint ventures may make it more difficult to determine the total economic obligations associated with them than calculating direct debt.

**Spending Exceeding Income**

Although Meta has not provided clear guidance on its capital spending after 2026, analysts’ estimates, according to the report, place total expected spending through 2028 at more than $500 billion, a level that exceeds the company’s expected operating cash flows during the same period.

This means that expanding AI infrastructure may require continued reliance on credit markets, whether through issuing new bonds or through leasing and financing arrangements tied to data center projects. Moreover, limited periodic disclosure of details regarding lease agreements may lead investors to await annual reports for a clearer picture of the scale of obligations, without any guarantee that these reports will provide sufficient detail on the varying terms of each project. The paradox lies in the divergent views of equity and bond markets toward Meta. According to the report, Meta’s bonds trade at yields that reflect higher risks than its strong AA credit rating suggests, with pricing converging toward that of Verizon, which is rated BBB. In contrast, this caution is not as clearly reflected in the equity market, where Meta’s stock trades at a price-to-earnings multiple more than double that of Verizon’s. This disparity reflects the differing nature of investors: bondholders are primarily concerned with the company’s ability to repay principal and interest, while equity investors bet on earnings growth and the future value of the business. An emphasis on growth expectations may lead to a reduced weighting of long-term obligations in equity valuations. The issue is not merely whether technology companies can attract financing, but rather the sustainability of their large-scale data center spending models and their ability to generate cash flows sufficient to cover debt, leases, and new investments. Historical experience shows that equity markets may overlook unclear or less visible obligations in balance sheets until financing conditions change or earnings slow. In such cases, the risks reflected in the bond market may gradually translate into downward pressure on equity valuations as well. For major technology companies, the continuation of the AI boom will depend on striking a balance between capital expansion and actual financial returns. For investors, reading the balance sheet in its broader sense—including lease agreements and project-related obligations—will become more important than relying solely on reported earnings and growth indicators.