Creditors are increasingly scrutinizing the financial obligations of major technology companies building out artificial intelligence (AI) infrastructure, as investments in data centres drive significant borrowing and leasing commitments. While AI-related enthusiasm has bolstered equity markets, bond investors are taking a more cautious approach, repricing debt and examining off-balance sheet liabilities more closely.
Large technology firms known as hyperscalers—companies that operate extensive data centre networks to support AI and cloud services—have undertaken substantial capital expenditures funded partly through debt issuance. For instance, Meta Platforms issued $25 billion in bonds earlier this year to finance its data centre expansion. Analysts incorporate this debt directly into leverage calculations, adjusting cash flow forecasts to account for increased interest expenses.
However, the financial picture is complicated by the sizeable lease obligations these firms have committed to for building and operating new data centres. These leases, often off-balance sheet items, represent significant future liabilities that creditors factor into their risk assessments despite being accounted for differently under current accounting standards. Since the introduction of ASC-842 in the United States, lease liabilities are capitalized and included when calculating “lease-adjusted leverage,” a key metric for assessing creditworthiness.
Meta disclosed approximately $347 billion in future lease obligations related to data centre projects not yet operational, contrasting with $84 billion in on-balance sheet debt expected by the end of 2025. The valuation of these leases depends on discount rates, but even optimistic assumptions suggest that lease liabilities remain immense. These figures highlight the potential scale of Meta’s total indebtedness beyond traditional borrowings.
The complexity and variability of data centre leases add further uncertainty. Unlike retail property leases, which are generally standardized, data centre leases can vary widely, making it difficult for creditors to assess risk precisely. Since comprehensive lease disclosures are typically provided only annually, surprises may emerge as new data centre projects progress and lease commitments accumulate.
Sell-side analysts estimate that Meta’s capital expenditures related to this growth could surpass $500 billion through 2028, potentially exceeding forecasted operating cash flow. This suggests ongoing reliance on credit markets for financing, either through bond issuance or leasing arrangements under entities such as the Hyperion venture, a partnership between Meta and Blue Owl focused on building a large data centre in Louisiana.
The combination of rising debt, operating leases, and evolving cash flow creates considerable uncertainty about Meta’s financial trajectory. Lease-adjusted leverage forecasts show a wider range of outcomes than is usual for a company with Meta’s current investment-grade rating. Market pricing reflects some of this caution; Meta’s bonds trade more in line with high-grade triple-B credits such as Verizon despite having a much stronger price-to-earnings ratio.
Historically, equity markets have often discounted off-balance sheet obligations until a tipping point forces recognition by the broader investment community. Creditors continue to refine how these lease commitments fit into their risk models, influencing the credit spreads demanded for holding hyperscaler debt. As scrutiny intensifies, bond valuations and equity multiples may adjust accordingly, potentially narrowing the current divergence between credit and equity market sentiments.