A substantial portion of AI infrastructure spending is locked in and not yet reflected as liabilities, signaling significant future capacity growth and financial risk for hyperscalers.

The surge in lease commitments feeds a broader AI-capital expenditure boom, with potential macro-financial risks if demand growth slows and infrastructure becomes a long-term liability rather than an immediately productive asset.

Oracle’s future obligations are tied to data centers anticipated to start between 2027 and 2029, contributing to leverage near 5.7x EBITDA when including existing obligations and prompting a downgrade to BBB- from S&P in July.

Ratings agencies may already weigh these commitments, but the scale shows how much AI-related capacity has not yet entered reported debt and leverage metrics.

There exists an accounting gap where future lease obligations aren’t immediately reflected as debt-like liabilities, potentially affecting leverage metrics and investor perception.

Big Tech players—including Microsoft, Meta, Oracle, Amazon, and Alphabet—hold about $1.09 trillion in uncommenced lease commitments for data centers powering AI, which are not yet recognized as liabilities on balance sheets.

The overall risk rests on whether demand for AI accelerates; if demand slows, firms could be stuck paying for large, long-lived capacity that cannot easily be scaled down.

Oracle alone carries the largest concentration risk with about $260 billion of uncommenced commitments versus $37.89 billion of recognised lease liabilities, with facilities slated to begin 2027–2029 and remaining for 15–19 years.

Total uncommenced commitments are almost four times the roughly $285 billion in reported lease liabilities, driven by accounting that records liabilities only when facilities become usable.

Many commitments are under construction and not recorded as standard lease liabilities, leading some analysts to view them as hidden debt that could pressure profits if AI demand undershoots.

The size and uncommenced nature of these commitments raise questions about balancing capex and operating leverage with actual AI revenue growth, with investors watching data center activation timelines and demand trends in quarterly results.

S&P Global Ratings has incorporated Oracle’s uncommenced leases into its adjusted-debt forecast, projecting leverage around 4.4x in fiscal 2027, and noting that the pipeline could affect ratings if demand slows.