The scale of off-balance-sheet AI commitments adds a layer of hidden leverage that traditional valuation metrics may be understating, a risk factor equity investors are likely to price in more explicitly if AI demand growth shows any signs of slowing. With Alphabet and Amazon already posting negative free cash flow, further reliance on capital markets to fund these obligations could pressure credit spreads and weigh on sentiment toward the broader AI infrastructure trade. The disclosures are likely to sharpen scrutiny of individual balance sheets heading into upcoming earnings, particularly for companies whose purchase and lease commitments are growing fastest relative to reported capex.
—
Big Tech’s real AI spending bill is running trillions of dollars ahead of what shows up on the balance sheet.
Summary:
- Nine major tech companies, including Alphabet, Meta, Amazon, Microsoft, Oracle and Nvidia, carry roughly $3 trillion in off-balance-sheet commitments tied mostly to AI infrastructure, according to a Wall Street Journal (gated) analysis.
- That figure dwarfs the roughly $600 billion in traditional capex these companies reported over the past year and is about triple their outstanding lease and long-term borrowing obligations.
- Uncommenced lease obligations, including Meta’s massive Hyperion data centre project in Louisiana, totalled $1.2 trillion, roughly four times higher than a year earlier.
- Chip and hardware purchase commitments across the companies stood at $1.9 trillion, with Alphabet’s commitments alone reaching $811 billion as of June, more than double the level reported three months prior.
- These obligations largely remain off balance sheet under current accounting rules until leases begin or products are delivered, making the companies’ true financial exposure harder for investors to assess.
- Alphabet and Amazon have both recently reported negative free cash flow, with commitments of this scale seen as adding risk if AI demand fails to meet expectations.
Big Tech’s disclosed capital expenditure on artificial intelligence infrastructure understates the true scale of its financial commitments by roughly $3 trillion, according to a Wall Street Journal analysis of securities filings from nine major technology companies. The gap stems from massive lease and purchase obligations tied to data centres and chips that remain off company balance sheets under current accounting rules.
The companies examined, including Alphabet, Meta, Amazon, Microsoft, Oracle, Nvidia, Broadcom, SpaceX and Advanced Micro Devices, reported combined traditional capex of about $600 billion over the past year, a figure dwarfed by their off-balance-sheet exposure. Obligations tied to leases that have not yet begun totalled $1.2 trillion, roughly four times the level disclosed a year earlier, while purchase commitments for hardware and other goods stood at $1.9 trillion. Under accounting rules, these obligations typically stay off the balance sheet until rent payments start or products are delivered.
Meta’s Hyperion data centre project in Louisiana illustrates how such deals accumulate hidden exposure. The company signed a lease with an initial four-year term and options to extend for up to two decades, guaranteeing bondholders will be made whole if it exits early, but because it considers that scenario unlikely, no liability appears on its books. Alphabet’s own commitments have grown especially fast, reaching $811 billion as of June, more than double the $332 billion reported just three months earlier, spanning technical infrastructure, inventory and energy agreements that in some cases extend to 2054.
The implications cut both ways. Optimists point to surging AI demand and hardware shortages as evidence that future revenue will comfortably cover these obligations. But the commitments are largely non-cancellable regardless of whether that demand materialises, and some companies once seen as having fortress balance sheets have needed to tap capital markets more frequently. Alphabet and Amazon have both recently posted negative free cash flow, meaning spending already exceeds operating cash flow before accounting for these additional trillions in future obligations.
Morgan Stanley accounting analysts warned in April that the growing size and complexity of these arrangements is making it increasingly difficult for investors to gauge the true leverage of the companies involved, a concern likely to weigh more heavily on markets if AI infrastructure spending continues to outpace revenue growth.
This article was written by Eamonn Sheridan at investinglive.com.