Big Five Tech Giants' AI Data Center Lease Commitments Exceed $1 Trillion
N.R. Finch
Microsoft, Meta, Oracle, Amazon, and Alphabet have committed roughly $1.09 trillion in leases for data centers not yet in service — nearly four times the $285 billion in lease liabilities on their books. This means → the real leverage behind AI infrastructure is far larger than reported balance sheets suggest, and whether demand materializes will determine if these commitments are assets or albatrosses.
$1 trillion in lease commitments — why doesn't it count as "debt"?
The five companies' combined uncommenced lease commitments total roughly $1.09 trillion, against only about $285 billion in recognized lease liabilities — a ratio of nearly 4 to 1.
In plain terms = contracts are signed and future payments are locked in, but because the data centers are not yet in service, accounting rules allow these obligations to stay off the liability line — disclosed only in footnotes.
Rating agencies may already factor in some of these commitments, but the vast majority of the exposure does not appear in reported leverage metrics.
How much has each company committed?
Microsoft leads the pipeline: $329.1 billion in uncommenced leases, against $88.52 billion in recognized lease liabilities.
Meta disclosed $279 billion, then signed an additional $68 billion in data-center leases in July — pushing the five-company known pipeline to roughly $1.16 trillion.
Oracle disclosed $260 billion, Alphabet $85.2 billion, and Amazon $137.21 billion — though Amazon's figure is less directly comparable because its lease portfolio also covers warehouses, offices, aircraft, and vehicles.
Why does Oracle stand out?
Oracle's uncommenced lease commitments of $260 billion are nearly seven times its recognized lease liabilities of $37.89 billion — the highest concentration ratio among the five.
These leases are expected to commence between fiscal years 2027 and 2029, with terms typically running 15 to 19 years. This means → once activated, Oracle will carry payment obligations stretching close to two decades.
Oracle itself has flagged the risk: lease durations and pricing may not match its customer contracts. If customers fail to renew or cannot perform, Oracle is directly exposed.
How stretched is Oracle's leverage?
As of late May, Oracle's borrowings stood at roughly 4.4 times EBITDA; adding recognized operating and finance lease liabilities raises that to about 5.7 times.
S&P Global Ratings said it has incorporated the $260 billion in uncommenced leases into its adjusted debt projections, forecasting leverage of about 4.4 times by fiscal 2027.
In plain terms = even with rating agencies already doing the math, Oracle carries the heaviest burden of the five.
Will these commitments prove to be assets — or albatrosses?
If AI compute demand keeps growing, these facilities underpin the next phase of cloud expansion — commitments become first-mover advantage.
If demand falls short, each company faces years of paying for high-cost, long-cycle capacity it cannot easily exit — commitments become sunk costs.
This reflects the central divide in how the market evaluates these off-balance-sheet obligations: the question is not "how much was spent" but "will the demand actually show up."
Content is for reference only, not financial advice.