Big Tech Accumulates $1.09 Trillion in Data Center Leases
Microsoft, Meta, Oracle, Amazon and Alphabet have committed about $1.09 trillion in future lease payments that have not yet begun, largely for AI data centers, Reuters reported on August 5. The total is nearly four times the approximately $285 billion of lease liabilities already recognized by the companies. Reuters' analysis identifies Oracle as carrying the largest apparent concentration risk.
Microsoft, Meta Platforms, Oracle, Amazon and Alphabet have committed about $1.09 trillion in future lease payments for facilities that have not yet entered service, mainly AI data centers, Reuters reported in an article published by iTnews on August 5. The aggregate is nearly four times the roughly $285 billion in lease liabilities already recognized on the five companies' balance sheets, based on company filings compiled by Reuters.
The difference reflects lease accounting treatment. Reuters reported that signed leases generally become recognized liabilities only when facilities are available for use; before then, companies disclose the expected payments in financial-statement notes. The reported $1.09 trillion figure therefore represents undiscounted future payments over many years and cannot be added directly to debt, while recognized lease liabilities are measured at present value.
Oracle's reported exposure
Reuters identified Oracle as having the largest apparent concentration risk among the companies reviewed. Oracle disclosed $260 billion of uncommenced commitments, compared with $37.89 billion of recognized lease liabilities, according to the report.
Reuters reported that Oracle's commitments are substantially for data centers expected to begin between fiscal 2027 and fiscal 2029, with terms generally spanning 15 to 19 years. The company has warned in its disclosures that lease duration, renewal terms and pricing may not align with customer contracts, which Reuters reported could create exposure if customers fail to renew or cannot perform.
Reuters' analysis of LSEG data and company filings put Oracle's borrowings at about 4.4 times trailing EBITDA at the end of May. Including recognized operating and finance lease liabilities raised that measure to about 5.7 times, Reuters reported.
What the commitments measure
The reported commitments capture a part of the AI infrastructure buildout that is contractually committed but not yet reflected in conventional balance-sheet lease liabilities. Rating agencies may already incorporate some of these obligations in their assessments, Reuters noted.
For cloud and infrastructure practitioners, the figures underline the long lead times and fixed-capacity economics behind GPU-intensive AI services. In comparable infrastructure cycles, long-duration facility commitments can improve access to scarce power and data-center capacity when utilization remains high, while creating material fixed-payment exposure if demand or customer contracts do not keep pace. The Reuters reporting does not establish whether the five companies' individual facilities will achieve the utilization or revenue needed to support those commitments.
Key Points
- 1Five hyperscalers have reported $1.09 trillion in uncommenced lease payments, exposing the scale of AI data-center capacity already contractually committed.
- 2Lease accounting delays balance-sheet recognition until facilities are usable, so reported liabilities understate future undiscounted payment commitments during construction.
- 3In comparable infrastructure cycles, long-term capacity contracts trade scarce compute and power access against fixed-cost exposure if utilization weakens.
Scoring Rationale
The reported commitments quantify the financial scale of AI data-center expansion across the largest cloud platforms. They are materially relevant to practitioners because long-term facility capacity, power availability, and cloud economics shape access to AI compute, although the report does not announce a new product or technical capability.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
