Analysis: Alphabet, Amazon, Meta, Microsoft, and Oracle, the top spenders on new US data centers, collectively added ~$350B in debt over the past five years
The largest builders of artificial intelligence data centers have doubled their debt load in the last five years …
Context & Ripple Effects
The debt increase follows a multiyear acceleration in AI-infrastructure investment: Microsoft, Meta, and Alphabet reported more than $32B of combined Q1 capital spending in 2024, while Microsoft, Alphabet, Amazon, and Meta raised first-half 2024 capex to $106B. By early 2026, Alphabet, Amazon, Meta, and Microsoft were forecasting roughly $650B of combined annual capex, driven by data-center construction.
The financing mix is also evolving alongside the spending boom. Related coverage shows Oracle, Meta, xAI, and CoreWeave using special-purpose vehicles to move portions of data-center debt off their balance sheets, making headline debt only one view of the infrastructure-financing burden.
First-order effects
- Alphabet, Amazon, Meta, Microsoft, and Oracle carry materially larger debt loads while continuing to fund U.S. data-center buildouts, increasing the importance of sustaining cash flow and access to financing.
- Oracle and other builders using SPVs can support construction with less debt recorded directly on the parent balance sheet, but the underlying projects still require debt service and operating demand.
Second-order effects
- The scale of borrowing and capex raises the financing bar for smaller AI-infrastructure providers, which must compete for capital and data-center capacity against companies with much larger balance sheets and customer bases.
- Cloud and AI providers face stronger pressure to turn newly built capacity into paid services; expansion into adjacent offerings, such as Meta's stated interest in cloud-like capabilities through Manus, becomes more strategically relevant.
Third-order effects
- If debt-backed construction and off-balance-sheet financing remain central to AI buildouts, AI infrastructure could become increasingly concentrated among platforms able to finance, operate, and monetize data centers at scale.
- The pattern shifts the AI investment debate from chip access alone toward the durability and transparency of the financing structures supporting long-lived data-center assets; the eventual demand utilization of that capacity remains the key uncertainty.
The trend: AI competition is becoming an infrastructure-financing contest, with hyperscalers pairing exceptional capex plans with larger and more complex debt structures to secure data-center capacity.