Supply-Chain and Grid Constraints Delay or Cancel Part of Planned U.S. Data Center Capacity

A substantial share of planned U.S. data center capacity may not be delivered on schedule. ProLift reports that between 30% and 50% of 16 GW of capacity scheduled to come online is subject to severe operational delays or outright cancellation. The estimate concerns planned capacity, not operational data centers.

The report identifies industrial supply chains and grid capacity as the primary bottlenecks. Sandstone Group is cited as estimating that nearly half of planned U.S. data centers are behind schedule, although the underlying methodology, observation period and affected-project list were not available in the supplied material.

The constraints are particularly material for AI-oriented deployments. Such facilities require substantially higher electrical loads and heavy engineering systems, including liquid cooling and backup diesel generation. The largest hyperscale campuses under construction are described as consuming 20 times as much electricity as a typical AI-focused data center; the source characterizes a typical AI-focused facility as using as much electricity as 100,000 households.

For operators, developers and capital providers, the reported delays suggest that the limiting factor in AI infrastructure expansion is increasingly physical execution rather than demand or compute procurement alone. Secured grid interconnection, access to critical electrical and cooling equipment, and the ability to manage complex construction and logistics could become differentiators. This is an inference from the reported constraints, rather than a confirmed market outcome.

The source also projects global hyperscaler spending at approximately $700 billion, but does not specify the relevant period or exact scope. That lack of definition limits its use as a direct comparison with the U.S. capacity estimate.