AI Data Center Growth Strains U.S. Electrical Grid

The surge in artificial intelligence and cloud computing is driving a critical mismatch between the rapid construction of data centers and the slower expansion of the U.S. power grid. The United States Department of Energy projects that data center electricity consumption will reach between 6.7% and 12% of total U.S. electricity by 2028. While tech firms can build facilities in 18 months, new power plants and transmission lines often require five or more years to complete.

Electric utilities are struggling with capacity planning due to speculative interconnection requests, termed phantom data centers, where companies file requests across multiple regions for a single project. American Electric Power reports that customers are seeking to connect approximately 190 gigawatts of additional load, which is five times the company's current system size. This artificial demand spike risks forcing utilities to either overbuild, increasing costs for ratepayers, or underbuild, leading to potential brownouts.

In North Dakota, officials warned lawmakers that new power plants are necessary to meet the needs of facilities that often require 100 megawatts or more. This rapid growth has already sparked legal disputes over infrastructure costs, specifically between Basin Electric Power Cooperative and Montana Electric Utilities. To mitigate these pressures, some hyperscalers are colocating facilities with nuclear or natural gas plants, while utilities are demanding more precise demand estimates and power purchase agreements.