US AI Data Center Power Requests Face Massive Shortfall

Wood Mackenzie projects that only about 28 percent of the 1,066 gigawatts of electricity requested for artificial intelligence data centers in the United States will actually materialize. The research firm attributes this gap to phantom projects, where developers submit speculative applications to multiple utilities to secure favorable deals, and unrealistic pitches from inexperienced firms.

These inflated requests have strained grid approval timelines and created risks for U.S. competitiveness in the global AI race. There are also concerns that infrastructure costs associated with these requests could increase utility bills for consumers. In response, some utilities, including Exelon Corp., have reduced their demand pipelines to prioritize credible projects.

State-level interventions have also emerged to manage the surge. In Texas, Governor Greg Abbott ordered a mandatory audit of all data centers seeking grid access on August 3, which effectively paused new approvals. The Electric Reliability Council of Texas is currently tracking 474 gigawatts of connection requests, with data centers representing approximately 90 percent of that total. Meanwhile, PJM Interconnection has implemented new measures to improve forecasting for large load requests across its service area of 67 million people.