AI wants electricity now. The electric grid needs years to catch up

The rapid expansion of artificial intelligence (AI) is outpacing the development of the electric grid, according to industry experts. Data centers that power AI technologies have drawn scrutiny for their potential to increase electric bills and create an imbalance in the power supply. Rob Gramlich, president of consulting firm Grid Strategies, stated that tech companies can develop AI infrastructure faster than the U.S. can construct the necessary electricity generation and transmission.

This mismatch in timing could leave the grid unable to meet the electricity demand projected to be nearly 12% of all U.S. electricity by 2030, according to the Lawrence Berkeley National Lab. The demand for electricity is growing faster than expected due to AI, and grid watchdog NERC projects a 69% increase in summer peak electricity demand in North America over the next decade.

The Western U.S. grid region also faces significant challenges, with planned data centers consuming up to 40% of the forecasted demand. The report highlights access to electricity as a bottleneck to data center development, with companies requesting more power and faster delivery rates than the grid can handle. U.S. power projects that came online in 2025 took a median of five years from requesting a grid connection to operating commercially, and utilities often do not connect new customers if they cannot reliably serve them.

This scarcity of power may lead to data centers waiting years for full service or accepting provisional connections that could be interrupted during grid strain. The lack of infrastructure to support electricity demand further exacerbates the issue, with the Department of Energy noting a pressing need for more infrastructure due to load growth from data centers.