AI hits power and land limits
- AI compute growth is colliding with limited data‑centre land and electricity capacity in hubs like northern Virginia. - IEA‑aligned analysis projects electricity demand from AI‑optimised data centres could more than quadruple by 2030. - Operators are rethinking site selection, utility partnerships, and distributed power to secure megawatts. (us.cnn.com)
Artificial intelligence is running into a basic limit: in the biggest data-center hubs, there is not enough power or land to keep building at the current pace. (cnn.com) Training and running AI models happens inside data centers, which are warehouses full of servers that draw electricity around the clock and need extra power for cooling. The International Energy Agency said on April 10, 2025 that electricity demand from AI-optimized data centers could more than quadruple by 2030. (iea.org) The same International Energy Agency report said total global electricity demand from data centers is set to more than double by 2030 to about 945 terawatt-hours, roughly equal to Japan’s current annual electricity use. In the United States, data centers are on course to account for almost half of electricity-demand growth through 2030. (iea.org) Northern Virginia sits at the center of that squeeze because it remains the biggest data-center market in North America, with 5.6 gigawatts of capacity in JLL’s mid-2025 market tally. CBRE said the region also delivered 1 gigawatt of new capacity in 2025, even as the market grappled with energy-sourcing constraints. (jll.com) (cbre.com) Utilities are already planning around that demand. Dominion Energy said in May 2025 that it had about 40 gigawatts of data-center capacity in various stages of contracting in Northern Virginia, up from 21 gigawatts in July 2024, and raised its 2025-2029 capital-spending forecast to $50.1 billion. (datacenterdynamics.com) Dominion also said it connected 15 new data centers in 2024, adding nearly 1 gigawatt of capacity, and expected to connect another 15 in 2025. The utility proposed a new rate class for customers with contracted loads of 25 megawatts or more so large users would cover more of their own costs. (datacenterdynamics.com) The grid strain is no longer theoretical. A Harvard Belfer Center policy brief said a voltage fluctuation in Northern Virginia in July 2024 disconnected 60 data centers at once, creating a 1,500-megawatt power surplus that required emergency action to avoid wider outages. (belfercenter.org) State officials in Virginia have been warning about the cost of keeping up. A December 2024 study prepared for the Virginia Joint Legislative Audit and Review Commission said unconstrained data-center growth could add almost $40 a month to residential power bills if infrastructure costs are broadly socialized. (wvtf.org) (jlarc.virginia.gov) That is pushing operators to look beyond the old model of picking the biggest internet hub and waiting for utility power to arrive. Industry and utility reports describe a shift toward new markets, direct utility partnerships, on-site generation, and distributed power systems that can secure megawatts faster than a traditional grid connection. (cnn.com) (bloomenergy.com) Dominion’s own long-term plan says Virginia is seeing its biggest power-demand growth since the years after World War II, driven heavily by data centers, and calls for an “all-of-the-above” buildout of generation, storage, transmission, and efficiency. The next fight is less about whether AI demand is real than about who gets the land, the wires, and the bill. (datacenterfrontier.com)