Particle.news

Download on the App Store

AI Data Center Boom Hits U.S. Grid Bottleneck, Forcing Power Workarounds

Years-long interconnection queues and equipment delays are pushing some developers to secure their own power.

Image
Despite being a major factor in growing energy demands, AI can also offer key solutions.
The Millstone nuclear power facility in Waterford.  (AP Photo/Steve Miller, File)
Image

Overview

  • A JLL report says data center vacancy has fallen to a record 2.3%, with half of new demand concentrated in Northern Virginia and Dallas and a construction pipeline that is 73% preleased.
  • Power requests now reach gigawatt scales, including single interconnection bids as large as 5 GW, while speculative filings flood utilities and create years-long backlogs such as roughly 40 GW of requests in Chicago.
  • Developers are increasingly turning to on-site solutions such as natural gas turbines and fuel cells and, in some cases, bypassing utilities, as critical equipment like transformers face four-to-five-year lead times.
  • The U.S. Department of Energy projects data centers could consume up to 12% of U.S. electricity by 2028, as commercial power rates have risen nearly 30% in five years and 75% of new builds target lower-cost markets.
  • Analysts note China’s state-led buildout and high reserve margins of 80%–100% versus roughly 15% in the U.S., even as AI-driven grid software shows promise in boosting resilience and hosting capacity.