US Data Centers Face 55-Month Grid Delays as 1.2-Gigawatt AI Projects Seek Power
Updated
Updated · worksinprogress.co · Jun 22
US Data Centers Face 55-Month Grid Delays as 1.2-Gigawatt AI Projects Seek Power
3 articles · Updated · worksinprogress.co · Jun 22
Summary
Grid interconnection—not power supply itself—is emerging as the main brake on US AI expansion, with giant campuses such as Stargate in Texas needing up to 1.2 gigawatts.
55 months was the median wait for a power plant to connect by 2023, up from less than 20 months in 2005, as first-come, first-served queues clog with speculative and duplicative requests.
143.5 gigawatts of data centers were seeking connection in ERCOT alone as of October 2025, far above the Texas grid’s 85.9-gigawatt record demand, pushing 62% of data centers to consider off-grid power.
xAI’s Memphis site illustrates the workaround: after initially getting only 8 megawatts from the grid, it installed 422 megawatts of on-site gas turbines while awaiting transmission upgrades.
The report argues the fix is procedural—auction faster queue access and expand flexible 'connect-and-manage' service—because the core problem is delivering electricity where it is needed, not generating it.
As AI's power demand outpaces the grid, will tech giants build their own energy empires to bypass the bottleneck?
Who will pay the trillion-dollar bill to modernize the grid for AI: tech companies or average American families?
What are the hidden environmental costs of the AI boom, from creating urban 'heat islands' to consuming trillions of gallons of water?
U.S. AI Data Center Power Crisis: The 7 GW Shortfall and Its Impact on National AI Leadership
Overview
The rapid expansion of AI infrastructure in the U.S. has collided with a severe and escalating power crisis, leading to massive delays and cancellations of crucial data center projects. By mid-2026, nearly half of all planned U.S. data centers for the year were canceled or significantly delayed, with only about 5 GW of the announced 12 GW capacity under active construction. This has stalled billions of dollars in planned infrastructure, directly impacting the projected $650 billion Big Tech spending cycle, the semiconductor supply chain, and the future trajectory of AI deployment across the country.