The United States is facing a new kind of energy crisis, driven not by households or factories, but by the exploding demand from artificial intelligence datacenters. 
As Big Tech races to build ever larger clusters of servers to train and run AI models, the nation’s outdated power grid is being pushed toward a critical breaking point.
Companies like Meta, Amazon, Google, and Microsoft are pouring billions into AI infrastructure. But each new datacenter requires staggering amounts of electricity, with a single interconnection request reaching up to 5 gigawatts-equivalent to powering more than five million homes. With grids still relying on infrastructure from the 1970s, the pace of upgrades simply cannot keep up with this exponential growth.
Goldman Sachs warns that energy demand from datacenters is already outpacing America’s ability to expand its power systems. Gartner projects that by 2027, as much as 40% of US datacenters may lack sufficient power to operate fully, with total demand soaring toward 500 terawatt-hours. Rising electricity bills are expected to follow: in high-datacenter regions, monthly costs could climb an additional $14–$37 by 2040.
Big Tech’s solution? Build their own power plants. Meta has begun constructing natural-gas-powered facilities for multi-gigawatt AI clusters, while Microsoft is reviving the Three Mile Island nuclear site to fuel its next generation of supercomputers. Amazon and Google are exploring similar paths, effectively turning technology giants into energy producers.
While this strategy may relieve long-term strain on the national grid, power plants take years to build-far slower than the pace at which datacenters are being deployed. For now, the mismatch leaves the US vulnerable to shortages and rising costs. Political leaders are under pressure to modernize the grid quickly, with the Trump administration’s proposed AI energy plan promising a path forward to stabilize America’s electricity system in the age of artificial intelligence.