AI Needs Natural Gas To Survive

As recent studies project a big rise in power generation demand from the big datacenters that are proliferating around the United States, the big question continues to focus in on what forms of generation will rise to meet the new demand. Most datacenters have plans to initially interconnect into local power grids, but the sheer magnitude of their energy needs threatens to outstrip the ability of grid managers to expand supply fast enough.

This hunger for more affordable, 24/7 baseload capacity is leading to a variety of proposed solutions, including President Donald Trump’s new executive orders focused on reviving the nation’s coal industry, scheduled to be signed Tuesday afternoon. But efforts to restart the permitting of new coal-fired power plants in the US will require additional policy changes, efforts which will take time and could ultimately fail. In the meantime, datacenter developers find themselves having to delay construction and completion dates until firm power supply can be secured.

Datacenters specific to AI technology require ever-increasing power loads. For instance, a single AI query can consume nearly ten times the power of a traditional internet search, and projections suggest that U.S. data center electricity consumption could double or even triple by 2030, rising from about 4-5% of total U.S. electricity today to as much as 9-12%. Globally, data centers could see usage climb from around 536 terawatt-hours (TWh) in 2025 to over 1,000 TWh by 2030. In January, a report from the American Security Project estimated that datacenters could consume about 12% of all U.S. power supply.