Pew counts two-thirds of planned US data centers in rural counties

Take Pew's count at face value and 67% of 1,500 is about 1,005 rural projects ^[1]. The 13% of roughly 3,000 operating sites that sit outside urban areas comes to about 390 ^[2]. The pipeline therefore holds about 2.6 rural projects for every rural data center running today ^[4]. Pew's total is "more than" 1,500, so that ratio is a floor ^[1].

The hyperscale figures say more about power. The 120 campuses under construction and the 460 in planning add up to 580 ^[5]. At the 100-megawatt threshold that defines the category, they come to at least 58 gigawatts of peak draw ^[6]. US data centers used about 192 terawatt-hours in 2024 ^[9]. Spread over the 8,760 hours in a year, that works out to an average load of about 22 gigawatts ^[7]. Peak draw and average use are different measures, and not every campus will run flat out. Even so, the floor of the hyperscale pipeline is more than two and a half times what the whole fleet averaged in 2024 ^[8]. Projections cited in the report put data centers at 9% to 17% of US electricity by 2030, two to almost four times the current share ^[10][9].

For decades the industry clustered near metro customers and fiber ^[15]. The report explains the move by saying urban grids in the old corridors are running out of room ^[18], while rural counties offer more available power capacity along with cheaper land and fewer permitting bottlenecks ^[14]. The same report says many developers are putting natural gas turbines, batteries and renewables on their own campuses to avoid local grid constraints ^[11]. Part of the reason is so they can tell residents their rates will not rise ^[17]. Some of the newest campuses are designed at gigawatt scale ^[7]. At the report's ratio of 80,000 homes per 100 megawatts ^[8], one of those draws what about 800,000 homes would ^[11]. I think what a rural county sells is acreage on which a developer can build its own supply, and a permitting process short enough to let it. Pew's figures, as reported, do not separate how much of the move comes from power and how much from land and permits.

Most of the eventual total depends on conversion: the 460 planning-stage hyperscale projects outnumber the 120 under construction almost four to one ^[10]. County politics could decide it instead. Some local governments offer tax incentives and fast-track permits, while others impose moratoriums or require environmental impact studies before construction ^[13]. Or on-site generation spreads until each campus is a private power plant with its own approvals to win. I'd expect the county vote to be the binding limit. The counter-thesis is that power stays the limit and local resistance fades as campuses supply themselves. The siting view is wrong if the projects that reach construction hold near Pew's 67% rural share as moratoriums spread, and right if the rural share falls between planning and construction ^[1].