From the front window of her home, Breanne Kidd once watched sunrise over farmland while sipping coffee before children arrived at her in-home daycare. Today, that view is obscured by cranes, steel and dust as construction advances on Meta’s 800-acre Bowling Green data center, along with an unexpected addition she says she was never told about: a natural gas power plant built to serve the facility.
“It’s not like we’re two streets away. We’re literally across the street,” Kidd said, pointing toward the Apollo Generating Station site about 25 miles (40.2 kilometers) south of Toledo. “I’m living next to a threat.”
The plant is part of a rapidly expanding wave of off-grid power projects being developed across the United States to meet surging electricity demand from the artificial intelligence and data center industry, according to a Reuters review of regulatory filings, public records and interviews with officials, residents, researchers and company executives.
Unlike traditional power plants, these facilities are designed to serve single industrial customers and are often moving from proposal to construction in a matter of weeks or months. That pace has allowed developers to bypass lengthy permitting processes, environmental impact studies and public hearings that typically accompany large-scale energy infrastructure.
Developers argue that privately contracted generation projects are exempt from many of those requirements, a position that has left local residents with limited visibility into plants that may affect air quality and contribute to climate change emissions.
Transparency has been further reduced in some cases through non-disclosure agreements with local governments, the use of shell companies and expedited permitting processes that limit public review, according to Reuters reporting.
“The AI industry’s off-the-grid natural-gas generation is emerging as one of the largest under-examined air-quality risks in the country,” said Michael Cork, a postdoctoral researcher at Harvard University.
The Apollo plant, designed to generate enough electricity for roughly 100,000 homes, was approved by the Ohio Power Siting Board on Feb. 3, less than three months after plans were submitted. A draft air permit was not released publicly until March, after construction had already begun, according to regulatory records. The paperwork initially identified the client only as a Meta subsidiary, Liames LLC.
Data from research firm Cleanview shared with Reuters shows at least 57 off-grid power projects are either proposed or under construction in the U.S. to serve individual data centers, with combined capacity of about 73,000 megawatts — enough to power tens of millions of homes.
Reuters identified more than a dozen projects approved in under a year with limited public notification. Two such facilities are already operational, including a plant tied to Elon Musk’s xAI outside Memphis and another in Ashburn, Virginia, serving Vantage Data Centers.
The speed and scale of approvals, as well as the regulatory mechanisms used to accelerate them, have not previously been widely documented. Most of the facilities rely on natural gas, which produces nitrogen oxides, fine particulate pollution linked to respiratory disease, and greenhouse gases that contribute to climate change.
Supporters of the projects argue they are necessary to meet the electricity demands of rapidly expanding artificial intelligence infrastructure and to prevent strain on public grids or increases in consumer electricity prices.
The Trump administration has pushed for faster permitting for AI-related infrastructure, citing competition with China. The Environmental Protection Agency and several states, including Ohio, West Virginia, Texas and Utah, have proposed or adopted measures aimed at accelerating approvals, according to a Reuters review of regulatory actions.
The EPA said it is pursuing a “deliberate, governance-driven approach” as it seeks to position the United States as the “AI Capital of the World,” while noting that permitting authority often rests with state and local agencies.
