Pennsylvania와 Texas 주 당국이 AI 개발사에 자가발전설비 구축 또는 신규 시설 확충 중단을 경고하며 전력망 부하를 이유로 제시했다.

Pennsylvania and Texas are taking steps to ensure companies planning massive data centers cover the power needs those projects create instead of pushing that demand onto everyday utility customers.

In Pennsylvania, Governor Josh Shapiro has imposed a moratorium on "any AI data center that does not bring, develop, and pay for their own power." Texas Governor Greg Abbott has stopped additional data centers from connecting to the state grid. This state-level resistance reflects growing concern over the enormous amount of electricity required by AI facilities.

Many proposed AI data centers are being paired with their own natural gas plants. While this setup keeps new costs from being passed on to households and small businesses, it could also mean less transparency around air pollution and climate impacts. The EPA issued a statement saying the Clean Air Act's Acid Rain Program does not apply to "islanded" power plants, or facilities built to serve a single site and not connected to the broader grid.

According to Mindy Goldstein, who heads Emory Law School's environmental law program, removing Acid Rain Program coverage eliminates one permit requirement for some projects, even though other Clean Air Act rules can still apply.

The AI boom has become as much an energy story as a technology story. Training and operating advanced AI systems requires enormous amounts of around-the-clock electricity, driving utilities and developers to race to add new power capacity. OpenAI has announced plans tied to a 9.2-gigawatt natural gas plant in Ohio. Amazon is planning a 7.65-gigawatt gas plant in Texas, while Nexus has proposed a 6-gigawatt facility there. Smaller projects are also moving forward, including a 90-megawatt plant in Georgia for a Serverfarm data center.

David Spence, a University of Texas professor focused on energy law and politics, noted: "For a long time, we just didn't have anybody building gas plants at all because wind and solar are cheaper. Then along came data centers with their incredibly large demands for power that has to be 24-7 and a lot of money to spend on these things."

AI's expansion brings both opportunities and risks. The technology could help optimize electric grids, improve wind and solar forecasting, and make buildings and factories more efficient. But it also brings serious concerns, including massive utility use, higher energy costs for customers, security risks, misuse, and broader social disruption if its expansion outpaces safeguards.

Much of the key decision-making is falling to state regulators, who will be responsible for permitting many of these projects. The EPA has said the guidance "expands opportunities for companies to develop and operate islanded power generation facilities for data centers. This will give developers greater flexibility in where and how quickly they can build new facilities while helping to protect the environment and reducing burdens on communities' electric grid." The agency has also proposed changes that would further streamline permitting for some "minor" pollution sources.

However, some communities and watchdog groups are pushing back. In Georgia, state regulators issued violation notices to VoltaGrid and Serverfarm after construction reportedly began before permits were secured. A lawsuit by the NAACP alleges xAI built a 495-megawatt plant in Mississippi to power a Memphis data center without even applying for permits. According to Goldstein, "This is a piecemeal approach to trying to clear the way for data center siting and for the siting of corresponding electric generating units. It seems like a continuous knockdown of dominoes to clear the way for these data centers, and they're leaving states and communities to hold it back."

Texas and Pennsylvania are providing a preview of what a bring-your-own-infrastructure approach looks like as AI developers chase cheap land and quick approvals. Similar battles centered on gas plants, grid strain, and electricity demands are emerging across the country. Nationwide, AI facilities already devour 6% of U.S. electricity, intensifying the power scramble and prompting state leaders to demand that developers shoulder the costs of their own expansion.