

With designs on the White House, some Democratic governors are one-upping each other with costly regulations.
I n June 2025, Pennsylvania Governor Josh Shapiro celebrated the largest private-sector investment in his state’s history: He announced that Amazon would be investing $20 billion into artificial-intelligence infrastructure in the Keystone State. His administration promised to help secure additional sites, build the necessary infrastructure, and move permits quickly.
By August 2026, Shapiro was singing a different tune. Data-center developers would have to bring or pay for their own power, meet new environmental and transparency standards, negotiate community-benefits agreements, and satisfy state officials before receiving incentives or expedited permitting. He removed data centers from Pennsylvania’s fast-track permitting program and declared that his state would not be “bullied” by developers.
Shapiro isn’t alone. Illinois Governor JB Pritzker, who signed his state’s data-center incentive program in 2019, froze new incentive agreements this year. Then came September. Virginia Governor Abigail Spanberger proposed a sweeping 2027 regulatory package. California Governor Gavin Newsom signed seven data-center oversight bills. Maryland Governor Wes Moore established a public rating system for large projects and called for repeal of his state’s data-center tax exemption.
Shapiro, Pritzker, Newsom, and Moore are routinely discussed as possible 2028 presidential contenders, while Spanberger has obvious potential for a national ticket. Moore made the national interest explicit when he condemned President Donald Trump’s regulatory policies and accused Big Tech of writing its own rules. None of these governors want to be portrayed in a Democratic primary as friendlier to Trump or Big Tech than the governor down the street.
People have legitimate concerns about data centers. I have wrestled with these issues for years as a Virginia legislator, state energy director, and now an adviser on PJM and large-load development. Those concerns can be addressed without sacrificing the enormous economic and strategic benefits that AI infrastructure can provide.
Data centers should pay the incremental generation, transmission, and distribution costs required to serve them. Speculative projects shouldn’t clog utility or regional transmission queues. Developers should provide accurate load forecasts, make binding financial commitments, and address genuine water and environmental impacts. Local communities deserve reliable information before making land-use decisions.
The Trump administration has embraced much of that reasonable framework. Its ratepayer-protection pledge calls on technology companies to build, bring, or buy the power their facilities need so those costs don’t land on households. At the same time, the administration’s AI Action Plan recognizes that rapid construction of data centers, generation, and semiconductor facilities is essential to competing with China.
That is the balance America needs. Data centers should pay their share, and projects that meet clear standards should expect a predictable and timely path to construction.
These governors are doing the opposite. “Ratepayer protection” is becoming a vehicle for political one-upmanship. Each governor has an incentive to add another condition, withdraw another incentive, announce another review, or attack the industry more aggressively than the last. “Guardrails” become handcuffs when requirements are duplicative, standards change after projects are underway, previously approved developments face new votes, and complying with one level of government offers no certainty at the next.
No single one of these rules will stop America’s AI buildout — but layer them together and they might. A developer may face local zoning approval, a state certification process, environmental reviews, utility requirements, a state commission proceeding, regional grid studies, and federal oversight. If every ambitious governor creates a different test and keeps changing it with the political winds, projects will be delayed, capital will move elsewhere, and power infrastructure won’t be built on time.
That is not merely an economic-development problem. The race for AI leadership is a race for computing capacity and the electricity that supports it. The country that can build and energize that infrastructure will have an advantage in military systems, intelligence, cybersecurity, advanced manufacturing, medicine, and nearly every other strategic field. China doesn’t need to defeat American innovation if American politics prevents it from being deployed at scale.
A better approach would establish durable national principles while preserving legitimate state and local authority. Large loads should pay the costs they cause. Developers should bring generation or make enforceable long-term commitments, disclose credible end users and expected loads, and lose their queue positions if they miss defined milestones. Communities should receive meaningful information and an opportunity to participate.
In return, governments should provide: firm review deadlines, concurrent rather than sequential permitting, respect for existing approvals, and protection against retroactive rule changes. A project that satisfies published requirements should be allowed to proceed without becoming collateral in the next campaign.
America’s governors are right to recognize that the data-center boom requires rules. But rules should solve identifiable problems, not provide applause lines for a presidential campaign. The Democratic nomination won’t be decided until 2028. The damage from its shadow primary may arrive much sooner.