Pennsylvania Data Centers: Rules After Backlash

Key takeaways

• Gov. Josh Shapiro says some data center developers ignore local communities.
• He wants companies to help pay for the power and roads they need.
• AI is pushing demand for large computing sites across the United States.
• Pennsylvania must balance new jobs with higher energy use and local concerns.

Pennsylvania data centers are large buildings packed with computers that store data and run online services. Gov. Josh Shapiro says some developers want the benefits without helping nearby towns. He wants clearer rules, fair costs and a stronger voice for residents before projects move ahead.

Shapiro made the comments in a CNBC interview on Tuesday. His remarks show a growing fight over the next wave of AI infrastructure. Companies need more computing power, but communities must live with the noise, water use and power demand.

Why are Pennsylvania data centers under pressure?

Data centers support cloud apps, video sites, online stores and AI tools. They run day and night, so they need steady electricity and cooling. A single large site can use as much power as a small city.

Shapiro did not reject data centers outright. Instead, he criticised developers that, in his words, “don’t care about our communities.” He said companies should work with local people rather than treat approval as a simple paperwork step.

That concern matters because a new site can change a town for decades. Construction may bring jobs and tax income. However, residents may also face new power lines, loud cooling equipment and pressure on roads.

“Data centers can bring investment and jobs, but developers must also pay their fair share and respect the communities hosting them.”

This is the core issue for Pennsylvania data centers: who receives the gains, and who pays the bill? The answer could shape whether projects win public support.

How much electricity could data centers need?

The US Department of Energy says data centers used about 176 terawatt-hours of electricity in 2023. A terawatt-hour equals one billion kilowatt-hours. The department expects use to reach between 325 and 580 terawatt-hours by 2028.

That range is wide because no one knows how fast AI growth will continue. Still, even the lower forecast would nearly double demand in five years. Pennsylvania data centers would add to a national race for new power supplies.

US data center electricity useTerawatt-hours per year17632558020232028 low2028 high

These figures come from the department’s energy research and grid analysis. They describe the whole country, not one state. Yet they explain why local leaders are asking who will fund new substations and transmission lines.

Who pays for the new power grid?

Utilities often recover grid costs through customer bills. That means families and small businesses could pay for upgrades built mainly for a huge industrial customer. Shapiro wants a fairer deal for Pennsylvania data centers.

One possible answer is a special connection charge. This fee would make a new project pay for the equipment needed to serve it. Another option is a long-term power contract, which promises steady payments to the utility.

The details matter. If fees are too high, companies may choose another state. If fees are too low, ordinary customers may carry the cost. Regulators must check both the price and the promise of new jobs.

Issue	Possible benefit	Main local worry
Construction	Jobs and supplier orders	Traffic and road wear
Power demand	Utility investment	Higher bills or new lines
Tax revenue	More public funds	Unclear long-term return
AI growth	New technology work	Heavy energy and water use

Pennsylvania sits inside PJM, the largest US power market. PJM coordinates electricity supplies across 13 states and Washington, DC. Its official market and planning data can help show whether the grid has enough power for new sites.

What does this mean for AI companies?

AI firms need special computer chips and large server halls. Those halls produce heat, so operators use air systems or water to cool them. A project can therefore affect both electricity supplies and local water planning.

Companies may respond by building smaller sites in several places. They may also sign deals for solar, wind or nuclear power. But a clean-energy contract does not always mean a project can run without the wider grid.

Pennsylvania data centers could still bring real economic value. Workers may build, maintain and secure the sites. Local firms may sell concrete, cables, cooling gear and repair services.

Shapiro’s message is that growth needs a social contract. In plain terms, companies should give communities a clear return for hosting them. That could include jobs, tax payments, lower grid costs or funds for schools and roads.

What should residents watch next?

Residents should ask four simple questions about each proposed site. How much power will it use? Who pays for the connection? How many permanent jobs will remain after construction? What happens if the company leaves?

They should also ask for water and noise studies. A data center may look like a quiet warehouse from outside. Inside, thousands of machines run constantly, and cooling systems can hum all day.

The state can make these answers easier to find. Public project filings should show expected power use, tax breaks, water needs and emergency plans. Clear data would help towns compare promises with likely costs.

Pennsylvania data centers are not only a technology story. They are also a test of local control, energy policy and public trust. Shapiro’s warning suggests the next projects will face tougher questions before they receive approval.

FAQs

What are Pennsylvania data centers?

They are large buildings filled with computers that store information and run digital services. They need constant electricity and cooling.

Why does Josh Shapiro oppose some data centers?

He says some developers overlook nearby communities. His concern is not the industry itself, but unfair costs and weak local input.

How much power could US data centers use?

The Energy Department estimates 325 to 580 terawatt-hours in 2028, up from 176 terawatt-hours in 2023.

Pennsylvania data centers now face a formal state test

The debate moved beyond rhetoric on 18 August 2026, when Gov. Josh Shapiro signed Executive Order 2026-05. The order directs state agencies to apply Responsible Infrastructure Development, or GRID, requirements to data-center proposals seeking state permits and incentives. The state says projects must address energy affordability, community engagement, jobs, transparency and environmental protection.

Everyone else is reporting community anger; we are explaining how that anger changes project economics. A developer that cannot document power supply, local consultation and public benefit may lose fast-track permitting or tax advantages. That shifts community acceptance from a public-relations issue into a financing and schedule risk.

The Pennsylvania governor’s office provides the primary order summary. Spotlight PA independently reported how the conditions affect permits and incentives. Inside Climate News documented resident opposition, while Pennsylvania Capital-Star reported a July survey in which 74% opposed a data center in their own community.

The numbers explain why the politics hardened

State officials have discussed reports of more than 100 proposed projects, but Shapiro said only a small number had the permits needed to operate. The gap between proposals and buildable sites matters. Speculative announcements can alarm residents and reserve land or grid capacity even when financing and customers are uncertain.

The public response is similarly uneven. One poll can produce a different result from another because wording and timing change. The safest reading is not that every voter opposes all Pennsylvania data centers. It is that support for national AI leadership does not translate into automatic approval for a large facility next door.

The Department of Energy’s national estimate—176 terawatt-hours in 2023 and 325–580 terawatt-hours in 2028—sets the scale. Pennsylvania’s projects are only part of that total, so the national figure should never be presented as the state’s consumption. It does show why grid planners cannot treat every large-load request as routine.

What a credible project must disclose

A bankable proposal needs more than a jobs headline. It should identify the power provider, peak demand, interconnection timetable, backup generation, water and cooling design, tax treatment, permanent employment and decommissioning plan. Residents also need a route to challenge assumptions before permits are final.

PJM’s queue and reliability studies are central because generation and transmission projects can take years. A developer may claim renewable energy support while still depending on the regional grid when wind or solar output is low. Matching annual clean-energy purchases does not remove the need for reliable capacity each hour.

For readers tracking the industry’s scale, our analysis of the India data-centre investment claim explains why announced capital, powered capacity and completed facilities are different measures. Our report on Lumentum’s AI optics growth shows another layer of the infrastructure supply chain.

The next approval will be the real test

The executive order does not ban Pennsylvania data centers. It creates leverage: projects seeking state help must prove they will not push unreasonable costs onto households or bypass local government. The practical test will be whether agencies publish comparable project data and enforce the same requirements across developers.

Customers and utilities also need enforceable demand forecasts. A large interconnection should not force ordinary ratepayers to finance equipment that becomes stranded if an AI tenant cancels. Deposits, minimum payments and phased construction can align the developer’s promises with the grid investments it triggers.

The clearest takeaway is that community consent has become a material project input. Developers that budget only for servers, buildings and power equipment—and not for transparency, mitigation and local benefits—now face a higher chance of delay or rejection.