PA Environment Digest Blog: Post-Gazette Guest Essay: Pennsylvania May Not Have Enough Electricity For A.I. Data Centers And The Rest Of Us

By Shaliee Kishore and Arindam Banerjee, Lehigh University

This guest essay first appeared in the Pittsburgh Post-Gazette on August 8, 2025--

With July and August come the dog days of summer, when electricity demand surges. The demand for cooling adds to the “business as usual” electricity needed to keep our economy humming.

In Pennsylvania, this increasingly means electricity needed to power Artificial Intelligence (A.I.) data centers. And that demand is only going to grow. What that will means for electricity customers is unclear.

A.I .power hogs

The state currently ranks 13th in its number of data centers. With recent regulation changes and incoming commitments, the growth may soon take off.

Amazon recently announced a $20 billion investment for two new large data centers.

And $90 billion of private sector investment was announced at the Pennsylvania Energy & Innovation Summit.

The companies promised job creation, workforce development and improved economic development.

But we don’t how this growth will affect consumers.

That’s because these new data centers will almost surely be used to support A.I. services, and A.I. data centers are power hogs.

Consider the rapidly growing number of A.I. tools available. These tools rely on large amounts of data to train complex A.I. models that respond to queries sent by users.

All of these processes run on specialized data servers housed in large data centers that require a significant amount of electricity.

Generating Lots Of Heat

This computing hardware generates a lot of heat.

Mechanisms for cooling data centers contribute to water use and impact and significantly add to their electricity demand.

A hyperscale data center, like those operated by major tech companies, can consume over 500,000 gallons of water per day.

Larger A.I. data centers need up to 100s of megawatts of power each. The largest currently each require a couple of gigawatts of power.

For comparison, the total capacity of the Susquehanna nuclear plant, one of the largest in the nation, is about 2.5 gigawatts.

This is the same amount of energy that can power between 2.5 and 5 million homes, not counting peak power usage.

Increasing Power Demand

As the capabilities and markets for A.I. grow, more A.I. data centers will be built and they will in turn demand more power from local and regional electricity systems.

Current estimates indicate that in the next three years, the electricity needed for A.I. data centers could be close to 12% of the total electricity demand in the U.S. That’s up from about 4% currently.

In some states, it’s even greater. In Virginia, for example, more than 25% of the electricity demand is from data centers.

And the size of the demand is not the only challenge for local electricity grids. A.I. data centers require uninterrupted power and constant cooling.

The power consumption of their servers can drastically vary over the course of minutes. This variation affects power quality for other electricity consumers.

Affordability & Reliability

Here’s the question for consumers. How affordable and reliable electricity will be after all these data centers come in?

In some cases, data centers near a large power plant would like to have some of the generated power diverted directly for their own use or have the generated power be credited in other locations.

The data center company usually contracts to pay a higher rate for the electricity to ensure electricity supply, making it more profitable for the power plant owners to enter into such agreements, as compared to supplying power more broadly.

These agreements raise questions about their implications for those outside major tech companies.

Normal, Necessary Demands

How will our normal, necessary demands be met if significant nuclear power generation is being diverted to one large-scale customer?

In the dog days of summer, for example, when people need cooling to stay healthy and productive, and even survive?

The electricity system will likely have to be upgraded to both incorporate power delivery to these large-scale customers and to ensure reliable power for the rest of us.

Who pays for these upgrades?

The long-standing answer has been everyone.

Regulators and policymakers will have to iron out details soon to accommodate the anticipated growth in data centers and to have buy-in from the rest of us.

Complications and clever solutions

The situation’s complicated. Experts on policy solutions must work closely with experts on A.I. algorithms, hardware design, electricity systems and markets, communications, water and building systems to reduce the energy and water use of A.I.

Sustained support is needed from federal agencies and industry, particularly for our state’s universities that innovate while training the nation’s future talents and reshaping the workforce.

Without clever solutions to pressing problems, a simple ChatGPT request during the dog hours of the dog days of summer might be too much for the electricity grid to handle.

Shalinee Kishore, professor of electrical and computer engineering, and Arindam Banerjee, professor of mechanical engineering and mechanics, both at Lehigh University, lead the university’s Center for Advancing Community Electrification Solutions.

Resource Links - PA Electric Grid:

-- PJM Electric Auction Impacts: 1 In 5 PA Households Report Problems Now Paying Energy Bills; Electric Utility Shutoffs Up 38.1% So Far This Year [PaEN]

-- PJM Electricity Auction: PJM Lost 2.8 Gigawatts Of Power Due To Reduced Reliability Rating Of Natural Gas Power Plants; Could Gain 12.2 Gigawatts By Increasing Reliability From Less Than 75% Now To An Achievable 90% [PaEN]

-- Guest Essay: Paying The Price For Natural Gas And A.I. Data Centers - By John Quigley, Senior Fellow, Kleinman Center for Energy Policy [PaEN]

-- 30 Stakeholder Comments Received By PUC On Adequacy Of Electricity Supplies In Pennsylvania; Increasing Natural Gas Power Plant Reliability To 90-95% Would Mean No Imminent Capacity Problem [PaEN]

-- PUC Invites Stakeholder Comments On The Issue Of The Adequacy Of Electricity Supplies In Pennsylvania [Background On Issue] [PaEN]

-- Guest Essay: Rewriting The Energy Story — Together - By Stephen M. DeFrank, Chairman, Pennsylvania Public Utility Commission [PaEN]

-- PUC: Brace For Higher Electric Bills As A Result Of Soaring Power Use; Consumers Should Review Energy Options [PaEN]

-- Guest Essay: Three Proposals To Protect Consumers From A.I. Price Increases: Bring Your Own Generation; Data Centers Must Pay Their Fair Share; An 'All Of The Above' Energy Future, Including Renewables - By Kevin Walker, CEO, Duquesne Light [PaEN]

-- Spotlight PA: Why Pennsylvanians May See Higher Electric Bills This Summer And Next [PJM Auction, A.I. Data Center Power Demands]

-- House Committee Told PA Needs An Energy Policy, Doing Nothing Means We Are Going To Run Out Of Power; Overreliance On One Fuel-- Gas-- Is Raising Electricity Prices [PaEN]

-- PA Elected Officials, Consumer, Clean Energy Advocates Demand Electric Grid Operator PJM Act To Hold Down Electricity Costs [PaEN]