Pennsylvania’s Coal Plants Were Supposed to Fade Out. AI Brought Them Back

AI doesn’t cry out. It hums—constantly, relentlessly, behind climate-controlled walls and steel doors. Its energy requirements are constant rather than sporadic or seasonal. That silent hunger is altering the electricity landscape in Pennsylvania in ways that are especially unexpected.

Many believed that coal was dying, but it has proven to be remarkably effective once more—not for heating houses, but for powering inferencing engines and neural networks.

Pennsylvania’s old gas and coal plants, many of which were already set to retire, have been extended or repurposed in recent years. Their output, which was previously intended for towns and trains, is now routed to large data centers that are used to develop and implement artificial intelligence.

Developers are selecting Pennsylvania for hyperscale centers by utilizing the state’s existing infrastructure, including old grids, permissive zoning, and sturdy infrastructure. These server racks are not your typical ones. They require output levels that only legacy plants can consistently supply, and they have constant demands for electricity around-the-clock.

At first, the town of Homer City lamented the closure of a coal plant. What many considered to be a soot-stained chapter came to an end in 2023 with its demolition. But not for very long. The same location was proposed as a high-throughput data energy hub in a matter of months. The goal of the new investors was to mine language models, not the land.

There are subtle echoes of that change throughout the region. Similar trends have surfaced in the areas surrounding Shamokin, Clearfield, and even portions of the Lehigh Valley in recent months. Permits for air discharge, water withdrawal, and fossil fuel flexibility are frequently included in projects that are wrapped in the rhetoric of “innovation.”

Opportunity is often emphasized in the public framing. And for good reason. Jobs in construction are back. Reservations for hotels increase. Coffee shops fill up. However, employment sharply declines after the center is constructed. Only thirty full-time employees may be required at a hyperscale facility.

This brings up an especially important question: Who gains in the long run?

Local leaders frequently find themselves torn between short-term economic optimism and environmental responsibility. They must approve permits that contain hundreds of pages of technical terms, such as transformer loads, stack emissions, and thermal outputs. Last fall, I went to a township meeting that lasted five hours, during which engineers discussed airflow patterns and locals merely asked if their children would be able to breathe clean air.

It’s surprising how few people doubt AI’s existence. Voice assistants, art generators, and perhaps even a more intelligent chatbot are examples of how AI is still perceived by the general public. However, the servers that run those tools need a physical footprint that is expanding more quickly than most people anticipated.

Developers are avoiding more expensive renewable transitions while accelerating deployment by incorporating older plants into these AI hubs. Refitting a mothballed coal site is significantly quicker than constructing a solar array with grid parity and battery support.

That speed is alluring. Pennsylvania is growing in popularity as a travel destination due to its legislative tone as well as its land. Bipartisan support for digital innovation is growing, and the factors driving it are being examined less closely.

Advocates for data centers have heavily relied on national security narratives in recent hearings. They maintained that in order to maintain America’s competitiveness, infrastructure must remain onshore and powered. In post-industrial communities where globalism has already had an impact, that is an especially compelling frame.

The problem is that AI is limited by geography. It penetrates it. It uses more water per megawatt hour than almost any other commercial process, mostly for cooling, and it draws from the same reservoirs and breathes the same air.

One proposed project in Clearfield County is expected to require two million gallons of water every day. That exceeds the total of the four nearby towns. The conflict between local strain and regional growth is difficult to ignore.

However, portraying this as a straightforward return to fossil fuels would be simplistic. The situation is more complex. Developers are fusing old industrial shells with digital aspirations through strategic repurposing. They’re dusting off what’s already connected and labeling it transitional, not raising coal flags.

The fact that AI is not energy-neutral is exceptionally evident. It does not survive on remote cloud servers or clever code alone. It is firmly rooted in the politics of energy, pipes, and plants.

But there is still hope. Numerous researchers think AI could speed up the adoption of clean energy, balance grids, and optimize power flows. Particularly alluring—and possibly true—is the notion that AI may eventually reduce emissions more than it produces.

However, when particulate matter increases this quarter, “eventually” doesn’t help.

Right now, the state has an opportunity to create a narrative about infrastructure that is more sustainable. Transparent reporting requirements, mandatory renewable offsets, and more stringent permitting for AI-linked energy draws would all result from this. Additionally, it would entail posing more challenging queries during approval, such as how much electricity is being used for community purposes versus servers? Which options were modeled?

Pennsylvania could present itself as a clean digital corridor instead of a covert machine learning refinery with targeted incentives and progressive legislation.

There is no denying that AI is a transformative force. However, oversight shouldn’t be flattened by its trajectory. As more AI companies physically enter towns that never requested to host them, the discourse must change.

I’ve talked to bar owners, teachers, and employees in these towns. Many are cautiously optimistic. However, they also recall what it was like when coal and steel first left. They have earned their caution.

AI is no longer a vague concept. It is shaped. Demand exists for it. Additionally, Pennsylvania has powerlines that go straight to locations that many people believed were already in the past.

Pennsylvania has the potential to become a national model if it can effectively navigate this moment by striking a balance between digital ambition and sustainable grounding. Not only to prepare for AI, but also to figure out how to develop technological futures without accumulating environmental debt.

It’s more than just sound policy. For any state looking to expand without erasing what it already knows, it’s a very adaptable approach.