The environmental cost of a data center is becoming harder to define precisely as the infrastructure behind artificial intelligence grows more complex. A server building may occupy a relatively contained footprint, yet the electricity, water, backup generation and transmission capacity required to keep it running can stretch across an entire region. Ted Smith, an environmental health researcher at the University of Louisville, says the answer changes substantially depending on where a facility is built, how it cools its equipment and how it obtains electricity. “Everything is moving so fast, we’re asking people to work without the evidence. And that can be really uncomfortable. I don’t know what the health effects are of living next a hyperscale data center. Unfortunately, I don’t know anyone who knows the effects of living next to a hyperscale data center,” he says.
Ohio has more than 200 data centers, making it the fifth-highest state in the country, according to the Ohio Consumers’ Counsel. Most of Ohio’s data centers are concentrated in Central Ohio, according to the Ohio Consumers’ Counsel, while the supplied regional data for Kentucky and Indiana does not provide a sufficiently current authoritative count to support the specific figures of about 60 and 125 facilities. Those totals tell only part of the story because a modest enterprise facility and a hyperscale AI campus can have radically different demands on electricity, cooling, water and land. The distinction becomes critical when projects reach hundreds of megawatts or potentially several gigawatts, because their resource requirements begin to resemble those of entire communities. The environmental debate is therefore shifting away from whether data centers consume resources toward a more difficult question: which resources, where, for how long and at whose expense?
AI Is Turning Infrastructure Into the New Constraint
For years, the data center industry could treat electricity as an enabling service that followed construction rather than as the central limitation on growth. That assumption is weakening as AI workloads require dense clusters of specialized processors and operators seek increasingly large blocks of continuous power. Dan Diorio, vice president for state policy at the Data Center Coalition, told Ohio lawmakers, “We are likely still at the early stages,” while adding, “There is still significant development to go as we see continuous projects seek to develop in the state.” His comments point to a market where demand for computing capacity is beginning to reshape decisions normally associated with utilities and industrial development. Credit-card processing, cloud storage, geolocation services and AI all contribute to that demand, but AI is pushing the economics toward larger and more power-intensive facilities.
The numbers illustrate why the industry has become an energy story rather than simply a technology story. Data centers across the U.S. consumed about 176 terawatt-hours of electricity in 2023, roughly equivalent to Ohio’s total electricity use that year and about 4% of U.S. electricity consumption. That share is expected to grow to 9% by 2030, according to the Ohio Consumers’ Counsel. PJM Interconnection, which coordinates wholesale electricity markets and transmission across 13 states and the District of Columbia, has identified data center expansion as a major driver of rising demand. Asim Haque, PJM’s executive vice president of governmental and member services, described the increase as a “generational increase in demand” when speaking to Ohio lawmakers. The significance is not simply that data centers use more power, but that they can require enormous amounts of electricity continuously, creating a very different planning challenge from ordinary commercial demand.
One Campus Can Resemble a City on the Grid
The most revealing comparison may come from the size of individual projects rather than the number of facilities. Haque told Ohio lawmakers, “Hyperscalers can range from 50 megawatts to 10 gigawatts down in Portsmouth,” referring to the proposed Pike County development. “A hyperscaler that is 50 or 100 megawatts, you’re looking at the size of Mansfield. A gigawatt, if you go to the New Albany data center campus for Meta, that is the size of Cleveland in terms of their consumption.” Those comparisons expose how quickly the definition of an industrial customer is changing as hyperscale development expands. A 50-megawatt facility already represents a substantial electricity load, while a gigawatt-class campus can transform the economics of local generation, transmission and capacity planning. A 10-gigawatt development pushes that question into another category altogether, requiring planners to think about energy supply on a scale traditionally associated with major metropolitan systems.
PJM’s long-term outlook points to a substantial increase in peak electricity demand as large-load customers, including data centers, expand across the region. PJM has described a major expansion in regional electricity demand, with its long-term planning pointing to peak demand reaching about 220,000 megawatts over the next 15 years. PJM expects peak demand to reach about 220,000 megawatts over the next 15 years, a level that would put substantially more pressure on generation and transmission infrastructure. Diorio has described data centers as a major contributor to the increase in electricity demand across the PJM footprint. Haque has emphasized that reliability and consumer prices represent the central challenges created by the growth rather than electricity availability alone. The environmental question follows directly from those concerns because every additional megawatt must ultimately come from some combination of generation technologies with different emissions and resource profiles.
The Power Source Matters More Than the Server
A facility connected to a predominantly renewable grid will carry a substantially different operational emissions profile from one supplied by coal or natural gas. Ohio currently relies heavily on generation that produces greenhouse-gas emissions, making the state’s power mix particularly important as computing demand accelerates. Jeff Bielicki, associate research director at Ohio State University’s Sustainability Institute, describes the issue in terms of the broader economy rather than the data center alone. “The emissions from the power plants and generators providing electricity are not inconsequential,” Bielicki says. “They have known effects on the environment, known effects on human health. That’s well understood, from local, regional, national to international communities.” His point changes the unit of analysis because the environmental footprint of computing follows the electrons that power it, regardless of where those electrons originate.
Haque told lawmakers that PJM was evaluating additional generation resources as Ohio and the wider region respond to rising electricity demand from large-load customers. Yet proposed generation does not automatically become operating generation, making the eventual mix of resources difficult to predict from project queues alone. The uncertainty around which proposed projects ultimately reach operation makes it difficult to know exactly what the future power mix will look like. The uncertainty creates a direct connection between data center development and broader energy policy because the environmental profile of new computing capacity depends on which projects actually reach commercial operation. “In Ohio, about 80% of the way we generate electricity for our economy is undermining the environment upon which our well-being depends,” Bielicki says. “Well-being is not just economic well-being. It’s also health and environmental well-being.”
The Real Environmental Metric Is Still Missing
What emerges from the data is not a simple argument for or against data centers, but a case for measuring infrastructure differently. The conventional approach asks how much electricity or water a building consumes, yet that method can miss the resources consumed upstream to generate its electricity or treat and replace its water. It can also overlook the infrastructure required to deliver those resources, including power plants, transmission lines, substations, pipelines and wastewater systems. A facility can therefore improve its internal efficiency while shifting environmental demand somewhere else in the system. That makes site-specific accounting more important than industry-wide averages as hyperscale projects become larger. The question is no longer simply how efficient a data center is, but how efficiently the entire network supporting it operates.
Meanwhile, the computing industry has another possibility that is unusual among major infrastructure sectors: it can use the same computing capacity driving its environmental footprint to study those effects. Bielicki argues that AI-enabled computing could accelerate research into energy technologies, new materials and human health. “When we think about data centers, we need to try to find a middle road, and increasingly put more effort into using the capacity of the AI enabled by the data centers to help address environmental and human health problems, [such as] new materials discovery, new innovation on energy technologies, new understandings of pathways of respiratory disease, and so on,” he said. The opportunity also changes the conversation from measuring damage after construction to designing facilities around measurable environmental performance from the beginning.
The Next Question Is Where the Burden Lands
The most important environmental distinction between data centers may ultimately be geography rather than technology. A facility built beside abundant low-carbon electricity and a resilient water supply presents a different set of pressures from one drawing power from fossil-heavy generation and water from a stressed aquifer. The same cooling technology can produce different outcomes depending on local climate, electricity sources and wastewater infrastructure. Sarah Hippensteel Hall has emphasized that communities need to consider those conditions before approving major facilities. “Understanding local conditions is extremely important,” Hall said. “A one-size-fits-all approach across Ohio will not fully reflect the differences between sites, watersheds or water resources. We see what happens other places when water is an afterthought.” Her warning captures the central weakness of treating data center development as a standardized infrastructure category.
