Artificial intelligence is driving a wave of massive new data centers across the U.S., raising concerns among communities and environmental groups about their demands for electricity and water, and their impact on nearby infrastructure and ecosystems.
But Daniel Cohan, a professor of civil and environmental engineering at Rice University, told Newsweek that the environmental impact can vary significantly depending on three variables: the amount of electricity a facility requires; the type of electricity generated—whether by natural gas, coal, nuclear power, wind or solar; and how quickly the projects are being built.
“The effect really comes down to three things,” Cohan told Newsweek. “It’s the total, the type and the timing.”
Read More on News
Timing is emerging as a key concern as the rapid pace of construction of some of the country’s largest AI data centers creates challenges the U.S. power system has rarely faced at this scale.
“So we’ve just never seen such large facilities come onto the grid at once,” he said. “That’s not all data centers by any means, but the largest ones being built are unprecedented in how much demand they have.”
That does not necessarily mean data centers cannot be built without creating environmental or reliability problems, Cohan said. Rather, he said, the way they are being built matters.
“If we had a staggered build-out of these data centers with plenty of planning for how we’re going to incorporate them on the grid, and if we were taking the steps that we need to be doing anyway to have a more robust grid nationally, … we could accommodate this,” he said.
The concern, he said, is what happens when enormous amounts of electricity demand are added in concentrated locations at high speed, before the grid and surrounding infrastructure have had time to adapt.
“But when you have so much being added in such concentrated locations so quickly, that’s when you run the risk of disruption, and run the risk of raising the costs and jeopardizing the reliability of power for people,” Cohan said.
Are Data Centers Bad for the Environment?
Data centers are not inherently environmentally damaging, Cohan said. Their impacts can vary substantially depending on how they are powered, cooled and operated—and where they are built.
“A data-center build-out done right could be a catalyst for building out a more robust grid powered by more renewables and battery storage,” Cohan said. “But done wrongly, it can lead to more inefficient and dirtier gas plants and diesel generators.”
That distinction is increasingly important as AI companies seek access to enormous amounts of electricity.
Cohan said data centers could, in theory, help accelerate investment in new generation and transmission infrastructure. More electricity demand can also mean more customers sharing the costs of maintaining the grid.
“Having more customers and more demand—whether it’s data centers, electric cars, electric heat pumps or electrifying our factories—if the policy and the management and planning is done right, having more customers can spread out the costs over more people and could bring down the costs,” he said.
“Again, that’s a big if. I’m not saying this is how it’s being done right now,” he continued. ‘The problem, he said, is the speed at which some projects are moving.
“But when it’s been done in this slapdash manner, and everyone is in this mad dash … you don’t have the proper planning,” Cohan said.
The construction of large data centers can also affect endangered and threatened species when projects require clearing forests, fragmenting habitat or building new roads and transmission infrastructure. In areas where development overlaps with sensitive habitats, those changes can add pressure to species that are already vulnerable.
It is also worth putting the environmental footprint of data centers in context. Their impacts can be significant at the local level, particularly where large facilities place new demands on electricity, water and infrastructure. But data centers are not the only major source of environmental pressure: commercial aviation, beef production, fast fashion and cryptocurrency mining all carry substantial environmental footprints of their own. The significance of a data center’s impact ultimately depends on what is being measured—and where.
What to Know About Data Center Water Usage
Water is another part of the environmental equation, although its importance can vary dramatically from one facility to another.
Yu-Feng F. Lin, a principal research hydrogeologist at the Illinois State Geological Survey, director of the Illinois Water Resources Center and a professor at the University of Illinois Urbana-Champaign, said data centers primarily use water as part of their cooling systems.
The computing chips inside AI servers generate significant heat, Lin said, and cooling systems have to remove that heat to prevent damage to the equipment.
“The water usage is mainly for cooling because their computing chips (GPU and CPU) emit a lot of heat,” Lin said. “Those expensive chips will be permanently damaged if they are overheated.”
But focusing only on water used directly at the data center can miss part of the picture.
Lin said operators may reduce direct water use by relying more heavily on electricity for cooling, but generating that additional electricity can itself require water.
“The electricity usage for cooling could be between 10 percent-40%, while more electricity usage might require less ‘direct’ water usage for cooling,” Lin said. “However, there might be more ‘indirect’ water usage for generating more electricity for cooling.”
“That’s why ‘Water-Energy Nexus’ is a complicated balance for data center design and operation,” he said.
Cohan similarly cautioned against treating data-center water consumption as a single national figure.
“If the power generation is coming from fossil fuels or nuclear plants, then those plants use enormous amounts of water to generate electricity,” Cohan said. “If it’s coming from wind, solar and batteries, then you may have almost no water use by wind or solar.”
“So whether it affects water is really a matter of how the cooling operates and how the data center gets its power,” he said.
Lin said the significance of water use ultimately depends heavily on local conditions.
“It won’t be easy to provide a general formula or a universal answer for this question because water stress usually is a very localized issue,” Lin said.
That means a data center’s water footprint can look very different depending on the local watershed, climate, cooling system and electricity source.
Lin also pointed to another challenge: facilities may eventually consume substantially more water and electricity than initially projected as their computing capacity is upgraded.
“One unique challenge from data centers is that their water and electricity usage might change significantly when they do a major upgrade of their computing capacity from the original design,” Lin said.
Why the US Is Building So Many Data Centers
The Trump administration has framed the rapid expansion of AI infrastructure as part of a broader competition with China. Trump has repeatedly argued that the United States cannot afford to slow the technology’s development.
“If we don’t win AI, we’re gonna be put in a very bad position,” Trump said in September. “We are leading China right now by a pretty good period.”
“I’m not going to stifle growth of something that will be bigger than the Industrial Revolution.”
That urgency is part of the backdrop to the extraordinary pace of data-center construction now underway across the country.
AI models require large quantities of computing power, particularly during training and increasingly as companies deploy AI services at scale. That demand has prompted technology companies and data-center developers to seek sites where they can secure large amounts of electricity, land and supporting infrastructure.
But the scale of some projects is fundamentally different from traditional industrial electricity users, Cohan said.
“There is no other [industry] that would consider trying multiple gigawatts of power at a time,” he said.
“A gigawatt is the output of a nuclear reactor or it’s the demand of a small-sized city,” he said. “Other factories, other industries are typically—even the biggest ones—might have a tenth of the demand of a large data center.”
The concentration of demand can create another unusual challenge for grid operators.
Cohan said a very large data center could potentially reduce its electricity demand almost instantaneously if it experiences a technical problem or switches to backup generation.
“With a tiny disruption, in a fraction of a second, all of that massive amount of power demand can disappear,” he said.
“It’s like all of a sudden if everyone in Austin turned off their lights and all their equipment at once.”
“Just like the grid can’t handle an instantaneous surge in demand, it’s actually extremely hard to handle an instantaneous falloff of demand,” Cohan said.
“At the biggest data-center campuses that are being built, you’ve got the equivalent of a few nuclear reactors’ worth of demand that could go off and on in a fraction of a second. That poses a real operational challenge.”
Cohan said greater flexibility in when data centers use electricity could help reduce some of those challenges.
“The part that gets neglected, that could have the biggest difference, is whether the timing could be flexible,” he said.
“In the way that it’s been done so far, with such rushed processes, we’re not realizing those opportunities that we had if we could be more flexible with our demand.”
Could AI Data Centers Mean More Fossil Fuels?
One of the biggest environmental questions surrounding the data-center boom is how the additional electricity will be generated.
Cohan said rapidly increasing demand could contribute to additional fossil-fuel generation, particularly when projects cannot quickly obtain enough power from the existing grid. “It could because of the flawed policies that we have in this country right now,” Cohan said when asked whether the data-center boom could lead to more fossil-fuel generation.
He argued that wind, solar and battery storage can be built quickly and used together to provide electricity as demand changes. But Cohan said the rush to provide power to some data centers can create incentives for developers to pursue natural-gas generation, including facilities built behind the meter rather than connecting to the broader grid.
“A lot of the bottlenecks is data centers trying to get permission from the grid,” Cohan said. “And so they say, ‘Oh, well, we can’t get permission. We’ll just build our own power source behind the meter or off the grid’ to supply our data center on site.”
“When they do that they’re almost always doing that with inefficient gas turbines and backing them up, in some cases, with diesel generators,” he said.
Cohan said that can make the environmental impact of the additional electricity demand worse than it would be if projects were planned alongside broader grid upgrades. “So even the power that’s coming from natural gas is coming in a dirtier and less efficient way than if things were being planned properly,” he said.
His broader point is that the data-center boom could produce very different environmental outcomes depending on the policies and infrastructure surrounding it.
“A data-center build-out done right could be a catalyst for building out a more robust grid powered by more renewables and battery storage,” Cohan said. “But done wrongly, it can lead to more inefficient and dirtier gas plants and diesel generators.”
Location Matters
The environmental impact of a data center can also vary considerably depending on where it is built. Cohan pointed to Texas and Virginia as major centers of data-center development and said the electricity grid serving a particular location can make a significant difference.
The biggest hubs are concentrated in a handful of states. As of early 2026, Virginia, Texas and Georgia had the largest numbers of planned data centers, followed by Illinois and Arizona, according to Pew Research Center. Much of the planned construction is concentrated in the South and Midwest.
At the same time, some states and local governments are pushing back against new development. New York lawmakers passed legislation calling for a one-year moratorium on permits for certain hyperscale data centers, while communities in states including Ohio, Georgia and North Carolina have pursued their own restrictions or moratoriums.
Cohan said that Electric Reliability Council of Texas (ERCOT), the independent system operator that manages Texas’ main electrical grid, has a “more streamlined approach for bringing new power generation online, faster than almost any other U.S. grid.”
Texas has rapidly added wind, solar and battery storage, while also relying on natural gas and other sources of generation.
“Put all that together, and if you have data centers being added in Texas, the Texas grid is able to integrate them more easily than other parts of the country,” Cohan said.
That does not mean every location in Texas has the same environmental profile. Water availability, extreme heat, transmission capacity and the local generation mix can all affect the consequences of a particular project.
Lin likewise emphasized the importance of local conditions, particularly for water.
“Water stress usually is a very localized issue,” he said.
Why Data Centers Are So Loud: What Communities Need to Know
The environmental footprint of a data center extends beyond electricity and water.
Large facilities can bring industrial-scale equipment, backup generators, cooling systems, substations and other infrastructure into communities.
Noise can come from cooling equipment, mechanical systems and backup power equipment, while the construction phase can bring additional traffic, heavy equipment and temporary disruption.
For communities near a proposed facility, those impacts can be particularly noticeable because data centers operate around the clock. Unlike a typical commercial building, a large data center may have cooling and mechanical equipment running continuously to maintain the conditions required for thousands of servers.
Who Pays for the Power?
The environmental and grid impacts of data centers also raise a question for communities: Who pays for the new infrastructure needed to power them?
The House passed the bipartisan Ratepayer Protection Act on September 16 by a 417-3 vote. The legislation would require state utility regulators to consider standards ensuring that large-load customers cover the incremental costs of generation, transmission and distribution upgrades needed to serve them.
The bill specifically covers non-residential customers with peak demand of at least 100 megawatts at a single site or campus—a threshold that can capture large data-center developments. That matters because bringing a massive new electricity customer onto the grid can require more than simply connecting a new building. Utilities may need to add power generation, build or upgrade transmission lines and make local distribution improvements to accommodate the additional demand.
Under the legislation, state regulators would consider requiring those large-load customers to recover the full incremental costs associated with those upgrades, as well as provide financial assurances to cover them. The bill would amend the Public Utility Regulatory Policies Act, known as PURPA, while leaving decisions about how to implement the standards with state regulators.
Supporters of the legislation have framed it as a way to prevent households and small businesses from subsidizing the infrastructure needed for large new electricity users.
The Senate is expected to take up the legislation this week.
For Cohan, making sure data centers pay their share is part of the larger question of whether the AI build-out can strengthen the grid without shifting its costs onto other consumers.
“It’s not that big data centers couldn’t be done,” Cohan said. “It takes a lot of planning. It takes a lot of infrastructure development. It takes policies to get things right so that the data centers are paying their fair share, and the costs don’t go up for other consumers.”
What Communities Should Ask Before Approving a Data Center
For residents and local officials, Cohan said the most useful framework is his “three T’s”: total, type and timing.
“The total amount of power this is going to be, and how does that compare to what’s needed on our part of the grid so far?” he said.
Then comes the type of power. “Is this going to be done in clean ways, or is this going to add new fossil generation?” Cohan said.
And finally, the timing. “Is this being designed in a way that can flex its power demand so that it’s not drawing power [during tight] conditions, so that it’s not driving up power prices, and so that it’s not risking a brownout on our grid?” he said.
Lin said residents should also ask developers for detailed information about both direct and indirect water consumption.
“The water usage for the required electricity will be critical because there might be significant ‘indirect’ water usage for the power generation,” Lin said.
Communities should also ask how a facility’s electricity and water demand could change over time as computing equipment is upgraded, Lin said.
For Cohan, those questions are ultimately about whether a project is being designed to work with—not simply draw from—the surrounding infrastructure.
“It’s not that big data centers couldn’t be done,” Cohan said. “It takes a lot of planning. It takes a lot of infrastructure development. It takes policies to get things right so that the data centers are paying their fair share, and the costs don’t go up for other consumers.”
The environmental footprint of an AI data center, in other words, may be determined not only by the computers inside it, but by the power plants, transmission lines, cooling systems and local resources needed to keep those computers running.
Newsweek’s reporters and editors used Martyn, our AI assistant, to produce this story. Learn more about Martyn here. Contact Newsweek editors on this story: Edward T. Cummins