Data centers are fueling a race to develop artificial intelligence tools that have the potential to revolutionize almost every sector of society, and Big Tech is receiving bipartisan support at nearly every level of government to aid in this global competition for AI dominance.
But these building projects that have become a unifying force for politicians are also galvanizing communities that object to seeing one of these behemoth campuses of computer servers pop up in their backyards.
RELATED: Pa. and N.J. want data centers to guarantee their own power supply
Last month in Plymouth Township, the zoning board held a public hearing to discuss a proposal to build a 2 million square-foot data center at 900 Conshohocken Road, transforming 10 buildings at the former Cleveland-Cliffs steel mill that sits on 66 acres of land along the Schuylkill River into a high-tech network.
When Montgomery County residents learned about the plan to build this facility near their homes, they began fighting back – citing concerns over the project's potential impacts on the environment and their energy bills.
"This is like a David versus Goliath fight, and we're trying to get everyone mobilized and understand this," Patti Smith, who organized a petition that received over 1,000 signatures, said before the hearing.
Then on Nov. 18, residents packed the room to voice their opinions at the zoning hearing, only to find out that the developer had to withdraw the application over a legal issue because the agreement of sale of the property was not finalized. While this appears to be a victory for the community activists, they anticipate the developer will refile the application and are concerned this move only allows more time for an ordinance to be drafted that would give the company a stronger legal standing when the proposal is revisited.
In neighboring Chester County, another data center that's being proposed at the former Pennhurst State School and Hospital site is also receiving strong opposition from the community, with a petition that's racked up over 12,000 signatures. That proposal could also stall as opposition is pushing for an ordinance to tighten regulations.
While these neighbors are standing up against Big Tech, they are also realizing they're confronting resource-rich developers and a government that in many ways is working to ease restrictions as the country tries to meet the insatiable energy demands created by the growing use of AI platforms like ChatGPT.
In June, Pennsylvania Gov. Josh Shapiro (D) announced that Amazon will spend at least $20 billion to develop cloud computing and AI innovation campuses in the state. The investment, which the company says will create at least 1,250 tech jobs, will start by building facilities in Falls Township in Bucks County and Salem Township in Luzerne County before exploring other communities for future sites.
Shortly after that investment was touted by state officials, President Donald Trump signed an executive order to accelerate federal permitting of data center development.
Then last month, governors from Pennsylvania, New Jersey, Maryland and Virginia (three Democrats and one Republican) submitted a joint proposal that would fast-track approvals for data center projects in which developers agree to build energy lines or substations and generate at least enough electricity to offset how much power the facility is expected use.
On Nov. 19, the U.S. Department of Energy approved a $1 billion loan to restart a nuclear reactor at Three Mile Island near Harrisburg to power Microsoft data centers. The renamed Crane Clean Energy Center is set to be back online in 2027, and Energy Secretary Chris Wright said the federal loan will help keep utility costs down for more than 65 million people in the 13-state territory that relies on the regional grid operated by PJM Interconnection.
"Everybody wants to be the biggest and best adopter of AI technology and make the most money for their state, but it's very short-sighted," said Genevieve Boland, one of the organizers of the Cleveland-Cliffs petition. "So the direction is coming from the federal government to essentially expedite data center approvals everywhere across the country and that has been trickled down to the state. Shapiro made a deal with Amazon and that was a huge deal. And then his mandate has trickled down to regional and local zoning boards to do whatever to approve these data centers."
Impact on electric bills
While the AI boom is putting pressure on existing electricity grids and causing a nearly unquenchable thirst for more energy resources, it remains to be seen how these demands will affect the bills of residential customers.
The U.S. Department of Energy reported last year that data centers consumed about 4.4% of the total U.S. electricity in 2023, and it predicts that will rise to 12% by 2028.
And the increase in the number of these facilities has coincided with a rise in costs. In Pennsylvania, the average electric bill has gone from $134.85 in 2022 to $160.57 this year, according to data from the U.S. Energy Information Administration.
But data centers might not be completely to blame for the higher bills, said Geoffrey Blanford, a principal technical executive at the Electric Power Research Institute, an independent nonprofit that does research for the nation's electric power industry.
"Most of the increase has been related to economy-wide inflation trends," he said. "Where it's been faster than inflation has been in places like California, which had a really steep increase that's mostly related to wildfire impacts. … Other places like New England, you've also seen higher prices. But again, that's more around (transmission and distribution) costs, grid modernization and resilience more so than data centers. So there are a lot of factors. Data centers are potentially one of them but really has not been the main one for what we've seen recently."
Researchers at Lawrence Berkeley National Laboratory found that the average retail price of electricity in Pennsylvania over the past six years has risen 7.4% faster than inflation, the New York Times reported. In New Jersey, it's up 6.1%. But it's actually down in other parts of the country — including much of the Southeast, Midwest and Mountain regions.
In addition to the quantity of data centers, the speed at which they are being built is also causing a power problem. Despite their size, these massive campuses can be built within a year or two in many cases. And the projects often require investments to improve the local grid, like adding new sources to generate energy or enhancing existing systems.
"They can be built pretty quickly, which means it puts a lot of pressure on the timeline for adding grid resources, which tend to be a little bit slower than how fast you can add a data center," Blanford said. "So it's been a major issue for power system planning."
Tech companies have now become both energy consumers and producers, and whether individuals and small businesses see a jump in their bills might be dependent on state regulators and lawmakers making sure those firms cover the expenses.
"In theory, there are potential benefits to the electric system from having a large load there," Blanford said. "… It can help with reliability, and it's not necessarily something to take for granted that it's going to increase electricity prices. In fact, it could decrease electricity prices by having more energy to spread the fixed costs over."
Another factor that could impact utility bills is the Trump administration's stand against renewable energy projects, said Sorelle Friedler, a professor of computer science at Haverford College who served as the assistant director for data and democracy under the Biden administration. The Interior Department's policies are causing permitting delays and the agency has been reviewing previously approved wind and solar farms.
"So we're actually, as a society right now, reducing the amount of energy that we could be generating while increasing the usage of it," Friedler said. "So I don't think that's going to be good for any of our energy bills."
To try to protect consumers, some politicians are stepping in. In New Jersey, Gov.-elect Mikie Sherrill campaigned on halting rising utility costs, promising to freeze rates and declare a "state of emergency on utility costs" on her first day in office.
"My priority is relief to New Jersey consumers, and I will bring everyone to the table to deliver it," she said in a statement.
Environmental impact
The AI boom is only a few years old so grasping the environmental impact can be challenging, but a study by Cornell University researchers published last month in the journal Nature Sustainability set out to quantify the industry's ecological footprint – and their estimations are staggering.
Depending on the speed of AI growth, researchers said U.S. data centers could eventually consume as much as 10 million Americans and emit as much carbon dioxide as 10 million cars per year, which would be equivalent of the annual resource consumption of the entire state of New York.
The Cleveland-Cliffs project in Plymouth Township calls for electricity generated by use natural gas turbines, which is cleaner than burning coal but still emits pollutants such as nitrogen oxides, carbon monoxides, formaldehyde and benzene. The Environmental Protection Agency warns that exposure to these types of emissions can have serious health effects, even though power plants have made great strides in reducing them over the past couple decades.
"As soon as we heard natural gas turbines … we became incredibly alarmed," said Smith, the organizer of the petition against the project. "… My preferred exposure to formaldehyde is honestly zero per year."
The cooling system needed to prevent servers from overheating also has an environmental impact. Google said in 2021 that its average data center consumed around 450,000 gallons of water per day, which it said was the "same amount of water used to irrigate 17 acres of turf lawn grass once."
In their study, the Cornell scientists recommend holding tech companies to similar standards put on vehicle manufacturers to mandate fuel efficiency – with tighter regulations, more transparency and benchmarks for energy usage.
"We're still early in this growth, it's in our hands now," Fengqi You, who authored the study, said recently. "It's not too late yet to do the planning and accounting for resource constraints that could let AI continue to grow."