Despite the prevalence and importance of data centers in modern life, the subject of data centers has never been more polarizing. Gallup reports that “7 in 10 Americans oppose constructing data centers for AI in their local area”, with women leading the opposition. Members of Congress have proposed a nationwide halt to data centers. The city of Monterey Park, California, has permanently banned data centers in city limits. Hubbard, Ohio, passed a one-year moratorium on data centers because, as one elected official put it, “we don’t know just anything about data centers, to tell you the truth.” Protests are breaking out in major cities. Americans are questioning whether data centers are contributing to nausea, headaches, and insomnia. Environmental concerns are being raised.

This Policy Circle Brief explains what data centers are, why communities have hosted them quietly for decades, how communities can thoughtfully engage with data center developers, and how to create a win-win scenario for the public.

Data center construction is one of the most talked about news stories today, but these facilities are not new to the U.S. According to Data Center Map, there are more than 4,600 data centers already up-and-running in the U.S.

While data centers are perceived to be a recent phenomenon, the concept is nearly eight decades old. Data centers are considered critical infrastructures on which our national security depends. 

In 1946, the first fully-functioning computer was unveiled at the University of Pennsylvania. Dubbed “ENIAC,” this device weighed 30 tons and occupied about 1,800 square feet. It required enormous power, complex cooling systems, and highly secure environments – the same challenges faced today by data centers. ENIAC has been called the earliest ancestor of the data center because it posed the same questions as data center operators face today: How do you power, cool, and protect a machine that can never turn off?

In the decade that followed, similar installations went up at West Point, the Pentagon, and CIA headquarters. Because these early machines ran defense and intelligence work, physical security and secrecy were built in from the start. By the 1960s, computing systems had grown enormous, filling dedicated rooms that needed their own specialized power and cooling. Yet almost no one interacted with them directly. The cost of building and running such facilities put them out of reach for all but governments, universities, and the largest corporations.

One of the most influential early commercial applications was the Semi-Automated Business Research Environment, or SABRE, developed jointly by American Airlines and IBM. The idea traces to a chance 1953 encounter between American’s president and an IBM salesman on a cross-country flight, and after years of development, the system took over American’s reservations nationwide in 1964. It ran on two IBM 7090 mainframes in a purpose-built data center in Briarcliff Manor, New York, linked to terminals in more than 50 cities, and cut the time to book a seat from about 90 minutes to seconds. Soon the largest commercial real-time data processing system in the world, SABRE was not the first computerized reservation system, but it was the most consequential, and it opened the door to enterprise-scale data centers.

Watch a video overview of the history of data centers below:

The growth of cloud computing

During the 1970s, computers got smaller and cheaper. Altair, Apple, Tandy Radio Shack, and later IBM introduced personal computers for hobbyists, enthusiasts, and small businesses through the late 1970s and early 1980s. The IBM PC, introduced in 1981, did not invent the personal computer, but its credibility helped bring computing into mainstream businesses and offices, and its design became the industry standard that most later PCs followed. These early machines operated as standalone devices, requiring users to physically carry data from one computer to another.

By the early 1990s, PCs were connecting to servers in the client-server model, largely considered the precursor to the need for data centers. A major shift occurred in the mid to late 1990s, when the dot-com boom occurred. Between 1993 and 1995, America Online (AOL) made email and the internet widely available to users. 

The explosive growth of the internet and online businesses during the mid-1990s created a massive and sudden demand for physical infrastructure to house and run all those websites and online services – what we now think of as data centers.

By 2000, more than 80 percent of American banks offered online banking. Then came the shift that reshaped everything: the launch of Amazon Web Services (AWS) in 2006 marked the start of modern cloud computing. Rather than every company buying and maintaining its own computers, cloud computing lets businesses and individuals rent computing power, storage, and software over the internet, on demand, from massive data centers run by providers like Amazon, Microsoft, and Google. It is the difference between owning a generator and simply plugging into the grid. As cloud computing has become woven into daily life, so has the demand for AI, and for the data centers that power both.

The takeaway: Modern cloud computing, which both drives and accompanies the expansion of AI, is a major factor driving the demand for data centers today. While data centers are the subject of heightened media scrutiny, they have been woven into the fabric of American life for decades.

What is new is their size, their concentration, and the speed at which communities are being asked to accommodate them.

A data center is a secure facility that houses servers and networking equipment for processing, storing, and transmitting digital information. People make use of data centers in their everyday life without even realizing it; all of the AI technology being used today runs through data centers. A 2025 Gallup survey found that nearly every American uses products that utilize AI, but two-thirds aren’t aware of it. When someone sends an email, streams a movie, runs a payroll, makes a credit card payment, or queries an AI tool, the computation occurs in a data center. 

Further, people expect all of this technology to be available 24-7, without fail. In the IT industry, this has been described as the “Five 9s” – that expectation that technology must be available and operable 99.999 percent of the time. This puts even greater pressure on data centers to be reliable and fully functional without fail. 

Major data center operators today include Amazon Web Services, Microsoft, Google, and Meta. These facilities operate continuously and are engineered for near-perfect uptime. 

Learn more about data centers from Google (3 minutes):

DIFFERENT TYPES OF DATA CENTERS

Data centers are often grouped into four categories:

A hyperscale data center is the largest class of data center, built by major cloud providers like Amazon, Microsoft, Google, and Meta to handle enormous cloud and AI workloads and to serve thousands of customers at once. What sets hyperscale apart is not just size but design: rather than being built for a fixed job, these facilities are engineered to keep growing, adding tens of thousands of servers without being rebuilt. By the common industry definition, a hyperscale facility has at least 5,000 servers, occupies 10,000 square feet or more, and draws at least 40 megawatts of power. In practice, most are far larger, and that 40-megawatt figure is now considered entry-level: new campuses built for AI typically range from 100 to 1,000 megawatts each, and the very largest are being planned in the thousands. To put that in perspective, the Congressional Research Service notes that a data center’s power draw is often compared to how many homes the same electricity could serve: roughly 100 megawatts is enough to power about 80,000 U.S. homes, so a large hyperscale campus can use as much electricity as 80,000 to 800,000 households. At that scale, these facilities have more in common with utility infrastructure than with a corporate office building.

An enterprise data center is a “company-owned data center used for internal data processes.” Large corporations, such as banks, hospitals, and retailers, may operate their own data centers to run internal systems. These enterprise facilities are typically smaller and purpose-built for a single organization. Both hyperscale and enterprise data centers have been part of community landscapes for decades; what is generating attention today is primarily the hyperscale category.

Multitenant, or “hotel-style,” data centers lease space, power, and bandwidth to many different companies at once. These are also called colocation or managed data centers, and a single facility can host hundreds or even thousands of tenants. The world’s largest colocation company is California-based Equinix, which operates hundreds of facilities worldwide. The single largest colocation facility, however, is Switch’s Citadel Campus near Reno, Nevada, whose tenants have included eBay, Microsoft, Google, and PayPal, a scale of up to 7.2 million square feet that shows just how large a single shared facility can grow..

An edge data center is a smaller, local facility that is built near its end users. Edge computing is the practice of processing data close to where it is generated, rather than routing it to a distant central facility. Proximity makes certain technology more possible at a higher performance level. When data has to travel a long distance to be processed and returned, there is a delay, measured in milliseconds. For most applications, that delay is invisible. But for a growing range of services, it matters enormously.

Here are a few examples of what edge computing enables:

• Self-driving and connected cars, reacting to the road in milliseconds, far too fast for a distant data center.
• Passport e-gates, scanning your passport and matching your face on the spot.
• Tap-to-pay and fraud checks, clearing or flagging a card the instant you tap.
• Telemedicine and remote diagnostics, bringing specialist care to rural patients.

Learn more about the different types of data centers in this video (2:30 minutes):

When asked about their use of AI technology in daily life, only 36 percent of Americans say they are using these tools. What is most likely coming to mind for these survey respondents is their use of websites and apps such as ChatGPT or Claude.ai. But in reality, everyone is using technology powered by data centers. 

• Everyday financial life: Every time you place a grocery pickup order, swipe your credit card, pay back a friend using Zelle or Venmo, or receive a direct deposit of your paycheck, your financial life is relying on data center infrastructure. Financial institutions require split-second processing and redundant storage to keep money moving and records accurate.
• Healthcare: From the moment a mother checks into the hospital, her care is tracked digitally. When her baby is born, the newborn is issued a barcoded wristband and its health information is entered into cloud-based electronic records. Every blood pressure reading, temperature check, medication, and clinician visit is logged in the cloud. All of it depends on reliable data center capacity, and a doctor or nurse who suddenly could not reach a patient’s records could face an immediate crisis.
• Government and public safety: Tax systems, benefits programs, 911, and law enforcement databases operate continuously and require the same 99.999 percent reliability as financial services.
• Supply chains and logistics: If you’ve ever received an Amazon Prime delivery, a gift that was shipped to you, or ordered clothing online from a major retailer then you used a data center. The systems that route packages, manage inventory, and coordinate freight, operate in real time. Every UPS and FedEx truck rolling down the street relies on data center and AI technology. Even a small delay in data processing can ripple into missed deliveries and supply disruptions.
• Communications: According to Pew, 91 percent of American adults have smartphones and nearly every American adult has a cell phone. All of these devices rely on data centers to make traditional phone calls, video calls, send email, use messaging apps, and many other functions these devices offer.

Go inside one of Google’s data centers in Alabama (6 minutes):

In May 2026, the Data Center Coalition released a report by PricewaterhouseCoopers quantifying the economic contributions of the U.S. data center industry for 2023 and 2024, at the national level and in all 50 states. Among its findings:

• The industry directly employed more than 1 million people in 2024 (1,005,080), in roles such as network engineers, IT security analysts, cooling-system technicians, and security personnel. Counting the jobs that data centers support elsewhere, each direct job is estimated to sustain about 4.5 additional jobs, the industry supported 5.5 million jobs in all, up 17% from 4.7 million the year before.
• The industry’s direct contribution to U.S. GDP was $926.9 billion in 2024, a 21% increase over 2023, and its total contribution, including indirect and induced effects, reached $1.7 trillion.
• Its total fiscal support to federal, state, and local governments was $204.4 billion in 2024, up 24% from the year before.

Verified: coal generated roughly 55-60% of China’s electricity in 2024 (about 55% of the total power mix, and China alone produces about 56% of the world’s coal-fired power). So the coal point is accurate and well documented. Here’s the tightened version with the coal detail integrated:

Data centers are the physical foundation of American competitiveness in the AI era, but that foundation ultimately rests on electricity. The Federal Reserve finds that the United States leads the world in AI infrastructure and controls an estimated 74% of global high-end AI computing capacity. 

But that lead is not guaranteed, because it depends on the one input where the U.S. trails: power. America is ahead on the advanced semiconductors that AI runs on, but China has a commanding lead in generating electricity, roughly 3,200 gigawatts of installed capacity to America’s 1,293, and in 2024 it added fifteen times more new capacity than the U.S. did. Much of that Chinese power still comes from coal, which supplies more than half of its electricity, but the sheer scale of generation is what matters for AI. OpenAI calls this mismatch between chips and power the “electron gap,” and it could tip the balance of AI computing between the two nations. 

The stakes are national: whichever country can produce the power to run AI at scale will shape markets and military capability for decades, and America’s edge in chips means little without the electricity to sustain it.

The services these facilities enable have broadened. Beyond streaming, email, and video calls, data centers now provide the computing behind:

• The artificial-intelligence tools a growing number of Americans use for everyday tasks, from drafting and research to health information and planning; and
• The infrastructure underpinning banking, medical records, and national-security systems.

Supporters argue this reframes the question, from “do we want a warehouse of servers nearby” to “do we want a stake in the infrastructure the modern economy increasingly runs on?”

JOB OPPORTUNITIES IN CONSTRUCTION AND OPERATIONS

When a data center comes to a community, it brings construction employment, often supporting thousands of local skilled tradespeople such as electricians, pipefitters, and steelworkers during a multi-year build, followed by a smaller base of well-paying permanent operations jobs.

PROPERTY TAXES

Data center facilities pay substantial property taxes, yet they ask relatively little of a community in return. The appeal is the ratio: a data center adds substantial tax revenue but almost no new residents, so it avoids the population-driven costs, new schools, expanded services, residential sprawl, that come with most large developments. It still requires road access, utilities, and emergency coverage, but those costs are modest next to what it pays in.

• In Loudoun County, Virginia, data centers occupy about 4% of the county’s commercial land but generate 38% of its general fund revenue, funding schools, public safety, libraries, parks, and mental-health and child-protection services, plus roads and bridges, all while the county has lowered homeowners’ property tax rate.
• In Quincy, Washington, the tax base from two decades of data center growth funded a new high school, hospital, fire station, and city hall, while homeowners’ tax rates fell rather than rose.
• In rural Richland Parish, Louisiana the construction of Meta’s Hyperion data center produced a surge in local sales tax revenue that the school district used to fund substantial teacher bonuses. In 2026, end-of-year payments for eligible certified employees reached $50,935, up from $10,200 the year before, in a parish where teacher salaries range from about $29,500 to $52,300. The district’s superintendent has said the bonuses help attract and retain certified teachers. The increase is tied largely to the multi-year construction phase, so district officials note the elevated revenue may level off once the facility is complete.

DATA CENTERS AND PUBLIC OPINION

Data centers are foundational infrastructure, much like roads, bridges, and the electric grid. Nearly everyone relies on the services they support, yet public opinion has turned sharply negative. A May 2026 Gallup poll found that roughly seven in ten U.S. adults would oppose an AI data center being built in their local area, even among people who use the AI tools those facilities run.

Why Data Centers are Drawing New Scrutiny

As this Brief has presented, data centers predate today’s AI boom, but the rise of tools like ChatGPT and Claude has drawn public attention to the infrastructure behind them. Because AI drives demand for computing power and data center capacity, the current construction boom has made these once-invisible facilities suddenly visible, and controversial.

Advocacy organizations have seized on that attention. On the environmental front, Food & Water Watch has become a prominent national organizer, known for a December 2025 coalition letter signed by more than 230 organizations and a push for state and federal moratoriums; it takes credit for the federal moratorium initiative advanced by Sen. Bernie Sanders and Rep. Alexandria Ocasio-Cortez. The Sierra Club has called data centers “fool’s gold” and issued state policy guidance urging protections against higher electricity bills and greater transparency. Others, including the NAACP, have turned to litigation.

Concerns have also been raised about health and quality of life. Children’s Health Defense reports that people near data centers experience ailments such as nausea and insomnia; the Environmental and Energy Study Institute has highlighted noise pollution; and the Southern Environmental Law Center commissioned a study by a Harvard biostatistician on air-quality risks. Some of these claims rest on solid evidence and some do not, which is itself the point: the volume of a concern is not the same as its weight, and sorting the two is the citizen’s task.

What the Concerns are Really about

Behind the campaigns are several specific concerns. Understanding what each does and does not involve is the difference between an informed judgment and a reaction to a slogan.

Electricity and cost. This is the concern the public rates most seriously, and it is well-founded in principle: a large data center can draw as much power as a small city, around the clock. The central question is who pays for the new generation and grid upgrades. The evidence is genuinely mixed, some analyses find large customers who cover their own infrastructure can hold down average rates, while others document real upward pressure on bills in fast-growing regions, so the outcome depends heavily on how a state’s utility commission allocates the costs.

Water. Water use is real but often misunderstood, because coverage conflates water withdrawn (much of it returned) with water consumed (evaporated and lost). Most of a data center’s water footprint is indirect, tied to generating its electricity rather than cooling its servers, and newer closed-loop designs sharply reduce on-site use. Water is a manageable issue in most locations and a serious one in the water-stressed regions where roughly two-thirds of recent facilities have been built. Location determines the stakes.

Emissions. Because data centers run on grid power, their emissions largely reflect whatever generates electricity locally. Their rapid growth pushes up total demand, and with it emissions, even as efficiency per unit of computing improves.

Quality of life. Noise, construction traffic, water discharge, and the visual footprint of a large industrial site are legitimate local concerns, and also among the most negotiable, through setbacks, screening, landscaping, and noise limits written into local approvals.

The point is not that these worries are baseless, several are well-founded, but that a national campaign cannot tell you how any of them apply to the specific project proposed in your community. Only the details can.

Organized Campaigns and the Question of Coordination

What began as scattered local objections has become a coordinated national movement, and reading the debate clearly means recognizing that. Some of that coordination is openly acknowledged by the advocacy groups above. But questions have also been raised about who is funding and directing parts of it.

Most prominently, a network of nonprofits connected to Neville Roy Singham, an American-born tech magnate living in Shanghai who openly aligns himself with the Chinese Communist Party, has drawn congressional scrutiny. Reports from groups including the Bitcoin Policy Institute allege that Singham-funded organizations such as CodePink and the Party for Socialism and Liberation have helped organize local opposition to U.S. AI infrastructure, and Sen. Tom Cotton has asked the Department of Justice to investigate possible Foreign Agents Registration Act violations. These are allegations under review, not settled findings, and it is worth being precise about what they do and do not show: even if some national organizing is coordinated or foreign-linked, that would not by itself make the underlying local concerns invalid, any more than industry-funded messaging in favor of data centers makes the benefits imaginary.

That is the balanced way to hold it. The data center debate is now crowded with polished messaging on every side, activist campaigns that overstate the science, and industry campaigns that gloss over real local costs. Tactics like Rep. Ocasio-Cortez holding up a jar of brown water at a 2026 EPA hearing are designed to move opinion, not to inform it. So are the industry’s rosiest projections. A citizen’s job is to discount the theater on both sides and look at the facts of the actual project.

More than “Not In My Backyard”

The phenomenon known as “Not In My Backyard” (NIMBY) describes opposition to nearby development, whether over property values, environmental effects, or a community’s changing character. Like housing, energy, and industrial projects before them, data centers now prompt these debates, and some data center proponents argue NIMBY sentiment has become a major roadblock to construction.

Part of the tension is a structural mismatch: the benefits of a data center spread across a large, dispersed population, while the most direct effects fall on the people who live nearest to it. But it would be a mistake to treat all opposition as unfamiliarity. A March 2026 Pew Research Center survey found that Americans who have heard the most about data centers hold the most negative views, among those who had heard “a lot,” two-thirds called them bad for home energy costs and 63% bad for the environment. Concern often rises with information, not in spite of it. That same survey shows a public weighing genuine trade-offs: negative on environment, energy costs, and quality of life, but positive on local jobs and tax revenue. Those are exactly the trade-offs a productive local debate should weigh.

A Growing Local Political Liability

Data centers have never mapped neatly onto the usual partisan divide, and heading into the 2026 elections they have become something rarer: an issue that cuts across party lines and across every level of government. As opposition has spread from a few hotspots into city council and county board races around the country, candidates in both parties increasingly treat their record on a nearby data center as something that can decide an election. What was once a routine economic-development vote has become one that officeholders across the spectrum now weigh carefully, mindful of how voters will judge where they came down.

Data centers sit at the intersection of multiple overlapping jurisdictions: federal agencies set the broad conditions for industry growth, independent commissions oversee interstate power grids, state legislatures and public utility regulators control incentives and policy, county and, local governments make most of the decisions that determine whether a specific project gets built, and utilities deliver the power and water required. Understanding who does what, and where the leverage points are at each level, is essential for communities navigating a proposal.

THE FEDERAL GOVERNMENT: SETTING THE FRAMEWORK

The Property Clause gives Congress authority over federal land, which it exercises through agencies like the Department of Energy, the Department of Defense, and the Department of the Interior. 

Most data centers are built on private land, but the federal government has recently moved to make some federal sites available: the Department of Energy has identified national-laboratory and former-weapons sites to host data centers paired with new power generation, and federal land also matters for the transmission lines and energy projects that serve these facilities. And because Congress has already delegated broad authority to federal agencies through laws like the National Environmental Policy Act (NEPA) and the Federal Power Act, a president can often direct those agencies to use that authority differently, adjusting how permits are reviewed or priorities are set, without a new act of Congress, though the agencies must still act within the limits those laws set.

An executive order does not create new federal power. It directs how the executive branch uses power it already has. For more on how the president’s powers and federal agency structure, see our Executive Branch Policy Circle Brief.

Approval for individual data centers still happens at the state and local level. But federal policy controls the environment the entire industry operates in, and that environment shifted dramatically in 2025.

MAKING DATA CENTER CONSTRUCTION A NATIONAL PRIORITY

On July 23, 2025, the Trump Administration released “Winning the Race: America’s AI Action Plan,” which made accelerating data center construction a national priority. The logic is direct: data centers supply the computing power that runs artificial intelligence, and AI has become a driver of research and innovation across medicine, science, defense, and industry. The same day, President Trump signed Executive Order 14318, “Accelerating Federal Permitting of Data Center Infrastructure.” 

The executive order states a priority “to facilitate the rapid and efficient buildout of this [AI] infrastructure by easing Federal regulatory burdens,” and that the Administration “will utilize federally owned land and resources for the expeditious and orderly development of data centers.”

The order works through two building-block definitions. A “Data Center Project” is a facility needing more than 100 megawatts of new electricity dedicated to AI. A “Covered Component Project” is the supporting infrastructure data centers depend on, or a facility that makes it: energy infrastructure like transmission lines, natural gas pipelines, and substations; power sources like natural gas, nuclear, and geothermal; semiconductors; and backup power. This second category matters because it reaches far beyond the data center building itself. A new power plant, a pipeline, or a chip factory built to serve a data center can all fall within it, and could sit miles away.

Financial support is then reserved for what the order calls a “Qualifying Project.” To qualify, a project must first be one of those two types, a Data Center Project or a Covered Component Project, and then also meet one of these tests:

• A committed investment of at least $500 million;
• A new electric load above 100 megawatts;
• A national security purpose; or
• A special designation by the Secretary of Commerce, Defense, Interior, or Energy.

Projects that clear both bars become eligible for federal loans, loan guarantees, grants, tax incentives, and purchase agreements, powerful tools for moving a project forward.

The order reaches into environmental review as well. It directs agencies to treat many of these projects as routine actions that “normally do not have a significant effect on the human environment,” and it presumes that when federal money covers less than half of a project’s cost, the project does not trigger a full National Environmental Policy Act review. In practice, that can remove one of the biggest sources of delay for a large project.

Taken together, the order shows how much leverage the federal government holds over the data center buildout, not by constructing facilities itself, but by deciding which projects get federal land, federal money, and a faster path through federal permitting.

Key federal agencies involved include:

• Department of Energy which manages grid planning, national laboratory sites, and energy research.
• The Environmental Protection Agency handles air quality permitting for backup generators and Clean Water Act Review.
• Army Corps of Engineers reviews Clean Water Act permits for transmission line crossings. 
• Department of Interior which controls federal land.

Acting on the order, the Department of Energy selected four initial federal sites — the Idaho National Laboratory, Oak Ridge Reservation in Tennessee, Paducah Gaseous Diffusion Plant in Kentucky, and the Savannah River Site in South Carolina — and invited private partners to build data centers paired with new power generation.

GRID REFEREE: FEDERAL ENERGY REGULATORY COMMISSION (FERC)

The Federal Energy Regulatory Commission (FERC) is an independent agency that regulates the interstate electricity transmission and wholesale power markets. As data centers have grown into some of the largest electricity customers in the country, FERC has become an increasingly active regulator of the issues they raise.

One of the priorities of communities is cost allocation: who pays for the grid upgrades these facilities require? In December 2025, FERC issued a landmark order to PJM Interconnection, the operator of the nation’s largest grid serving 67 million across 13 states and the District of Columbia. The order stated: “As technology leaps forward, clear and fair regulations must keep pace to support advancement, help prevent price volatility, and promote competition, ultimately benefiting consumers by keeping electricity costs manageable.”

The line between federal and state authority here is contested. Under the Federal Power Act, states, not the federal government, regulate the sale of electricity to homes and businesses and how large new customers connect to the grid. When FERC proposed national rules for connecting large users like data centers, the National Association of Regulatory Utility Commissioners, which represents state regulators, pushed back, arguing the rules would intrude on state authority.

How this tension is resolved will shape how much authority local communities retain over the energy-cost consequences of data center development.

STATE GOVERNMENT: INCENTIVES, ENERGY POLICY, AND THE COMPETITION FOR INVESTMENT

The federal government sets the backdrop, but the states hold most of the levers that determine whether a data center gets built, how it is powered, and who pays for it.

That responsibility has grown heavier because the demand was foreseeable. After roughly 15 years of nearly flat electricity consumption, U.S. demand has been rising again over the last five years. The largest drivers are data centers and a wave of new manufacturing, with electric vehicles and the broader electrification of daily life adding to the trend.

States and their utility regulators have had that trend in front of them for years, which frames the central question now facing every statehouse: have their energy policies kept pace with a need for more generation and grid capacity that was visible well before the AI boom made it urgent? Building that capacity is itself contested over how much, what mix of sources, and who pays, but ensuring an adequate, reliable, and affordable power supply is among the oldest responsibilities of state government, and data centers have turned a slow-building challenge into an immediate one. 

HOW STATES EXERCISE THIS POWER

Three bodies of state government hold the main levers, each shaping a different part of the picture.

1. Legislatures set the rules. State legislatures write the laws that govern both the incentives to attract data centers and the protections that decide who bears their costs. Their choices fall into a few areas:

• • Tax incentives. More than three dozen states offer dedicated tax breaks to attract data centers, according to the National Conference of State Legislatures, ranging from sales-tax exemptions on servers to multi-decade property-tax abatements.
  • Ratepayer protections. A growing number of legislatures are writing rules to ensure data centers pay for the new power and grid upgrades their demand requires, rather than shifting those costs onto households and small businesses.
  • Siting authority. Legislatures decide whether local governments control where data centers are built, or whether that authority is centralized at the state level. States have gone in opposite directions: Georgia and Virginia have largely left siting to local zoning, while West Virginia’s 2025 Power Generation and Consumption Act (HB 2014) took the opposite approach, creating “high impact data center districts”  that are exempt from county and municipal zoning, noise, and land-use ordinances, the first law of its kind in the country.

2. Governors recruit investment and set energy strategy. Governors court data center projects, sign or veto the legislation above, and set the state’s broader energy direction, including how much new generation to encourage and of what kind. A governor focused on economic development may champion incentives; another focused on ratepayers or reliability may push protections or new generation instead.
3. State Utility Commissions decide the economics. State utility commissions approve utility rates, review the long-term plans for building new generation and transmission, and determine how the cost of serving a large new customer is spread across everyone else. This is where the abstract question of “who pays” becomes a concrete number on a monthly bill.

These commissions are also where a hidden feature of the grid comes into play: no state is an island. Electricity flows freely across state lines through regional grids, and almost 10% of all U.S. electricity is traded between states. The power that lights a home in one state is often generated in another.

Making sense of the tax debate. Because the word “tax” appears on both sides of the incentive question, it is easy to confuse two different things:

• A tax exemption is revenue the state chooses not to collect. To win a project, a state may waive the sales tax a company would owe on its servers, equipment, and sometimes its electricity. On a purchase of that size, the savings can reach into the hundreds of millions, money the state forgoes up front.
• Tax revenue is what the project still generates once it is built: property taxes on the land and buildings, plus the economic activity, construction, jobs, and local spending that produces further revenue.

In short, a state gives up some revenue up front (the exemption) in the hope of gaining more over time (property taxes and jobs). Whether that trade pays off is a genuine, ongoing debate: supporters argue the incentives attract investment that expands the tax base and creates jobs a community would not otherwise have; critics question whether decades of forgone taxes are justified, especially since data centers employ relatively few permanent workers. That balance, what a community gives up against what it gains, is what state and local officials must weigh.

States are answering these questions differently, and the range is instructive. Minnesota has paired its incentives with strong ratepayer protections, ending the sales-tax exemption on the electricity data centers consume and directing regulators to make sure large users, not households, cover the costs they impose. Utah has focused on power supply, making it easier for the largest electricity users to secure their own generation rather than lean on the existing grid. West Virginia has taken the opposite tack on control, centralizing authority at the state level and limiting how much say local governments have over where and how data centers are built. Each reflects a different answer to the same question: how to welcome the investment while managing its costs.

Yet even as states set the framework, much of what a community actually experiences-the zoning decision on a specific parcel, the conditions attached to a permit, the negotiation over noise and setbacks-is decided closer to home. Except where a state has centralized that authority, as West Virginia did, the most consequential decisions about an individual project often rest with county boards, city councils, and planning commissions. It is at this local level that the abstract of state policy becomes a concrete proposal in a specific neighborhood.

COUNTY AND LOCAL GOVERNMENT: THE FRONT LINE OF DECISION-MAKING

For most data center proposals, this is where the decisions get made. County boards, city councils, and their planning and zoning commissions hold the powers that determine whether a specific project is built, where, and on what terms. Understanding what these bodies actually decide, and in which meetings, is the key to engaging effectively.

What local governments decide

Their authority falls into a handful of concrete powers:

• Zoning: Whether a data center is even allowed on a given parcel, and under what category (industrial, utility, or a purpose-built “data center” designation). Changing what a parcel may be used for typically requires a rezoning, decided by the county board or city council after a recommendation from the planning commission.
• Land-use and site-plan approval: Whether a specific proposed building meets local requirements for setbacks, height, lot coverage, screening, and traffic.
• Special-use (or conditional-use) permits: Many communities require a data center to obtain a special-use permit, which lets officials attach binding conditions, on noise, water use, hours of construction, landscaping, or decommissioning, to an individual project.
• Development and community benefit agreements: Negotiated contracts in which a developer commits to specific local benefits (infrastructure, hiring, contributions to schools or emergency services) in exchange for approval.
• Timing: The power to pause. A temporary moratorium lets a community stop accepting applications while it studies a project’s water, energy, and zoning implications and writes clearer rules.

WHERE THE DECISIONS GET MADE, IN PUBLIC AND BEFORE

Most of the formal, binding decisions happen in public meetings that residents can attend and speak at:

• The planning (or zoning) commission holds hearings on rezonings, special-use permits, and site plans, and makes recommendations.
• The county board of supervisors or city council casts the binding votes on rezonings, permits, ordinances, and moratoriums.
• The board of zoning appeals handles variances and appeals of zoning decisions.

Each of these bodies is required to hold public comment periods, which is where an informed resident has the most direct influence.

But a great deal of the shaping happens before any of that, and often before the public knows a project exists. By the time a proposal reaches a public hearing, key terms may already be well advanced:

• Developers frequently meet privately with economic-development officials before an application is ever filed.
• Land is often acquired quietly, sometimes through intermediaries, so the buyer’s identity and intentions stay confidential.
• Utilities and developers may negotiate power and infrastructure arrangements before the project becomes public.
• Tax incentives and development agreements are sometimes substantially negotiated before formal hearings begin.

Non-disclosure agreements are common in this early phase. They have legitimate uses, companies protecting land prices and competitive plans, but their use with public officials has become contentious enough that lawmakers in at least ten states have proposed banning or limiting them for data center and economic-development deals, often with bipartisan support. Some proposals would simply require that key details like a project’s power and water use be disclosed to the public.

Writing the rules in advance

When an ordinance says nothing about data centers, approval can turn on how a zoning officer reads a vague label like “industrial,” “utility,” or “technology facility,” a gap that leaves both the community and the developer guessing. 

To avoid that, a growing number of communities are updating their codes before a developer arrives, setting clear standards for siting, water, noise, setbacks, and increasingly, design and landscaping. For example, Atlanta barred data centers within a half-mile of its Beltline trail and any transit station. In June 2025, the city council went further, voting to require a special-use permit for every new data center citywide, bar them outright in certain neighborhoods, and require applicants to disclose their water and energy demands and any tree removal. 

Appearance has become a real point of leverage: because there are only so many viable sites, local governments have gained the upper hand in negotiations and are demanding buildings that fit their surroundings rather than the anonymous windowless boxes that fuel much of the opposition. 

Developers have responded: the architecture firm Gensler, whose data center work has grown rapidly, now designs facilities with landscaped grounds, green roofs, and living walls, including a Phoenix campus wrapped in weathering steel and paired with public space. For a community writing its rules early, design and landscaping standards are among the most achievable things to ask for, and among the most visible in daily life.

Learning from communities that have done it before

Not every community is starting from scratch. Some have hosted large-scale digital infrastructure long enough to know what works. MidAmerica Industrial Park in Pryor, Oklahoma, is one such place. Fifteen years after Google opened its first data center there, the 9,000-acre site has grown into one of the largest industrial parks in the country. It is home to more than 80 companies, including Google, DuPont, Airgas, and Berry Global, employing over 4,500 workers across technology, aerospace, advanced manufacturing, and logistics. Its experience points to the through-line of this whole section: the communities that fare best tend to be the ones that plan ahead for water, energy, and infrastructure, and use the local powers described above deliberately, rather than reacting to each new tenant as it arrives.

Many communities have made these projects work, turning a large new development into lasting local benefit. Their experiences show that the outcome is rarely accidental: it comes from clear agreements, honest engagement, and companies that hold to their commitments over time.

The practical takeaway is that no single level of government controls data center policy entirely. A proposal that clears local zoning may still require state environmental permits. A facility approved at all levels still needs a utility interconnection agreement. Federal land use policy sets the envelope, but states and localities set the terms within it.

For communities, this means engagement needs to happen at multiple levels simultaneously:

• At the local planning board and board of supervisors: to shape zoning standards, require special use permits, and negotiate development agreements and community benefits agreements.
• At the state legislature and governor’s office: to influence incentive design, ratepayer protection requirements, and what local governments are authorized to require.
• At the state public utility commission: to ensure that grid upgrade costs are fairly assigned and that rate impact analyses are publicly available before approvals.
• At FERC proceedings (for communities in multi-state grid regions like Pennsylvania, New Jersey, and Maryland): to ensure that co-location arrangements and large load interconnection rules protect existing customers.

ENERGY DEMAND AND THE OPPORTUNITY

Picture the inside of a personal computer. It has a processor that runs applications, drives for storing data, a network card linking it to the world, and a fan that keeps it from overheating. A data center is the same idea, but vastly larger, shuttling data around the clock for cloud computing, streaming, financial services, and medical care. That scale is why a data center’s energy footprint looks nothing like a home computer’s.

This is a real challenge for communities and the grid. But it is also, increasingly, a source of real solutions, and the rest of this section explains both.

Why Data Centers Use So Much Power

Roughly two-thirds of a data center’s electric demand comes directly from running IT equipment. All of that processing generates heat, and a substantial share of the remaining electricity goes to keeping the machines cool enough to keep working. The industry measures this with Power Usage Effectiveness (PUE), the ratio of total facility power to the power that actually reaches the computing equipment. A PUE of 1.0 would mean none of the power is lost to overhead like cooling and lighting. According to the Uptime Institute’s 2025 Global Data Center Survey, the global average PUE was 1.54 in 2025, essentially unchanged for the past several years, though leading hyperscale operators like Google (1.09) and Microsoft (1.17) report to achieve higher efficiency.

Because data centers run nonstop, they also carry batteries and backup generators in case of a power failure, along with fire-suppression systems and around-the-clock physical and cyber security. Outages are genuinely costly: According to Uptime Institute’s Annual Outage Analysis 2025, 57% of operators said their most recent significant outage cost more than $100,000, and one in five said it cost more than $1 million.

The Scale of Growth

In 2023, U.S. data centers used approximately 176 terawatt-hours of electricity, according to the Department of Energy and Lawrence Berkeley National Laboratory. That load has tripled over the past decade. Data centers consumed about 4.4 percent of total U.S. electricity in 2023, a share projected to rise to between 6.7 and 12 percent by 2028 as AI adoption accelerates.

A single large data center can draw as much power as a small city, and that power cannot simply be switched on. Utilities must plan years ahead so the grid can absorb new demand without affecting existing customers, and regulators decide who bears the cost of any necessary upgrades.

Where the Power Comes from

A data center runs around the clock, every hour of every day. That single fact shapes everything about how it is powered: it needs electricity that is constant and reliable, not power that comes and goes. This is why wind and solar, for all their growth, cannot run a data center on their own. The sun sets and the wind drops, but the servers never stop, so even operators with ambitious renewable goals depend on firm, always-available power to fill the gaps, whether from natural gas, nuclear, hydropower, or renewables paired with enough battery storage to smooth the shortfalls.

There are two basic ways a data center gets that power: it can plug into the regional electric grid, or it can generate its own.

Most still rely on the grid. A data center does not run on a special fuel blend of its own; when it draws from the grid, the mix powering it simply reflects whatever generates electricity in that part of the country, which varies widely between, say, Virginia and Texas. Nationally, the grid runs largely on natural gas, the largest single source, alongside renewables, nuclear, and a declining share of coal.

But in the fastest-growing markets, plugging in has become the hard part. The grid is running short of spare capacity, and the wait to connect a large new facility has stretched out: in parts of Northern Virginia, utility connection timelines have grown from three or four years to as many as seven. 

That bottleneck is changing how data centers get built. Rather than wait, a growing number of developers are choosing to bring their own power, building on-site natural gas plants, contracting directly for nuclear output through the grid, or acquiring generation to run alongside or independent of the public grid. The idea has support at high levels: the U.S. Energy Secretary has said developers who cannot find available power should “bring the energy with [them].” Monitoring Analytics is the independent watchdog for PJM.  The operator that coordinate the grid across 13 states from Washington, D.C. to Chicago and has argued that self-generation may be the only realistic way to add large new loads to a system with no room to spare. Bringing its  own power can ease the strain on existing ratepayers. However, it raises open questions about how to regulate generation that sits outside the traditional utility system.

This points to a larger issue often lost in the debate. As energy writer Emmet Penney argues in The Free Press, rising electricity bills trace less to data centers than to years of underinvestment in the power grid. By that view, data center demand did not cause America’s power squeeze so much as expose it. Whether the answer is building far more capacity or pacing growth until it exists, the underlying challenge predates the AI boom.

A Historical Precedent: Aluminum Smelters in the Pacific Northwest

Concentrated industrial electricity demand is not new. In the mid-20th century, much of the nation’s aluminum smelting clustered in the Pacific Northwest to tap cheap hydropower, becoming some of the region’s largest power users, and the grid was built to serve them. The IEA draws the same parallel to data centers today. The lesson is that large, predictable demand is something a grid can plan and build around.

Data Centers as Long-Term Customers for New Power

Large data centers consume power predictably and at scale, which means a hyperscale operator signing a long-term power purchase agreement becomes a guaranteed buyer for generation that might not otherwise have been built. When Microsoft signed a 20-year agreement with Constellation Energy to restart Three Mile Island Unit 1, rebranded the Crane Clean Energy Center, and is expected back online by 2027 or 2028. While Microsoft contracts for the plant’s output, the restart adds roughly 835 megawatts of net-new carbon-free generation to the PJM grid, strengthening overall supply for the region. A similar dynamic is underway in Iowa, where Google signed a 25-year agreement with NextEra Energy to restart the 615-megawatt Duane Arnold plant, targeted to return to service by early 2029.

Small Modular Reactors: Promising but Early

Small modular reactors are a newer option still, designed to generate up to 300 megawatts each. Built largely in a factory rather than on site, they are intended to be faster to construct and easier to site, including directly alongside the data centers they would power. 

Compact reactors are not a new idea in themselves. The U.S. Navy has powered submarines and aircraft carriers with more specialized versions of small reactors for some 70 years. Adapting the concept into standardized, factory-built units that can sell power to the grid at competitive cost is the unproven part. No commercial small modular reactor is yet operating in the United States, and cost, regulatory timelines, and financing risk all remain open questions.

Competing for the Same Grid: Manufacturing and AI

In 2022, two pieces of federal legislation – the CHIPS Act and Inflation Reduction Act – triggered a surge in domestic manufacturing just as the launch of ChatGPT set off an AI data center buildout that neither piece of legislation had anticipated. U.S. manufacturing construction has more than doubled since then, driven by semiconductor fabrication and battery plants coming online, adding to the same grid that data centers are straining.

The Global Context: A Smaller Piece than it Seems

Despite the headlines, data centers are not the single largest driver of rising electricity demand. The IEA’s Energy and AI report puts data centers at roughly one-tenth of global electricity demand growth through 2030, less than the contribution from industrial motors, air conditioning, or electric vehicles. The grid will need to grow regardless, to meet an increasingly electrified economy with or without AI. The question is not whether growth is coming, but whether it is planned for, paid for fairly, and matched with the kind of reliable, diverse generation that benefits everyone connected to the grid, not just the facility that prompted it.

WHY WATER IS A COMMUNITY CONCERN

If energy is the first major point of contention when a community debates welcoming a data center, water is the second. It deserves careful attention, and it also deserves precision, because much of the public conversation blends together measures that are not the same thing. Getting the concepts straight is the difference between a productive local debate and one driven by frightening but misleading numbers.

Two words get used interchangeably in water coverage, and telling them apart resolves much of the confusion.

• Withdrawal is water drawn from a source. A great deal of it, especially the water that cools conventional power plants, is returned to the river or lake afterward, largely unchanged.
• Consumption is water that is used and not returned, usually because it evaporates. This is the water actually removed from the local supply.

For a deliberately contrarian treatment of why national water fears are often overstated, see independent environmental writer Andy Masley’s “The AI water issue is fake” (note the provocative title; Masley writes from a pro-AI perspective, and he is careful to agree that local water stress is a real problem worth planning for).

The two ways a data center uses water

A data center’s water footprint has two very different components, and separating them is the key to understanding any specific project.

• Directly, on site, to cool the servers. This is the water most people picture, and it is the water a community’s own utility or aquifer supplies.
• Indirectly, off site, to generate its electricity. Most thermal power plants use water for their own cooling, so every kilowatt-hour a data center draws carries a hidden water cost at the power plant.

The indirect share is the larger one. Lawrence Berkeley National Laboratory’s 2024 federal report found that U.S. data centers consumed about 66 billion liters (roughly 17 billion gallons) directly for cooling in 2023.  “The total indirect water footprint of U.S. data centers is nearly 800 billion liters, attributed to water consumed indirectly through electricity use, based on the regional electricity grid mix for U.S. data center locations.” Roughly 80 percent of a data center’s total water footprint is indirect, tied to power generation rather than on-site cooling. The practical implication is important: for many facilities, the single most effective way to shrink the water footprint is to power them with electricity sources that use little or no water, such as wind, solar, or gas paired with dry cooling. A facility’s electricity source can matter as much to its water profile as its cooling equipment.

Understanding the cooling technologies a project may present

When a developer describes how a proposed data center will be cooled, the choices fall along two axes: where the heat is captured (at the chip, at the server, or across the room) and how the heat is ultimately rejected (using water, or using air). Each combination trades water against energy, cost, and complexity. The key ones for a community to understand:

• Air cooling (building level). The traditional approach: fans move cold air across the servers, and the heat is carried away by the room’s cooling system. Air cooling is simple and uses little or no water directly, but it hits a physical wall as computing density rises. The intensely hot chips in modern AI servers generate more heat than moving air can practically remove, which is why purely air-cooled designs are fading for AI workloads.
• Evaporative (water) cooling. To boost efficiency, many facilities pair air cooling with evaporative cooling towers: a portion of water is deliberately evaporated to shed heat, the same principle by which sweat cools skin. This is water-efficient in energy terms but consumes water, drawn most often from the same supply that serves homes and farms. The hotter and drier the climate, the more water it uses.
• Liquid cooling at the chip (direct-to-chip). Instead of cooling the whole room, coolant is piped through cold plates mounted directly on the hottest components, the processors and AI accelerators. This removes heat far more effectively, which is why it has become the leading approach for dense AI hardware. Crucially, it can be run as a sealed loop that recirculates the same coolant, using very little fresh water over the life of the facility.

The central tradeoff runs between water and energy. Broadly, water-based (evaporative) cooling uses more water but less electricity, while dry and air-based approaches use less water but more electricity, sometimes noticeably more in hot climates. There is rarely a free lunch: a facility that eliminates its cooling water may raise its power use, which carries its own (indirect) water cost at the power plant. The right balance depends heavily on local conditions, which is why no single technology is “best” everywhere.

How the major operators are navigating this. The largest companies are pursuing visibly different strategies, which is instructive for communities evaluating a proposal:

• Microsoft has committed that its next-generation designs will use a closed-loop, chip-level system that consumes essentially zero water for cooling: once filled at construction, it recirculates the same water continuously rather than evaporating it. Microsoft says the approach eliminates more than 125 million liters of water per facility each year and has cut its water-use effectiveness from 0.49 to 0.30 liters per kilowatt-hour since 2021. The design is real but still early: it is so far deployed only at its Wisconsin campus, with pilots planned in Phoenix, and it modestly raises power use, the water-for-energy trade in action.
• Google uses direct-to-chip liquid cooling for its hottest AI chips but still relies on evaporative cooling at many sites, and in 2024 consumed about 8.1 billion gallons of water, a reminder that even a leader in efficiency has a large absolute footprint as it grows.
• Meta and others have run their data halls warmer and drier to cut water use in their existing evaporative-cooled facilities, and their newest designs go further still, moving to closed-loop cooling that uses almost no water at all.

The lesson for a community is not to prefer one technology by name, but to ask which approach a specific project will use, and why it fits the local climate and water supply.

What actually determines the outcome: technology, scale, and location

Three things, taken together, determine how much a data center’s water use should worry a given community.

Technology deployment is moving in a water-saving direction, but unevenly. Direct-to-chip and closed-loop designs can cut on-site water dramatically, and the industry is adopting them quickly for AI workloads. But they raise energy use, they are being deployed first at flagship sites rather than everywhere, and older evaporative facilities will remain in service for years. A project’s actual cooling design, not the industry’s best-case direction, is what a community should evaluate.

The overall trajectory is upward, even as efficiency improves. Because water use tracks electricity use, and data center electricity demand is climbing steeply, total water consumption is projected to rise even as per-facility efficiency gets better. Lawrence Berkeley National Laboratory projects that U.S. data centers’ direct water consumption could grow from about 17 billion gallons in 2023 to between 38 and 73 billion gallons by 2028, a two- to four-fold increase. In national terms, this remains a small share of U.S. freshwater use, far less than agriculture, but the growth is real and concentrated.

Location is decisive. A national total can be reassuring while a local situation is alarming, and the local one is where the policy questions live. About two-thirds of the data centers built since 2022 have gone up in water-stressed regions. A single large facility drawing on a strained system in a drought-prone area can place real pressure on that community even if AI’s nationwide water use stays modest. 

The right question is therefore rarely “does AI use too much water” in the aggregate. It is “can this community, in this location, supply this facility, with this cooling design, without shortchanging existing users?” That is a question about siting, cooling technology, water source, and transparent reporting, and it is one a community can actually answer, project by project.

BEYOND THE CAMPUS: INNOVATIONS IN THE DATA CENTER FOOTPRINT

A traditional hyperscale data center requires years of planning and hundreds of acres, and that footprint is itself a major source of community friction. Across the industry, a range of innovations are tackling that constraint from different angles. Some are already running in commercial production today. Others remain genuinely experimental, with real questions still unresolved. Together they illustrate an industry actively searching for ways to shrink, relocate, or entirely sidestep the footprint a data center leaves on the ground.

Modular and Mobile Data Centers

Not all data centers look like the massive campus facilities that dominate news headlines. A fast-growing alternative is the modular, or prefabricated, data center: standardized units built and tested in a factory before being shipped to a site. On arrival, each unit connects to local power and network and begins running, often within weeks of delivery. The smallest deployments fit inside a single shipping container, and larger configurations scale by simply adding modules as demand grows.

This changes who can host digital infrastructure and how fast a project can become viable. Where a hyperscale campus needs years of planning, a modular deployment can be located almost anywhere there is power and a network connection.

Going off Land

For all the attention paid to where data centers get built, a more provocative question is emerging at the edges of the industry: Do they need to sit on land at all? Faced with mounting constraints on land, water, and power, companies are testing designs that would have seemed like science fiction a decade ago: sealed facilities on the ocean floor, platforms floating at sea, and computing hardware in orbit. None of these are mainstream yet, but some have already crossed from experiment into commercial operation.

Data centers are the factories of the digital economy, and they have been for decades. They are not simply AI infrastructure. They underpin the financial system, the healthcare system, public safety, supply chains, and the daily communications of modern life.

In the U.S., data centers are expected to consume more electricity for processing data than all energy-intensive manufacturing combined by 2030, including aluminum, steel, cement, and chemical production. 

At the same time, the industry is innovating at a pace that makes today’s concerns a moving target. 

Zero-water cooling is no longer a concept — it is being deployed. Modular facilities are making it possible to bring digital infrastructure to communities that could never host a traditional campus. And the long-term relationships in places like Delaware and Mecklenburg County show that when communities negotiate clearly, hold companies to commitments, and engage as genuine partners rather than passive hosts, the outcomes can be durable and beneficial.

The communities best positioned to navigate this moment are those that go in informed, understanding both what data centers actually are and what good agreements actually look like.

Data centers are not decided in Washington. They are decided in county commission meetings, in planning board hearings, on utility commission dockets, and in school board discussions about future workforce needs. The decisions happening right now in communities across the country will shape what gets built, where, on whose terms, and who benefits. The sections below are designed to help you move from informed reader to active participant.

Drive Awareness and Education

We rely on data centers every time we send a message, stream a video, use a navigation app, receive medical test results, or pay with a card. Hospitals, emergency systems, and public safety infrastructure all run on them.

Yet the debate now unfolding in communities across the country often gets stuck. Projects are either waved through without scrutiny or met with reflexive opposition, when what is really needed is an informed conversation that weighs both the benefits and the genuine tradeoffs.

Helping your community have that conversation is where you can make the biggest difference.

Host a Policy Circle on data centers in your community. The Policy Circle’s roundtable format is designed exactly for this kind of conversation: bringing together informed citizens, local officials, business owners, and subject matter experts to work through a complex policy issue together. Consider inviting a local official, a representative from a local utility, or a representative from a data center company to join the conversation. Learn how to start your own Circle and get support from The Policy Circle as you get started. 

Share this Policy Circle Brief with your local officials and business association. Many local decision-makers are navigating data center proposals without access to information that covers both the opportunity and the tradeoffs. You can be the person who provides that.

Strengthen Communities

A data center that treats a community as a partner rather than a host creates measurably different outcomes than one that simply arrives, extracts, and departs.

Advocate for community benefit agreements (CBAs). A CBA is a legally binding contract in which a developer commits to specific, enforceable benefits in exchange for community support. Done well, it turns vague promises into measurable obligations: local hiring, contributions to schools and emergency services, environmental monitoring with public reporting, infrastructure investment, and decommissioning guarantees if the operator leaves.

In late 2025, the city of Lancaster, Pennsylvania negotiated a CBA with the developers of a data center campus before construction was approved.  The agreement shows what a community can secure when it negotiates specifics in writing rather than accepting general assurances.

The most important point about timing: advocating for your county to adopt a CBA requirement before the next proposal arrives is far more effective than trying to negotiate one under deadline pressure after a developer has already applied.

Connect data center workforce opportunities to local education pipelines. As the Uptime Institute and CBRE’s 2025 North America Data Center Trends reports both document, the single biggest constraint on data center construction and operation in 2026 is not power or land but trained workers. Data centers employ electricians, HVAC technicians, network engineers, cybersecurity specialists, facilities managers, and operations staff, many of whom do not require four-year degrees but do require specific certifications and training. If a data center is coming to your community, the question to ask immediately is: where will the workforce come from, and can it come from here? Connect your local community college, high school career and technical education programs, and workforce development board with the operator before construction begins, not after.

Identify who is and is not in the room. When a data center proposal arrives in your community, developers will have already been speaking with local officials, utilities, and economic development agencies for months. Find out who is representing the community’s interests, what they know, and where the gaps are. Your county’s economic development office, planning department, and elected supervisors are the starting points. You should also ask who is not in the room, such as low income or disadvantaged communities, and make sure they get a seat at the table and that their interests are represented. 

Build long-term relationships, not just one-time responses. The most effective community advocates are people who build ongoing relationships with local officials, utility representatives, and economic development agencies, who follow proceedings consistently, who share information with neighbors between meetings, and who position themselves as trusted, informed voices before any specific controversy arrives. That is the kind of civic presence that shapes outcomes, not just responses.

Reach a Position of Influence

The most consequential conversations about data centers are happening at the local level, in the very bodies that decide whether a project gets built and under what conditions. Getting a seat at that table before a specific project arrives is far more powerful than showing up after a decision has already been made.

Apply for The Policy Circle’s Civic Leadership Engagement Roadmap (CLER) Program to join a cohort of like-minded women learning better skills to be effective business and civic leaders in their communities.

Find your local planning and zoning commission, and consider becoming a member. Every county and municipality has one. Planning commissions review land use applications, make recommendations on zoning changes, and hold public hearings on major development projects including data centers. To find your local planning commission, search your local government website or reach out to your local municipal officials. Terms are typically several years and require no technical background – only genuine civic interest and a willingness to engage.

Look for data center task forces being formed in your area. In some areas, task forces are formed at the city or county-level. For example, Prince George’s County, Maryland established a formal Data Center Task Force in April 2025 specifically to evaluate energy demand impacts, environmental effects, and quality of life considerations near proposed facilities. Linn County, Iowa developed a data center ordinance through a public planning process in early 2026, with open commission meetings and public comment periods at every stage.

Identify who is already in the room. When a data center proposal arrives in your community, developers will have already been speaking with local officials, utilities, and economic development agencies for months. Find out who is representing the community’s interests, what they know, and where the gaps are. Your county’s economic development office, planning department, and elected supervisors are the starting points.

Impact Legislation

Policy made without informed community voices tends to serve whoever showed up. Lawmakers at the state and local level are writing data center legislation right now, often without sufficient understanding of the tradeoffs between economic development, energy demand, water use, and community impact. 

Your role is to help them understand the full picture.

Know what legislation is active in your state. Data center policy is moving fast. States including Illinois, New York, Maryland, California, and Virginia have introduced or passed legislation on data center water reporting, energy cost allocation, environmental review, and tax incentives in 2025 and 2026. The National Conference of State Legislatures tracks energy and technology bills by state. Your state legislator’s website will list their committee assignments, and the relevant committees are typically those covering energy, environment, commerce, economic development, or technology.