Key Points
- While data center growth can increase system demand in the future, it can also enable new energy supply for all ratepayers.
- The environmental impact of data centers is comparable to, or even less than, that of other industries in similar locations.
- Local communities see real benefits from data center investment; those benefits can be maximized when coupled to broad-based tax policies that treat inputs and investments for data centers as well as they do for other businesses.
- Taxpayers have a stake in the policy debate surrounding data centers as ratepayers, workers, and users of the digital economy.
Introduction
Pete Sepp, NTU President
What should taxpayers think about data centers? The question is not an academic one, since the facilities that house the Internet-based architecture for the next big leap in how technology will shape our economy and society are being built in increasing numbers right now. Part of the answer to this question is rooted in the past. After all, like factories, warehouses, office buildings, and entire business zones, data centers have both general and specific needs in order to contribute to the nation’s economy.
What is a data center? Simply put, it is a structure that houses computers, memory storage, telecommunications equipment, and energy sources (backup or sometimes primary) to facilitate the flow of information—perhaps the most important element in our economy today, since data drives virtually everything Americans do. Whether it is manufacturing automobiles, supporting online retail transactions for groceries, enabling medical researchers to develop lifesaving drugs, giving children and adults instruction online, or simply allowing friends and relatives to stay in touch, the fast, secure flow of information undergirds it all.
Taxpayers should demand accountability, transparency, and fiscal responsibility from both public officials and private sector actors when it comes to any economic development, and data centers should be no different. Governments at all levels have provided productive (and sometimes, less productive) policy paths forward on data centers. This report is intended to provide lessons that taxpayers across the nation can take to heart as they work to ensure that data centers leave us all better off in the next big step in America’s technologic and economic evolution. It can be done, because it is already being done in many parts of the country. Let us move forward, thoughtfully and confidently, to the next successful chapter in American economic history.
Energy and Environment
Data centers can increase both the supply and demand for energy.
The amount of energy that data centers consume is important to taxpayers, and rightfully so. With the Energy Information Administration estimating that there are nearly 150 million residential energy customers in the U.S.—similar to the number of American households for tax purposes—nearly every taxpayer is also a ratepayer. As a result, the growth in data center construction has many taxpayers concerned about how their power bill will be affected. While the energy industry functions within an imperfect market, supply and demand will still have a balancing effect on power costs.
At a basic level, residential energy prices are determined by local supply and demand. When data center developers build within a community, the extra demand on the community power supply can put pressure on household electricity prices to increase. Despite this, data center development can actually lower prices when developers plan responsibly and operate within a sound regulatory framework that encourages supply.
Data center energy usage varies depending on the size of the facility and its use. Estimates show that, in 2024, data centers in the U.S. accounted for approximately 4.4% of annual electricity consumption. A single data center can use anywhere from 1 to 100 megawatts (MW) of electricity depending on its size and purpose, with hyperscalers (large data centers housing hardware for one company) using 100 MW or more.
In a typical data center, 50% to 70% of energy use is for computing, servers, and storage. For less efficient data centers (those operating at a power usage effectiveness (PUE) of ~2) nearly 40% of power consumed goes toward cooling. Some hyperscale centers focused on efficiency have reduced cooling down to only 7% of energy consumption, at a much lower PUE.
The Institute for Energy Research compared data from the Energy Information Administration on electricity prices and sales to the states with the most data centers. It found no relationship between the number of data centers and current electricity prices. According to the analysis, “the top 10 data center states (Virginia, Texas, California, Illinois, Ohio, etc.) average 14.46¢/kWh in 2025, virtually identical to the 14.39¢/kWh average for all other states.” In terms of the speed with which rates have increased in the last decade, there is a minor correlation between the states with the most data centers and an accelerated price change.
Even more recently, an April 2026 update from the Lawrence Berkeley National Laboratory (LBNL) of its closely-watched annual “Retail Electricity Price Trends and Drivers” report demonstrated that a variety of complex factors contribute to electricity costs, often in ways that contradict the narratives of data center opponents. Nationally, the single biggest driver of retail price increases between 2024 and 2025 was the simple base costs for fuel; 30 states saw a jump of 2% or more due to this factor. The report also noted that states in the PJM grid area (the mid-Atlantic and Midwest) saw average wholesale costs rise by 0.9 cents per kilowatt hour in 2025, with another 0.6 cents projected for this year.
While the PJM area includes mid-Atlantic and Midwest states, Virginia, which has the largest number of data centers in the country, was not part of this trend. New England states also continued to experience relatively high prices compared to the rest of the country, despite the fact that data centers tend to be rare there outside of Massachusetts. One of the most important findings was that “[s]tate-level load growth [i.e., demand for electricity] was associated with declining all-sector average retail prices in recent decades, including from 2019 to 2025 in most states.” From 2019–2025, LBNL estimated that the highest load-growth states saw average all-sector prices drop by over 1 cent per kilowatt hour. Beneath their surface, these trends make sense. Greater demand pushes expanded grid capacity, and large power customers (such as data centers) shoulder a greater share of fixed upgrade costs that might otherwise fall on residential customers.
In areas where electricity prices increase, it is often a symptom of our nation’s aging power infrastructure and the inability to increase supply despite abundant energy resources. Much of the U.S. electrical grid was developed in the 1970s and is approaching or surpassing its intended lifespan. Furthermore, grid systems have an extensive backlog of new renewable and fossil fuel sources that wait years for regulatory approval to come online while coal and other power plants retire. In some areas, new power sources are taking eight years to come online, far longer than in 2008, when the period to bring those sources to generation was less than two years.
There are two mechanisms that can lower residential energy prices in areas with increased data center development: diffusion of fixed costs and “behind the meter” power generation. Diffusion of fixed costs occurs when an electrical grid gains new customers, spreading costs out among a broader consumer base and reducing prices. “Behind the meter” generation allows data centers to operate a power source independently, leaving local grid prices unaffected.
With local utilities struggling to bring new power sources online, regulators in certain areas now allow data center developers to generate energy “behind the meter,” operating separately from the grid. These developers may choose to operate natural gas turbines, fuel cells, solar power, or other systems to meet their own needs.
The interplay between supply and demand in the grid is best demonstrated through contrasting two state examples, Pennsylvania and North Dakota. Pennsylvania’s energy generation and consumption is largely dictated by PJM Interconnection, the nation’s largest grid operator. With PJM’s eight-year interconnection waitlist, per-megawatt day energy prices in the region have increased nearly tenfold from 2024 to 2025 as energy use in the area has surged. Meanwhile, Pennsylvania is the nation’s second largest natural gas producer and second largest net supplier of energy to other states.
In contrast to Pennsylvania’s inability to keep up with increased demand, data center development is decreasing energy prices in North Dakota. While this is partially due to diffusion of fixed costs, it is further encouraged by investment from developers. Some North Dakota developers have tapped into underutilized energy sources that are already online, resulting in increased public utility profits that are then passed directly to consumers via a transmission credit program.
Utilities in other parts of the country are being made whole in their grid investments thanks to data center energy development. A recent independent economic analysis performed for Amazon showed that, in certain regions, their projects actually generated a net financial surplus. For example, Entergy Mississippi’s CEO recently noted that his company was making a $300 million grid investment as a direct consequence of this windfall:
Typically, these kinds of large-scale upgrades would translate to higher electricity bills for our customers. But thanks to the influx of new customers like Amazon coming to Mississippi, we’re able to fund these critical reliability improvements without passing any added costs on to our residential and small business customers. It’s a true win-win: We’re delivering a more robust, resilient grid, while ensuring our rates remain well below the national average.
Meanwhile, Entergy Louisiana is currently finalizing an arrangement with Meta in Richland Parish for a massive data center campus that could be Meta’s largest such project yet. As a bonus, Entergy projects $2 billion in relief for ratepayers over a 20-year cycle.
Difficult-to-predict data center energy demand growth makes the case for new supply even stronger.
Certainty over the number of data centers that will be built in the coming years remains a challenge, but reforms to strengthen the grid are necessary regardless. In many parts of the country, critical infrastructure is decades old. Taxpayers should support actions that will improve regional power stability and construction, which, in turn, will improve the stability of revenues for state and local governments. Simplifying the permitting process and expanding options for production will help ensure that both households and businesses have access to consistent, affordable energy.
While there are concrete numbers on how much energy data centers utilize on-site, projections of total energy requirements in the future vary widely from 325 TWh (terawatt hours) to 580 TWh.
The disparity between the estimates can be ascribed to two major factors. The first issue with five to fifteen year projections for data center energy use is how many data centers will be built in the intervening years. Estimates depend upon how long the current AI boom will continue, to what extent demand will increase, and whether technological improvements will decrease energy use. Uncertainty around all of these facets of the future of data center expansion leave projections with many assumptions and less precision.
The other main factor that leads to such disparate estimates of energy use is the role of utility companies and regional transmission organizations in production and delivery. In the industry, it is not uncommon for companies to inflate estimates for a variety of reasons.
The Federal Energy Regulatory Commission (FERC) requires companies to estimate how much energy they will need to produce and sell with restrictions on new construction based on these projections. In 2024, FERC released Order No. 1920, which increased long-term report requirements from 15 to 20 years, meaning an additional five years of conjecture on load and construction needs. If an energy company underestimates its requirements a decade or more before they are needed, it becomes substantially more difficult to obtain permits for new construction.
Thomas Aiello, Vice President for Federal Affairs at National Taxpayers Union, recently described the main issue in the U.S. regulatory landscape:
Lengthy permitting timelines are not the result of strong environmental standards, but of a broken regulatory framework that prioritizes process over results. Projects that meet clear legal requirements are routinely delayed for years due to overlapping agency reviews, unclear timelines, and excessive litigation. These delays ultimately hurt workers, consumers, and taxpayers, while providing little additional environmental benefit.
The process creates perverse incentives to inflate how much energy will be needed to avoid a possible shortage or congestion, yet the instinct to overestimate is better than the alternative. Former FERC Chair Richard Glick was quoted as saying “You don’t want to overbuild. But I would say that the consequences of underbuilding are a lot worse than the consequences of overbuilding.” Whether or not specific projections prove accurate, the cost of insufficient supply far exceeds the cost of building more generation than is ultimately needed.
In the past, generation and transmission companies inflated numbers when perceived paradigm shifts were on the horizon, such as the electrification of appliances and the rapid proliferation of electric vehicles. Some leaders in the production and transmission industry have called out the inflated numbers surrounding data centers. In 2025, the CEO of Vistra Energy, a Texas-based, Fortune 275 energy company, stated that “We think these interconnect queues . . . may be overstated anywhere from three to five times what might actually materialize either in regulated markets or competitive markets.”
Beyond the question of how much energy data centers need is the question of cost and whether tech companies are offloading the cost of power and new infrastructure onto consumers. Fortunately, regulatory tools already exist to ensure that new supply gets built without burdening existing ratepayers, such as large-load tariffs.
A modification of utility tariffs which have been used for decades, large-load tariffs are agreements between utility companies and heavy energy consumers, particularly cryptocurrency companies and data centers. Rather than distributing the costs and risks associated with expanding the grid and interconnection as in traditional utility tariffs, large-load tariffs require the associated customer to pay for a minimum power draw. This encourages utilities to build new infrastructure knowing that there is a guaranteed customer. Some large-load tariffs even include exit fees in the event a customer no longer needs the large load of power.
Private sector entities are actually embracing these policies without heavy-handed government regulations or the threat of tax hikes. As just one example, Google’s Capacity Commitment Framework (introduced in 2025) and Clean Transition Tariff (introduced in 2024) both include protections against cross-subsidization from other ratepayers.
As of this writing, 30 states have approved large-load tariffs for utility companies with proposals awaiting approval in 5 more states. Seven of the approved or pending tariffs cater specifically to data centers and cryptocurrency customers.
Power purchase agreements (PPAs) also provide data center operators an option to contract energy while funding new supply. Typically deriving power from renewable energy sources, PPAs allow large scale consumers to buy longer contracts for stable energy. This is important for data centers that need consistent access to energy as generative AI usage ramps up. PPAs give purchasers peace of mind over periods up to 20 years and funds from agreements can be used by power companies to build out new generation which expands supply for the entire grid.
Even in the unlikely event that the upper limit predictions of data center energy use come to pass, the combination of large-load tariffs, PPAs, behind the meter generation, and other reforms give policymakers a toolkit to expand supply while protecting ratepayers. The benefits of increasing energy supply will not only alleviate challenges of increased demand from data centers, but will also allow households to increase energy usage in a cost-efficient manner when switching to electric appliances or otherwise making residential upgrades.
While the majority of power for data centers comes from fossil fuels today, there is significant traction in the industry to move toward renewables. The biggest tech companies are building hyperscaler locations with PPAs that are completely based on renewable energy. Equinix and Digital Realty, two of the major players in the leasing and colocation (data centers which rent out space for smaller companies to run hardware) industry, are also moving toward greater renewable energy use.
Additionally, nuclear energy is seeing a renaissance largely driven by the demand from data centers. Microsoft was involved in reopening the shuttered Three Mile Island plant and Meta is working with several companies to build out new nuclear capacity.
The environmental impact of data centers is comparable to that of other industries in similar locations.
All too often, taxpayers have been an afterthought in governments’ policies toward the environment. Excessive “precautionary principle”-based regulations on certain substances have endangered the continued development of affordable products—including flooring, medical supplies, and automotive parts—that have made government purchases of structures, health care, and vehicle fleets, among others, less expensive. Government mismanagement of environmental cleanups, some of them caused by agencies like the Department of Defense, has likewise run up the tab for taxpayers. At the same time, public-private partnerships in environmental oversight have tremendous potential for ecological and taxpayer protection, if governments will allow them to happen.
Data center development has three main environmental impacts: land usage, water consumption, and emissions. Environmental impacts are often the most prominent concern for taxpayers in areas with increased development. Across the country, small, vocal groups are pressuring localities not to build data centers. Taxpayers should resist this degrowth alarmism which is largely based on misconstrued claims. While environmental concerns should be taken seriously, the controversy surrounding data centers is unfounded. The impacts of data centers are manageable and comparable to that of other industries. Reasonable stewardship on the part of data center builders can result in cleaner water and fewer emissions. Residents in prospective sites should encourage this stewardship, not oppose construction outright.
Land acquisition is often one of the first major upfront costs of projects and affects how much the campus can grow in the future. While developers are incentivized to locate near population centers to reduce data latency from producer to consumer, they may also tend to locate in more rural areas due to vast land supply at lower cost.
Data center campuses can occupy as little as 10 acres to as much as over 1,000 acres. The average data center in the U.S. occupies about 200 acres, while the average farm occupies just over that amount at over 300 acres. While acreage requirements for industrial factories vary widely, an electric vehicle manufacturing facility typically occupies around 500 acres.
As with any construction for residential or commercial purposes, building in an area shrinks that area’s total supply of land and could raise housing prices depending on the desirability for families to remain in the area. At times, rising home valuations can be due to land improvement from data center development rather than a reduction in the supply of local land. For example, some data centers are constructed on rehabilitated brownfields, which are properties no longer in use due to prior industrial contamination. Federal, state, and local governments can encourage developers to clean up and rebuild these sites.
Water use is the most talked about, criticized, and misunderstood element of the data center and AI boom. Statistics related to water use at data centers vary wildly from source to source. The range in estimates stems primarily from whether one factors in just direct use or includes indirect water use.
Direct use is primarily involved in cooling. The computations data centers perform generate immense heat and evaporative cooling systems help keep the computers running properly. Most of the water used onsite evaporates into the atmosphere, while some is reused. At times, the water looping through the cooling cycle has to be treated when mineral concentrations become too high.
Indirect use includes water consumed offsite; primarily this is in energy generation as natural gas, coal, and nuclear power all use water. Most power generation requires water boiled to steam to turn turbines. Typically this water is drawn from surface sources like rivers and lakes and, in many cases, much of the water is recirculated through the system multiple times.
The amount of water a single data center uses in a year is comparable to water use in most industries. A higher estimate according to the Berkeley Lab Report is that data centers use approximately 628 million gallons of water per day. This is about 25% of the water used in U.S. cotton production in 2023. Brian Potter, Senior Infrastructure Fellow with the Institute for Progress, notes that other U.S. industries use much more water, including “aquaculture at 7.5 billion gallons a day, mining at 4 billion gallons a day, and water for livestock at 2 billion gallons a day.”
The biggest concern surrounding water usage for data centers originates in areas where water scarcity is a common occurrence. It seems counterintuitive to build high-water-use facilities in dry states like Arizona, Nevada, and Colorado, but the capacity for solar in the desert shifts incentives. Using solar power, either on site or from existing facilities, helps companies like Meta and Google hit their climate targets.
For future construction, tech companies should consider area-appropriate development as much a priority as energy prices and climate goals. The agricultural sector in the United States already shirks responsibility by growing water intensive crops like alfalfa, corn, and almonds in arid regions (often calling upon taxpayers to underwrite the process). Use of free cooling, utilizing cold ambient air outside the center, would also help lower water use in northern latitudes.
While data center developers must be responsible water stewards in areas with low water supply, it is important to understand how unique local circumstances can exaggerate concerns nationwide. A viral article in late 2025 accused Amazon of causing harm to the local water supply in eastern Oregon due to a large data center constructed in the area in 2011. Water containing elevated nitrate levels after decades of agricultural runoff is pumped from the local supply to Amazon’s data centers and is then recycled back into the same supply. The outgoing water is more concentrated than the incoming water because of evaporation—but data center cooling itself did not taint the water supply.
As with any development, air quality is also a concern for data centers. Carbon dioxide and other emissions from a data center are almost entirely related to the carbon footprint of the power source it is using, as, unlike agriculture and manufacturing, there is relatively little airborne byproduct of computing. Data centers are estimated to contribute less than 2% of U.S. carbon emissions, compared to 10% for the agricultural sector and 12% for the manufacturing sector. Despite a smaller air quality impact, data centers are highly regulated under the same emissions standards and quotas as other energy-intensive industries.
Investment
Data center development can benefit local communities through infrastructure, services and growth.
Data center developers often share community concerns in the areas where they choose to locate. Aging power and transportation infrastructure and inadequate government services affect not only local taxpayers but also developers attempting to build in the area. Taxpayers should view aspiring developers as partners in improving their locality and generating economic growth.
Physical infrastructure is critical for data center development. This includes roadways to facilitate construction, reliable broadband connectivity, and access to power and water resources. Upgrades to this infrastructure generally benefit the community at large. For example, in Huntsville, Alabama, data center development by Meta improved rural broadband access in surrounding areas.
Corporate social responsibility initiatives often drive philanthropic funding of community events and organizations. Data center developers in Lancaster, Pennsylvania, have agreed to contribute $20 million to local economic development funds. Developers in Henrico County, Virginia, have committed $60 million to a local affordable housing trust. Likewise, Amazon has created the Amazon Community Fund to provide funding for workforce development and training in areas where it has data center campuses.
Economic growth is an important yet understated benefit of development through an influx of full-time jobs. During the construction phase of development, a data center can employ around 1,000 or more workers, a number that dips significantly once the facility is fully operational. During the construction phase, these workers contribute to the local economy through purchasing goods and services. Other jobs, both full- and part-time, can be longer-lasting, but are diffused throughout the supply chain. Server equipment manufacturing is one example of indirect employment provided by data centers.
Tax policy is one piece of a larger toolkit for local leaders to attract data center investment.
Government incentives, including tax exemptions and credits, are meant to lure data centers to otherwise underdeveloped areas. While policymakers can be tempted to use tax breaks and incentives to attract developers, data center developers more often choose their location based on a variety of criteria. Tax incentives can also have negative impacts, such as supporting behavior that would occur anyway without incentives or reducing local tax revenue through tax base narrowing. Taxpayers should be wary of local officials using tax incentives to draw construction of any particular business, especially as other tools may be more effective at attracting data centers.
Some states have long offered incentives to businesses to build data centers, particularly Virginia. As the demand for data centers grows, most states have joined the fray, offering various tax benefits for building both colocation and hyperscaler data centers. The majority of these are sales and use tax incentives toward initial construction.
Many of these incentives were adopted in the late 2000s and 2010s as cloud computing and social media use exploded and the first wave of data centers were constructed. States saw new data centers as a job creator and possible revenue generator. Hoping to attract construction, state lawmakers approved tax exemptions (again, primarily on sales and use taxes) aimed at data centers. Many of these bills required a minimum investment and a minimum number of full time jobs created. Some even stipulated wage minimums equivalent to 100% or 150% of the county or state average.
Even with tax exemptions, data centers still generate local tax revenue. Loudoun County, part of Virginia’s Data Center Alley, recently cut taxes in response to the magnitude of the property taxes collected from data centers. In 2025, 38% of the county’s revenue came from data centers. Projections for 2026 showed data centers would generate over $1.3 billion just in personal property tax on equipment.
Despite these revenues, Virginia lawmakers are considering changing course. The State Senate budget bill for 2026 would end data center tax incentives at the end of the year, accelerating the sunsetting of the provisions eight years early. Legislators in Arizona, Georgia, Michigan, and Maryland have introduced bills in the last twelve months to repeal tax incentives aimed at data centers, some of which are simply portions of the existing tax code for other businesses that were extended to data centers.
There is evidence for both sides of this trend at the local level. Many municipalities introduced property tax exemptions for data centers over the past two decades to bring in jobs and investment. The city council of Abilene, Texas, approved an 85% property tax abatement for the data center developer Crusoe in 2025. Earlier this year, the city of Marysville, Ohio approved a 100% tax break over 15 years for a data center which is projected to be completed in late 2034.
Others, meanwhile, are either repealing those exemptions or are prohibiting outright the siting of data centers within their jurisdictions, often with state government level concurrence. Arizona Governor Katie Hobbs turned heel in her State of the State address, speaking out against the tax exemptions for data centers that she supported previously in the state legislature. Temporary bans on further construction are being considered in 11 states.
Niche tax incentives increase complexity of the tax code and shift the tax burden. The best tax incentives are transparent, low rates applied broadly across all taxpayers, or provisions that benefit all business activity, such as full and immediate expensing of capital investments. States and localities may not need to continue offering special tax treatment to data centers, and could instead focus on clarifying that they qualify for sound tax policies provided to other types of economic activity, e.g., sales and use tax exemptions for equipment purchases or exemptions for business tangible personal property in capital-intensive operations. As competition for preferred sites increases, some companies have started committing to not taking advantage of certain tax breaks.
In early 2026, Microsoft announced it was declining tax breaks in St. Joseph County, Indiana, including all property tax abatements. This is part of a broader pledge from the company to not seek any utility deals or local tax breaks, though Microsoft leases many facilities instead of building them. Google, as part of the Capacity Commitment Framework mentioned earlier as well as other programs, is also working toward broader-based tax and regulatory policy that can be a model for all business development, not just data centers.
Some states report that data centers are not taking advantage of industry-specific tax breaks and instead prefer broad-based tax incentives. Florida has over 100 data centers but reports that none have applied for the state’s sales and use tax exemption for data centers. Likewise, few companies in Oklahoma have taken advantage of the state’s industry-specific tax breaks, instead preferring place-based and workforce-focused incentives programs. Google recently announced that it will develop a new data center in Oklahoma and will partner with local universities for workforce development initiatives.
If state and local tax abatements taper off, companies building data centers may slow other community contributions like workforce development and direct infrastructure investment, but the long-term effects could actually remain positive. Revenue from taxes could make up for these losses and, in many municipalities, would likely enable officials to lower property taxes across the board.
Outlook for Taxpayers
Taxpayers have a stake in data center development as ratepayers, workers, and users of the digital economy.
Taxpayers will continue to have a stake in data center development as ratepayers, workers, and users of the digital economy. Consistent tax, budget, and regulatory actions would be the key to establishing a policy environment that all business sectors, including the tech sector, could rely upon in the future to thrive. Structural safeguards such as Truth in Taxation reforms, limits on government tax and expenditure growth, and conformity with federal expensing provisions could all contribute to this healthy climate.
Data centers are, and will remain, a major policy topic for the foreseeable future. The direction of our economy, society, and national security depends on their development. In his most recent State of the Union address, President Trump himself announced a “Ratepayer Protection Pledge” that numerous tech companies have signed. According to Trump:
We have an old grid. It could never handle the kind of numbers, the amount of electricity that's needed. So I’m telling them, they can build their own plant. They’re going to produce their own electricity. It will ensure the company’s ability to get electricity, while at the same time, lowering prices of electricity for you [consumers and taxpayers].
On their own, the President’s words guarantee a prominent place for data centers in the policy debate. Combined with the actions of state and local governments, taxpayers will have a major stake in ensuring this conversation includes their voices, and more importantly, protects their interests.