Many claims are made about data centers and why we need them.
This page looks at talking points that come up often in the debate over data center development in West Virginia, from supporters and opponents alike. For each one, we note was is accurate and attempt to clarify any misconceptions.
Clarifying common talking points
"If you use technology and the internet but oppose data centers, you are a hypocrite."
Nearly everything people do online, from banking and streaming to telehealth and email, runs through data centers. That demand is real, and anyone who uses these services relies on that infrastructure.
Going deeper
Most local debate is not about whether data centers should exist. It is about which kind is proposed, at what scale, in what location, with what conditions and accountability measures, and who bears the costs. Data centers range from a server room to campuses drawing hundreds of megawatts, and these differ greatly in their effects on surrounding communities and ecosystems.
Relying on a service does not settle the question of where or how the infrastructure behind it should be built.
The "hypocrite" framing implies that data center development is a binary decision: no data centers or data centers. However, a significant part of the current opposition lies in how these facilities are being planned (behind closed doors with no local input or control) and in where they are being planned (on active farmlands, in close proximity to protected land and sites, in close proximity to historical landmarks, schools, churhces, and residential communities, and in communities with a pre-exisiting disproportionate burden from extractive industry). Much of the planned growth now is happening quickly with little to no deliberation of potential impacts to the surrounding communities, environment, or ecosystems. This speed and lack of adequate review leaves many communities feeling uncertain about the future of places they call home.
"Data centers are inevitable, so we should accept them."
Demand for data center capacity is growing quickly. Lawrence Berkeley National Laboratory estimates that U.S. data centers used 4.7% of the nation's electricity in 2024 and projects 9.5% to 15.3% by 2030.
Most of that growth comes from AI. Berkeley Lab finds that rising electricity use is driven primarily by growth in both the number and the power of AI servers shipped each year. In its central projection, AI servers account for 84% of all server energy and 55% of total U.S. data center electricity by 2030. The question of whether this growth is inevitable is therefore largely a question about AI.
Going deeper
Researchers and policymakers have argued that deployment of the technology data centers are built to power is outpacing the study of its consequences, and some warn about risks that would be difficult to reverse once systems are widely in use. The United States has no comprehensive federal law governing how AI is developed or used, and questions about privacy, bias, labor, misuse, and accountability remain unsettled. While AI has the potential to offer significant benefits to society, it could also cause significant harm -- and that balance is a matter of human choice.
Inevitability implies technology is an entity that evolves independent of human choice. In practice, it's growth is determinted by the companies making design, development, and investment decisions, and by the public policies that govern them.
With regard to data centers, people across the country are fighting against the enchroachment of these massive facilities in their communities and lawmakers are responding to the concerns of their constitutients. There are over 800 advocacy groups in 49 states, and according to a March 2026 Gallup Poll, 71% of Americans are opposed to AI data center construction in their community. This opposition crosses political lines with a majority of all Democrats, Independents, and Republicans polled either somewhat or strongly opposed.
Governments at every level are acting on these concerns. Local governments have adopted moratoriums to pause development while they write rules, and others have adopted ordinances addressing setbacks, noise, water withdrawals, and decommissioning. About 30 state legislatures have introduced or adopted rules on data center siting, electricity, and water use. These choices will determine whether data center development proceeds responsibly and sustainably, or whether decisions are made without adequate foresight into potential harms.
To call data center development inevitable assumes that real people are not behind the decisions about where and how they are built, and implies that communities must accept data center development at any cost without any meaningful say. It also ignores the possibility of shifts in technology usage, like when software and hardware become more efficient overtime and require less space and energy usage. Even the projections shared here do not model shifts in technology, markets, regulation, or grid capacity — any of which could change the outcome.
Source: Berkeley Lab, United States Data Center Energy Usage Report: 2025 Update (2026)
"Data center companies contribute to the surrounding community."
Data centers pay property taxes, and some developers make additional commitments such as local donations, infrastructure improvements, support for fire departments and schools, or training programs. Some follow through on these commitments, and the tax revenue a large facility generates can be substantial.
Going deeper
Commitments made in press releases and flyers are good-faith statements, not obligations. What makes a commitment enforceable is a community benefit agreement: a negotiated, binding contract between a developer and the host community, signed before approval. Elsewhere, these agreements typically cover local hiring targets, water and noise limits stricter than permit minimums, funding for schools and emergency services, road maintenance, independent monitoring, decommissioning bonds, and penalties if commitments are not met.
Community benefit agreements get signed because the community holds something the developer needs, such as a rezoning, a variance, or another discretionary approval. HB 2014 removed that leverage of local communities to negotiate a community benefit agreement for certified projects. It exempts them from county zoning, land use, and permitting ordinances, and certification is based on meeting the law's definition of a high-impact data center, not on any community commitments. There is no point in the process where a county can say "yes, if." That is not a failure of local officials; it is how the statute is built.
HB 2014 also changes who receives the tax revenue. For certified projects, 50% goes to the state's personal income tax reduction fund, 30% to the host county, and 10% is split among the remaining counties by population, 5% Economic Enhancement Grant Fund administered by the Water Development Authority, and 5% to the Electric Grid Stabilization and Security Fund. Without HB 2014, most of that revenue would have stayed with the host county, which also bears the local costs of roads, emergency response, and other services.
"Data center growth is driven by everyday consumer use."
Consumer services such as streaming, social media, and cloud storage are a long- standing source of data center demand, and consumers increasingly use AI tools directly.
Going deeper
Berkeley Lab attributes most recent and projected growth to servers built for AI. In its central projection, AI servers account for 55% of all U.S. data center electricity by 2030, and most of that AI power goes to running models for users (inference) rather than training them. Public data does not break this demand down by who the end user is, whether consumers, businesses, or government, so the share driven by each cannot be stated with confidence.
Source: Berkeley Lab, 2025 Update (2026), Sections 3.4.3 and 3.6
"Data centers use little water, especially with closed-loop cooling."
Closed-loop systems can sharply reduce the water a data center consumes on site. In a closed loop, a liquid coolant circulates through sealed pipes, often directly to plates on the chips, absorbs their heat, and returns to be cooled and reused. The coolant itself is not consumed, apart from an initial fill and occasional top-ups. Nationally, Berkeley Lab estimates that U.S. data centers consumed about 17 billion gallons of water on site for cooling in 2023.
Going deeper
"Closed loop" describes the inside of the system. The heat still has to leave the building, and how it leaves determines how much water is consumed:
- Dry coolers release heat to the air through radiators and fans. They consume almost no water but use more electricity, especially in hot weather.
- Cooling towers release heat by evaporating water. They use less electricity, but the evaporated water is consumed.
- Hybrid or adiabatic systems run dry most of the time and spray water only on hot days, so their consumption depends heavily on climate.
Water consumption also depends on local temperatures, how many hours a year a system relies on evaporation, and the quality of the water supply, which affects how often water must be flushed and replaced.
The tradeoff is between water and energy. Saving water on site usually means using more electricity, and generating electricity consumes water too. Berkeley Lab estimates that in 2023, the power plants supplying U.S. data centers consumed about 211 billion gallons, roughly 12 times what data centers consumed on site. Dividing both by the 176 terawatt-hours data centers used that year gives national averages of about 97 million gallons consumed on site and 1.2 billion gallons consumed at power plants for every terawatt-hour (TWh).
Most importantly, water effects are local. A national total that is small compared with agriculture or power generation says little about what a single large facility means for a particular aquifer or stream, especially during drought.
Source: Berkeley Lab, 2024 United States Data Center Energy Usage Report
"We need more data centers if we want to keep using the internet."
Online services depend on data centers, and as those services grow, so does the need for computing capacity.
Going deeper
The current build-out is driven mainly by AI, not by conventional internet services. In Berkeley Lab's analysis, growth in U.S. data center electricity comes primarily from the number and power of AI servers shipped each year. Hardware does become more efficient with each generation, but the report finds that growth in computing demand has more than offset those gains. Whether facilities on the scale now proposed are needed to maintain everyday internet use is a different question from whether they are needed for AI expansion.
"Data centers are a matter of national security and the AI race with China."
What is accurate
AI capability is widely treated as a matter of economic and national security, and computing capacity is a large part of it. West Virginia officials have framed HB 2014 in these terms and Governor Morrisey said it would help the country compete with China in the technology race.
Going deeper
Where the two countries stand. The United States holds a large lead in AI computing capacity. As of May 2025, Epoch AI estimated that the U.S. hosted about 75% of the world's AI supercomputer performance, with China second at about 15%. The U.S. also leads in private investment ($109.1 billion in 2024, compared with $9.3 billion in China) and produced 40 notable AI models in 2024 to China's 15. The quality gap has narrowed sharply, however. Stanford's AI Index found that performance differences between leading U.S. and Chinese models on major benchmarks shrank from double digits in 2023 to near parity in 2024. China also leads in AI publications and patents.
Constraints on the U.S. build-out. Berkeley Lab's estimates imply that U.S. data centers will need about 148 gigawatts of grid connection capacity by 2030, an increase of about 17 gigawatts a year. The report identifies grid capacity, supply chains, workforce availability, and water resources as factors that may limit growth. The most advanced chips also depend on a small number of overseas manufacturers.
Export controls. Beginning in 2022, the Biden administration restricted sales of advanced AI chips to China, citing national security. The Trump administration banned sales of Nvidia's H20, a chip designed to comply with those rules, in April 2025, then allowed them in August in exchange for 15% of the revenue. In December 2025, it went further, permitting sales of the more powerful H200, which had been blocked under Biden-era rules, to approved Chinese customers in exchange for 25% of the revenue. Nvidia's newest chips remain restricted. Supporters called the policy a balanced compromise. Critics, including former national security officials and members of Congress, said it would help Chinese firms close the gap with U.S. models.
National security arguments apply to AI capacity in general, not to any particular site. How the U.S. builds that capacity, and what it allows competitors to buy, bear on the outcome more than how much it builds.
Source: Epoch AI, AI supercomputer performance share by country (May 2025)
Source: Stanford HAI, AI Index 2025
Source: Berkeley Lab, 2025 Update (2026), Section 2 and Conclusion
"These projects will bring good-paying, permanent jobs."
Data centers create construction jobs during the build and a smaller number of permanent, often well-paid positions. The Bedington Project announcement estimates more than 1,000 temporary construction jobs and more than 125 permanent jobs, and the developer's flyer lists pay of $90,000 to $110,000 for those positions.
Going deeper
Job numbers are estimates, and it helps to know what kind of estimate you are hearing. Announcement figures usually come from the developer, sometimes from economic impact models that add "indirect" jobs (at suppliers) and "induced" jobs (from local spending) to the job count reported at the facility. A widely cited U.S. Chamber of Commerce study, for example, found that a typical large data center employs 1,688 local workers during construction and supports 157 local jobs a year once running, but that 157 includes jobs along the supply chain, not only staff at the facility. It is fair to ask which kind of job is being counted, and over how many years.
What kinds of jobs. During construction, these are skilled trades such as electricians and pipefitters, often specialized crews that travel between projects. Construction jobs are temporary and spread across the build. Once a facility is running, permanent jobs include facility technicians, electrical and mechanical maintenance, security, and operations staff. Staffing does not scale with size: a facility drawing ten times the power does not employ ten times the people.
Is any of it binding? HB 2014 certification is based on a facility's power capacity, not on job commitments. Without a community benefit agreement, nothing requires a developer to meet its job estimates or to hire local residents, and no West Virginia project has one. At a Martinsburg Chamber of Commerce event in May 2026, the economic development director of Loudoun County, Virginia, home to one of the largest concentrations of data centers in the world, said data centers did not create a large number of jobs there; the benefit was tax revenue. Loudoun controls its own zoning and keeps its own tax revenue, the two things HB 2014 changes here.
The questions that matter are: how many permanent jobs, what share go to local residents, at what wages, and is any of it binding?
"Residents voted against zoning, so they chose this."
West Virginia law allows counties to put a zoning ordinance to voters, and voters can reject it. Where countywide zoning has not been adopted, land use controls are more limited.
Source: W. Va. Code § 8A-7-7, election on a zoning ordinance
Going deeper
For projects certified under HB 2014, local zoning would not apply either way. The law exempts certified microgrid districts and high-impact data center projects from county and municipal zoning, land use, noise, lighting, and permitting ordinances, and bars counties from enacting rules that would limit them. A county's past vote on zoning therefore does not determine whether it could regulate a certified project today. Jefferson County, which has zoning, is subject to the same exemption.
"The opposition to data centers is funded by China."
The claim has been made publicly. In June 2026, Republicans on the House Energy and Commerce Committee said there was "strong evidence" that foreign influence linked to China was contributing to data center opposition, citing a think tank report that described Chinese state media and some advocacy groups as part of a broader ecosystem opposing AI infrastructure.
Going deeper
Reporting by NPR found little direct evidence that local opposition groups are funded by China, and the groups named have denied it. Opposition is also widespread: a Gallup poll found that 71% of Americans are somewhat or strongly opposed building data centers in their own communities. We are not aware of any foreign funding for opposition in the Eastern Panhandle or any part of West Virginia.
Source: 119th Congress House of Representatives Committee on Energy and Commerce (June 4, 2026)
Heard something we have not addressed?
Send it to us. We will look into it and add it here. [email protected]