Research Note: The Societal Impact of Data Centers
Introduction
Data centers are becoming foundational infrastructure for the next phase of economic, scientific, and civic life. They power not just “high tech” items like cloud computing, AI, and cybersecurity, but also bread and butter “blue chip” industries like finance, logistics, healthcare, and the software tools that are becoming increasingly ubiquitous in everyday office settings. But the current buildout is also creating serious local burdens: electricity demand, water consumption, diesel backup pollution, noise, land-use conflict, and ever-higher burdens on the ratepayer. This research note will argue that the goal of data center policy should not be to stop data centers, but rather to make them pay the full cost of their footprint, locate them intelligently, and require cleaner, more efficient operation.
The Cost of Data Centers
Modern digital systems are no longer just websites and file storage. Data centers now host the computational backbone for artificial intelligence, enterprise software, scientific modeling, national security workloads, streaming, banking, logistics, and public cloud services. AI especially changes the scale of demand because model training and inference require dense GPU and TPU clusters, advanced cooling, high-capacity power delivery, and low-latency networking. The uses of data centers are therefore real. Without data centers, there is no scalable AI ecosystem, no modern cloud economy, no large-scale digital public infrastructure, and limited capacity for advanced modeling in medicine, climate, defense, education, and manufacturing. The key question is not whether society needs them, but whether the benefits justify the costs under current development patterns. The International Energy Agency projects global data-center electricity consumption will roughly double from about 485 TWh in 2025 to 950 TWh in 2030, reaching around 3% of global electricity demand; AI-focused data centers grow even faster, tripling over that period.1 For comparison, the total electricity usage in the United States in 2025 was roughly 4,000 TWh,2 and the Lawrence Berkeley National Laboratory estimates data-center energy use rose from about 60 TWh in 2014–2016 to 176 TWh in 2023, coming to comprise 4.4% of total U.S. electricity consumption, with projected 2028 use between 325 and 580 TWh, equal to 6.7%–12% of U.S. electricity consumption.3 These numbers are staggering.4 In the United States, although our total energy production is largely keeping pace with exploding demand, data center clusters in places like Pennsylvania, Virginia, and Texas with aging energy infrastructure already nearing capacity threaten to raise electricity rates and make brownouts a looming reality.
Are There Societal Benefits?
Proponents of data centers claim that they can produce real economic gain, in the form of large capital investment, construction employment, property tax revenue, and infrastructure modernization. However, these claims are highly circumstantial and often transient. For example, the National Conference of State Legislatures notes that although hyperscale data centers can support more than 1,500 construction jobs, only around 10-30 permanent positions result. 5 Another purported benefit to communities is tax revenue. The NCSL reports that data centers now pay enough in property taxes to fund the entirety of Loudoun County, Virginia’s general operations budget.6 However, Virginia incentivizes data center development by exempting their services from sales tax. Loudoun County does not disclose precisely how much revenue this nets, but claims that it comprises 38% of the general revenue collected, which would put the total at approximately $1.1 billion. A rough approximation of the average sales of a 10MW data center at Loudoun County’s 6% sales tax means that Loudoun County residents subsidize data centers there to a tune of 4-6 times (that is, $4 to $6 billion) what those data centers return to the community. This analysis does not include impacts to electricity or water utility rates. So while there may be some positive externalities in the form of temporary employment and potential local capital outlays, the economic benefits are likely to be uneven and inordinately weighted to the immediate term. This means policymakers should be wary of evaluating data centers like traditional industry. They may be valuable infrastructure, but they are not major long term employment engines or revenue centers.
Resource Burdens and Local Harms
Data centers carry the potential for major negative externalities and direct harm to local communities and the environment. The largest burden is electricity. AI servers, GPUs, storage, networking, cooling, and backup systems impose concentrated demand on local grids. Berkeley Lab states that AI hardware ended the era of relatively flat data-center energy use, with demand more than doubling between 2017 and 2023. That creates three risks: more fossil generation in the near term, higher grid upgrade costs, and ratepayer cross-subsidies if utilities socialize infrastructure costs across households and small businesses. Water is the second major burden. Cooling systems vary, but water-cooled chillers and evaporative systems can consume substantial local water. Berkeley Lab notes that water-cooled chillers are efficient but raise concerns about local water consumption and availability, while some liquid-cooling designs are energy-efficient yet still consume substantial water when paired with evaporative systems. The World Resources Institute reports that mid-sized data centers can use up to 300,000 gallons of water per day, while large facilities can consume up to 5 million gallons daily, and that AI-related U.S. data centers could require up to 32 billion gallons annually by 2028.7 Air pollution is the third major concern. Many facilities rely on diesel backup generators, which are tested regularly and used during outages. Washington State’s Department of Ecology states that data-center generators emit diesel particles and nitrogen dioxide, both of which can harm respiratory health, especially for people with asthma or frequent exposure.8 Noise and quality-of-life impacts are also becoming increasingly visible. The Environmental and Energy Study Institute notes that data centers can generate noise from servers, cooling equipment, backup diesel generators, and onsite power plants; industrial diesel generators can reach very high noise levels, and some facilities are located close to homes, schools, and parks.9 These harms are not abstract. They fall on nearby residents, often without giving those residents proportional economic benefit.
Policy recommendations
This note concludes with a list of eight policy recommendations that would significantly ameliorate the damages caused by data centers, allowing the United States to protect residents and the environment without compromising economic advantages.
- Require full resource disclosure before approval
- Every proposed facility above a 1MW threshold should disclose projected electricity demand, peak load, water use, cooling method, backup generation type, expected emissions, noise profile, and grid-upgrade requirements. This should be public, standardized, and updated annually after operation begins.
- Make data centers pay for grid upgrades
- Policymakers should prevent residential and small-business ratepayers from subsidizing dedicated transmission, substations, and generation capacity for data center facilities. “Bring your own power” requirements which require facilities to produce their own power on-site or paying into a fund to modernize the public electric grid should be non-negotiables. Additionally, large load tariffs and upfront interconnection payments should be implemented to protect the ratepayer from spillover damage.
- Tie permits to clean power and additionality
- A data center should not receive approval merely because it buys renewable energy credits. It should be required to demonstrate additional clean generation, storage, or firm clean capacity that matches its load profile. Otherwise, “green” procurement simply redistributes fossil fuel electricity generation elsewhere.
- Restrict water-intensive cooling in stressed watersheds**
- States should require water-impact assessments and prohibit evaporative cooling in water-stressed regions. Dry cooling, closed-loop systems, recycled wastewater, and non-potable water should be preferred. Where water use is unavoidable, facilities should pay water-impact fees dedicated to local water infrastructure and drought resilience.
- Phase down diesel backup generators
- Permits should require best-available emissions controls, strict generator testing windows, public reporting of runtime, and transition plans towards cleaner backup systems such as batteries, fuel cells, renewable microgrids, or low-emission generators. At minimum, new facilities should be required to use the cleanest feasible generator standards and avoid siting backup-generator yards near homes, schools, or health facilities.
- Establish noise, setback, and design standards
- Local governments should adopt data-center-specific zoning rules: minimum setbacks from residences, acoustic enclosures, limits on low-frequency noise, shielded lighting, tree buffers, and independent post-construction sound testing. Noise regulation should not be left to generic industrial zoning standards that were not designed for 24/7 cooling and generator infrastructure. Although data centers have been around in some form since the 1950s, today’s modern variants are considerably different and produce much more noise.
- Reform subsidies
- Tax incentives should be capped, disclosed, time-limited, and conditional on exemplary behavior. Conditions should include clean energy, water efficiency, local tax contribution, community benefits, and should include enforceable clawbacks for incentives. Given the limited permanent job creation discussed earlier in this note, subsidies should not be justified primarily as employment policy.
- Create community benefit requirements
- Host communities should receive direct benefits: utility-bill stabilization funds, local infrastructure payments, emergency-service funding, broadband investments, workforce training, and pollution monitoring. Residents living near facilities should have standing in permitting and renewal processes.
-
IEA (2026), Key Questions on Energy and AI, IEA, Paris https://www.iea.org/reports/key-questions-on-energy-and-ai.
-
US Energy Information Administration, Use of Electricity, 2026. https://www.eia.gov/energyexplained/electricity/use-of-electricity.php
-
Berkeley Lab (2024), 2024 United States Data Center Energy Usage Report. https://eta-publications.lbl.gov/sites/default/files/2024-12/lbnl-2024-united-states-data-center-energy-usage-report_1.pdf
-
It should be noted in fairness that the compute efficiency of AI is increasing in leaps and bounds, and these projections, while best guesses, may be projecting against a misleading trend.
-
NCSL, “Subsidizing Servers: How States Are Competing to Attract Data Centers”, 2026. https://www.ncsl.org/fiscal/subsidizing-servers-how-states-are-competing-to-attract-data-centers
-
Ibid.
-
World Resources Institute, “From Energy Use to Air Quality, the Many Ways Data Centers Affect US Communities”, 2026. https://www.wri.org/insights/us-data-center-growth-impacts
-
Department of Ecology, State of Washington, “Diesel pollution from data centers” https://ecology.wa.gov/air-climate/air-quality/data-centers
-
EESI, “Communities Are Raising Noise Pollution Concerns About Data Centers”, 2026. https://www.eesi.org/articles/view/communities-are-raising-noise-pollution-concernsabout-data-centers