Large computing facilities built for artificial intelligence have turned into a local grievance in many places, mostly over electricity bills, water and land. Candidates in both major US parties are now shaping messages around that discontent.
Key takeaways
- Data centres, the warehouse-scale buildings that house computing hardware for cloud services and artificial intelligence, have become a subject of organised local opposition in parts of the United States.
- The most common complaints concern electricity demand and its effect on household bills, water used for cooling, noise, land use and the size of tax incentives offered to attract the facilities.
- Because these disputes play out in counties and suburbs rather than only in national debate, they cut across the usual party lines and are attractive to candidates in competitive races.
- Politicians of different parties can oppose the same project for different stated reasons, ranging from consumer costs to environmental impact to scepticism about subsidies for large corporations.
- The scale of any single facility’s power or water use, and the precise effect on a given household’s bill, is often disputed or not publicly disclosed, which makes verification difficult for voters and journalists alike.
What is actually happening?
Across a number of US states, proposals to build or expand large data centres have run into organised resistance at the local level. Planning meetings, county board votes and zoning hearings that once attracted little attention have drawn crowds. Some projects have been delayed, scaled back or refused; others have proceeded over objections.
At the same time, the subject has migrated upwards into state and federal campaigning. Candidates have begun to reference electricity prices, grid strain and the terms of tax deals in their platforms and advertising. Because the 2026 midterm elections are approaching, campaign messaging is being tested and refined, and data centres appear to be one of the themes under examination by strategists in both parties.
It is important to separate two things that are easily conflated. One is the documented existence of local opposition and of political interest in it. The other is the size of the effect these facilities have on any particular utility bill or water table, which is contested and varies enormously by location, by the design of the facility and by how a given state regulates its utilities.
Why is this surfacing now?
The immediate driver is the construction boom associated with artificial intelligence. Training and running large models requires dense clusters of specialised processors, which consume far more power per rack than the general-purpose servers that dominated earlier generations of data centre. That has increased the electrical footprint of new builds and, in some regions, has coincided with rising demand forecasts from grid operators after years of relatively flat growth.
Household energy costs are a durable political subject in their own right. When bills rise and a large, visible industrial facility has recently been approved nearby, voters draw a connection whether or not the causal chain is straightforward. Utility rate structures are complicated; the perception of who pays for new transmission lines and generation capacity is often simpler than the accounting reality.
The electoral calendar supplies the rest of the explanation. Midterm campaigns reward issues that are concrete, local and not already claimed by the other side. A dispute about a building at the edge of town fits that description better than most national arguments.
Background a newcomer needs
A data centre is a building, or campus of buildings, filled with racks of computers, storage and networking equipment. It requires a reliable electricity supply, a cooling system to remove the heat that the equipment generates, and high-capacity fibre connections. Cooling may use air, water, or a combination; water-based systems can consume significant volumes, though designs vary and some recirculate.
These facilities have existed for decades and underpin ordinary services such as email, streaming and online banking. What has changed is scale and density. Sites are being planned with power requirements comparable to those of heavy industry, which brings them into contact with the parts of the regulatory system that govern generation, transmission and long-term utility planning.
States and counties have long competed for such investment using tax abatements and other incentives, on the argument that construction creates jobs and the completed facility adds to the local tax base. Critics respond that operational employment at a finished data centre is modest relative to the capital involved, and that the incentives can be generous. Both observations can be true at once; the balance depends on the specific agreement, and those agreements are not always fully public.
Who is affected, and how?
Residents living near proposed sites are the most direct constituency. Their concerns tend to be practical: construction traffic, the low continuous noise of cooling equipment, the visual effect of large windowless structures and substations, and the loss of farmland or open ground.
Utility customers across a wider area form a second group. If new generation or transmission is built to serve large loads, the question of how those costs are allocated between the new customer and existing ratepayers becomes central. Regulatory commissions in various states have been examining special tariff classes for very large users, precisely to address this.
Local governments occupy an awkward middle position, weighing tax revenue against constituent objections. Workers and construction trades often support projects for the building phase. The technology companies and specialist developers behind the facilities have an interest in predictable permitting. And, more diffusely, anyone who uses cloud or AI services depends on this infrastructure existing somewhere.
Where do informed people disagree?
There is genuine disagreement about the magnitude of the problem. Some analysts hold that large loads, properly priced and paired with new generation, can be absorbed without harm to ratepayers and may even spread fixed network costs across a larger base. Others argue that timelines do not line up: capacity is contracted quickly while new generation and transmission take years, leaving a gap filled by existing, sometimes more expensive, supply.
Water is similarly contested. Consumption depends heavily on cooling design and climate, and figures reported for one site cannot be transferred to another. In water-stressed regions the objection carries more weight than in others.
There is also disagreement about the appropriate level of decision-making. Some favour local control, arguing that communities should be able to refuse projects. Others contend that a patchwork of local decisions produces arbitrary outcomes and that state-level standards on power procurement, cost allocation and disclosure would serve residents better.
Finally, the politics themselves are disputed. Whether opposition to data centres actually moves votes in a general election, as opposed to animating a committed minority at hearings, is an open empirical question that campaigns are effectively testing in real time.
What are the practical implications?
For communities, the practical consequence is that permitting processes are receiving more scrutiny. Requests for binding commitments on noise limits, water sourcing, and payments in lieu of taxes have become more common features of negotiations.
For utility regulation, the implication is a growing body of proceedings about how to treat very large customers. Questions include whether such customers should bear the cost of dedicated infrastructure, whether they must contract for their own generation, and what happens if a facility is announced but never built.
For the industry, the implication is that siting is no longer primarily a technical exercise. Availability of land, power and fibre still matters, but so does the likelihood of local approval, which has pushed some developers towards earlier community engagement or towards jurisdictions with clearer rules.
For politics, the implication is that a technical infrastructure question has acquired a populist register. That can improve accountability by forcing disclosure of terms that were previously negotiated quietly. It can also produce campaign claims that outrun the available evidence, particularly about the precise contribution of any single facility to a bill.
What to watch next
Watch state utility commissions, which are where the cost-allocation question is being decided in practical terms rather than rhetorically. Watch whether legislatures pass disclosure requirements covering power draw, water use and incentive terms, since better data would settle several arguments that are currently unresolvable.
Watch the outcomes of contested local votes: a pattern of refusals would change developer behaviour more quickly than any statute. Watch whether campaign messaging on this subject persists into the closing weeks of the midterm campaigns or is quietly dropped, which would indicate what internal polling shows. And watch technical developments in cooling and chip efficiency, which could alter the resource footprint of new builds and, over time, the terms of the dispute itself.
Frequently asked questions
Why do data centres use so much electricity?
They house large numbers of computers running continuously, plus the cooling systems needed to remove the heat those computers produce. Facilities built for artificial intelligence use specialised processors that draw considerably more power per unit of space than earlier server hardware, so a modern AI-oriented site can require far more electricity than a conventional data centre of similar physical size.
Do data centres raise household electricity bills?
This is disputed and depends on local circumstances. The concern is that if new generation or transmission is built to serve a large customer, some of that cost may be recovered from all ratepayers. Whether this happens, and by how much, depends on how a state’s regulators allocate costs and design tariffs. Reliable, facility-specific figures are frequently not publicly available.
Why do data centres need water?
Many use water as part of their cooling systems, because evaporating water is an efficient way to remove heat. Consumption varies widely: some designs recirculate water, some use air cooling with little water at all, and some use large volumes. The significance of the use depends heavily on local water availability, which is why the objection is stronger in dry regions.
Do data centres create many jobs?
Construction phases typically employ substantial numbers of workers for a limited period. Ongoing operation of a completed facility generally requires a comparatively small permanent staff, since the buildings are highly automated. This gap between capital investment and long-term employment is one of the recurring points of contention when tax incentives are being debated at county or state level.
Why is this a bipartisan issue?
Opposition draws on several distinct arguments that map onto different political traditions: consumer protection against rising bills, environmental concern about power and water, scepticism about subsidies for large corporations, and defence of local land use and property values. Because a project can be opposed from more than one direction, candidates of different parties can each find a version of the argument that suits their existing platform.
Can local communities actually stop a data centre?
Sometimes. Land use and zoning decisions are usually made at county or municipal level in the United States, which gives local bodies real authority over whether a project proceeds. However, rules vary by state, some jurisdictions have pre-approved industrial zones, and developers can revise proposals or look elsewhere. Outcomes therefore differ considerably from one place to another.
Sources and further reading
- The US Energy Information Administration, for published data on electricity generation, consumption by sector and regional demand trends.
- Regional grid operators and their published long-term load forecasts, which describe expected demand growth and the assumptions behind it.
- State public utility commissions, whose dockets contain filings and rulings on tariffs for very large electricity customers.
- Local government planning and zoning records, which set out the terms of specific proposals and the conditions attached to approvals.
Surfaced from the reddit:technology signal “local opposition to data centres”. AI-assisted draft, editorially reviewed.
