Blog · Analysis · Last reviewed June 23, 2026

The Machine Needs a Town

AI is not weightless. It needs land, water, substations, transmission lines, permits, backup power, tax deals, and a community willing to absorb the footprint. The cloud has become a zoning question.

For this essay, the town is the full host system around a compute campus: local government, utility territory, watershed, neighborhood, emergency services, tax base, ratepayers, and the people asked to live beside infrastructure built for distant model users.

The Cloud Gets a Postal Address

For years, the public language around computation made infrastructure disappear. Software was "in the cloud." Models lived behind chat boxes. Intelligence arrived as a subscription, an app, or a workplace feature. That language is no longer adequate. Generative AI has made the physical substrate visible again.

A data center is not an abstraction to the place that hosts it. It is a campus, a building envelope, a cooling system, a set of substations, a transmission problem, a backup-power plan, a tax arrangement, a water user, a construction project, and a long-term political relationship. When AI companies ask for more computation, they are also asking towns, counties, utilities, and ratepayers to absorb more physical load.

This is the first serious correction to the mythology of artificial intelligence. The machine does not float above society. It lands somewhere. That is why this essay should be read beside AI Data Centers, AI Energy and Grid Load, The Data Center Becomes a Civic Machine, and The Interconnection Queue Becomes AI Governance.

Current Context

As of June 23, 2026, the data-center debate is no longer only a technology story. It is a utility, land-use, water, ratepayer, and local-democracy story. The International Energy Agency estimated that data centers used about 415 terawatt-hours of electricity in 2024, roughly 1.5% of global electricity consumption, and projected about 945 terawatt-hours by 2030 in its base case. In the United States, the Department of Energy announced a Lawrence Berkeley National Laboratory report estimating 176 terawatt-hours of data-center electricity use in 2023, about 4.4% of U.S. electricity, with a projected range of 325 to 580 terawatt-hours by 2028.

The regulatory context has also hardened. On June 18, 2026, the Federal Energy Regulatory Commission issued show-cause orders to the six regional grid operators under its jurisdiction, directing them to justify or reform tariff rules for data centers and other large energy users. FERC's stated issues include study processes, transparency into transmission costs, co-location and behind-the-meter generation, flexible large-load service, and preventing cost shifting. NERC's May 2026 guideline for emerging large loads, including data centers, recommends stronger modeling, data collection, operational coordination, event recording, and reliability practices. In plain terms: a large compute campus is now a grid-planning object, not only a zoning file.

Local governments are moving too. Charlotte's City Council voted on June 8, 2026, to approve a 150-day moratorium on new data centers, running until November 5, 2026, while an interdisciplinary city and county group studies policy responses. Charlotte is only one case, but it shows the new civic pattern: when existing zoning categories cannot answer energy, water, noise, visual-impact, and cumulative-buildout questions, a pause can become part of governance rather than an anti-technology gesture.

The current question is therefore sharper than "will AI need more data centers?" It is whether the public record around those data centers can identify the load, water source, cooling design, cost allocation, public incentives, emergency-service burden, noise mitigation, ratepayer exposure, and local benefits before a private campus becomes a public obligation.

What Communities Are Fighting

Recent reporting and public records show a consistent pattern: communities are not simply rejecting technology. They are asking who receives the benefit and who carries the cost. Local residents see land conversion, traffic, noise, water demand, grid upgrades, diesel backup generators, visual impact, and public subsidies. The benefits are often described in broader economic terms: regional investment, construction jobs, tax revenue, and cloud capacity whose value may be captured far away from the host community.

That mismatch creates political friction. A town can be told it is participating in the AI future while receiving a warehouse-scale neighbor, higher infrastructure pressure, and a small number of long-term jobs compared with more labor-intensive industries. Even where a project is legal and economically attractive, the public question remains: did the community meaningfully choose this future, or was it processed through a permitting system designed for a slower technological era?

Charlotte's 2026 moratorium is a useful signal because the official city FAQ does not present the pause as permanent rejection. It says the city is not accepting applications for new data centers during the moratorium, while already approved or complete applications may continue; the city also says staff will research policy options and engage stakeholders. Moratoria are blunt tools, but they appear when ordinary planning categories cannot keep up with a new development pattern. The politics is not only "yes" or "no" to data centers. It is whether local government has enough information, legal leverage, and time to negotiate the terms of the footprint.

The stronger question is cumulative. One facility can be governed as a parcel. A cluster becomes a regional infrastructure decision. If several campuses appear inside the same utility territory or watershed, ordinary project-by-project review may miss the combined effect on substations, transmission, water planning, construction traffic, emergency services, and political bargaining power.

Water, Power, Noise, Heat

The impact of any particular data center depends on design: cooling method, climate, power source, water source, backup system, utilization, and local grid conditions. A dry-cooled facility, a water-intensive evaporative system, and a campus paired with new renewable generation are not identical. But the common categories of impact are now clear enough to govern.

Power is the largest public issue. AI training and inference increase demand for high-density computing, and high-density computing increases demand for reliable electricity. EIA's 2026 outlook says data-center server energy use is a major factor in renewed electricity-demand growth, while its Virginia analysis shows how concentrated load can reshape one region's commercial electricity sales and summer peak-demand forecast. The International Energy Agency similarly treats data centers as a major new source of electricity demand, especially in countries already hosting large computing clusters.

Water is the second public issue. Some facilities use water directly for cooling; others use electricity from power plants that consume water elsewhere. Lawrence Berkeley National Laboratory's data-center efficiency materials distinguish direct water consumption for cooling from indirect water consumption through electricity generation, and ISO/IEC 30134-9 defines water usage effectiveness as a data-center key performance indicator. Those metrics help, but they do not replace local judgment. Communities in water-stressed regions have a legitimate interest in source, withdrawal, consumption, discharge, drought performance, and opportunity cost.

Noise and heat are easier to dismiss until they are local. Cooling fans, backup generators, transformer hum, truck traffic, and construction cycles can become quality-of-life issues. Waste heat may be manageable, reusable, or technically ordinary, but it is still heat. A community asked to host machine cognition is also asked to host machine exhaust.

Rural Siting and Democratic Bypass

Developers have incentives to seek cheap land, fast approvals, large parcels, grid access, and limited organized opposition. Recent reporting has described a shift toward rural locations partly because they can reduce the public friction found in cities and suburbs. That does not make rural siting inherently wrong. Rural communities need tax bases, jobs, and infrastructure too. But the pattern deserves scrutiny.

Rural governments may have smaller planning staffs, fewer technical consultants, less legal budget, and less experience negotiating with hyperscale technology firms. A county may be asked to evaluate grid impact, water impact, noise standards, tax abatements, emergency planning, and long-term land-use consequences while facing a company with vastly greater technical and legal capacity.

That asymmetry matters. Democratic consent is not just a public hearing held at the required time. Consent requires understandable information, enough time to study it, independent expertise, and real ability to say no or demand different terms. It also requires protection against nondisclosure-driven governance: a town cannot deliberate about a public footprint if the important facts are hidden in confidential utility, tax, or development negotiations.

Ratepayers and Hidden Subsidy

The hardest politics may happen through utility bills. If a hyperscale customer pays directly for new generation, transmission, and interconnection, the public burden is different from a scenario where grid upgrades enter a utility rate base and are spread across ordinary customers. The distinction is technical, but morally central.

Consumer advocates have warned that AI data-center buildouts can expose households to higher electricity costs if utilities overbuild for large customers or socialize infrastructure expenses. FERC's June 2026 large-load action uses different institutional language, but it points at the same problem: tariffs need to address transmission-cost transparency, cost-shifting prevention, and large-load flexibility. Utilities also face a planning dilemma: reject large loads and risk losing economic development, or build for them and risk leaving the public with stranded costs if demand projections change.

This is why electricity governance should be part of AI governance. A model can be sold as private innovation while its infrastructure costs are negotiated through public utility commissions, tax abatements, zoning boards, bond markets, and regional transmission planning. The public may never use the model directly, yet still pay for the system that makes it possible.

The New Company Town

The old company town organized housing, wages, stores, policing, and local politics around a dominant employer. The AI data-center town is different, but the risk has a family resemblance: a community can become dependent on a powerful outside firm whose primary commitments are elsewhere.

Data centers do not usually employ people at the scale of factories after construction ends. Their value is capital-intensive rather than labor-intensive. That changes the bargain. A locality may receive tax revenue and short-term construction work, but the ongoing social relationship can be thin: a guarded campus, a few specialized jobs, a large utility footprint, and a corporate presence with little everyday civic life.

When a town becomes a host for computation, it should ask what kind of reciprocity it is receiving. Is the company funding schools, emergency services, grid resilience, water conservation, housing, noise mitigation, public reporting, and local technical capacity? Or is it merely buying enough land and political permission to turn local capacity into remote intelligence? This is where siting policy meets public compute governance: public accommodation for AI infrastructure should create public rights, not only private throughput.

A serious data-center approval process should begin with disclosure. Communities need plain-language reporting on projected electricity demand, peak load, water use, backup generation, noise, expected jobs, tax incentives, emergency-service needs, construction timelines, and decommissioning plans. Claims about sustainability should be specific enough to audit.

Second, local governments need independent technical review paid for by applicants but controlled by the public authority. A small county should not have to accept a developer's analysis as the only expert account of a project.

Third, public utility commissions should protect ordinary ratepayers. Large-load customers should bear the costs they create unless there is a transparent public reason for sharing them. If a utility builds for speculative AI demand, the risk should not quietly migrate to households.

Fourth, communities should negotiate community benefit agreements where the footprint is large. That can include water conservation funding, local resilience investments, apprenticeship programs, noise controls, emergency equipment, public dashboards, and enforceable performance standards.

Fifth, approvals should carry operating conditions. A data center's promise at the hearing should become a monitorable obligation: load ramp schedule, water source, noise limits, backup-generator restrictions, emergency contacts, curtailment commitments, public reporting cadence, and remedies if promised jobs, tax revenue, or infrastructure benefits fail.

Sixth, temporary moratoria can be legitimate when planning systems are outpaced. A pause is not necessarily anti-technology. Sometimes it is the only way a community can recover the ability to think.

What This Changes

Intelligence always has a body.

AI is often marketed as a cognitive event: language, reasoning, agents, automation, creativity. But cognition has supply chains. Synthetic intelligence occupies land, consumes power, moves water, produces heat, needs metals, depends on labor, and reshapes local politics. The machine that reflects society back to itself is built out of very ordinary things: substations, cooling loops, zoning maps, rate cases, and land deals.

This changes the politics of AI. The question is not only whether models are safe, biased, useful, or economically disruptive. The question is also where the machine is allowed to incarnate. Which towns become its body? Which watersheds cool its thought? Which households underwrite its electricity? Which rural boards negotiate with companies larger than their public budgets? Which communities become sacrifice zones for someone else's automation layer?

The machine needs a town. The town deserves more than a promise that the future will be efficient.

Source Discipline

Data-center claims need source separation. IEA, DOE, LBNL, EIA, FERC, NERC, ISO, and Charlotte city materials establish official estimates, regulatory actions, standards, guidance, or local government status. They do not prove that any specific project is safe, fair, clean, popular, or economically beneficial. Those judgments require project-level permits, utility filings, tax agreements, water records, noise studies, rate cases, and enforceable contracts.

Secondary reporting is useful for community experience and political pattern recognition, but it should not substitute for primary records where those records exist. A news story can show that residents objected or that a developer targeted a rural site. It cannot by itself settle who pays for a substation, whether water rights are adequate, whether clean-power claims are deliverable, or whether a tax abatement produces net public benefit.

The article treats "data center" and "AI data center" carefully. Not every data center is a frontier AI training cluster. Annual electricity use, site peak load, server electricity, total facility electricity, water withdrawal, water consumption, cooling type, backup generation, and ratepayer exposure are different claims. Current-source claims were checked against the named sources on June 23, 2026.

Sources

Return to Blog