HomeCountiesArticlesMethodologyContactHow Municipal Leaders Can Turn a Data Center from a Fight Into a Community Asset — by Writing the Terms Themselves
The default municipal response to a proposed data center has become the moratorium: pause, study, delay. It is an understandable reflex — residents worry about water, power, noise, and traffic, and the developer’s promises often outrun the community’s benefit. But a moratorium is a posture, not a plan. It signals that a community can only say no, never shape a yes. This briefing argues for the harder and more valuable move: architecting the deal. A municipality that writes specific, enforceable conditions into its zoning code and development agreements — closed-loop cooling to protect water, on-site clean generation sized above the facility’s own needs, transparent public reporting, and real remediation deadlines with teeth — can convert an anxiety into an asset. The technology to meet these conditions already exists and is being deployed today. The legal instruments to require them are ones cities already use. The one genuine constraint — that a city cannot by itself rewrite how power is sold to the grid — is navigable if leaders understand where their authority ends and the state’s begins. Delay shows fear. Design shows foresight. And because the binding constraint on AI’s progress is no longer chips but power and the places to put it, the community that learns to architect these deals is not tolerating the future — it is helping build it. This is a blueprint for design.
Before discussing how to structure a data center deal, it is worth being clear-eyed about why a community should want one at all — because the case rests on something bigger than tax revenue. StrataHelm has been open about where we stand: we take the risks of artificial intelligence seriously, and we are also genuinely optimistic about what it will do for humanity. That optimism is not vague cheerleading. It rests on capabilities already appearing in the peer-reviewed record.
As we argued in our position statement, Where We Stand, we believe AI will be one of the most beneficial technologies of our era — and as we detailed in Would You Give Up Your Job to Save a Life?, that belief is grounded in concrete results: AI systems catching cancers earlier in large randomized screening trials, flagging sepsis hours before it turns fatal, discovering novel antibiotics against drug-resistant superbugs, and bringing specialist-grade diagnosis into ordinary clinics. Add to that the clean-energy, climate, and scientific breakthroughs on the horizon, and the stakes become clear. Every one of those benefits runs on computation.
And here is the part that connects directly to the fight over a data center in your community: the thing now standing between us and more of those breakthroughs is not ideas, and increasingly it is not even chips. It is physical infrastructure — specifically, power and the places to put it.
For a couple of years, the scarce resource in AI was the chips themselves — whether you could get the graphics processors that train and run these models. That era is ending. In 2026 the binding constraint has moved downstream to something far more physical: electricity, and the ability to connect a facility to the grid. Analysts now describe the industry as power-bound rather than GPU-bound, with one grid observer putting it bluntly that AI has become a power-and-water business wearing a semiconductor costume (Spheron, 2026; Scienceshot, 2026). At NVIDIA’s 2026 developer conference, CEO Jensen Huang reframed the economics of the whole sector around a single formula — revenue equals tokens per watt times available gigawatts — making power infrastructure, not chips, the primary lever on how much AI the world can actually produce (Schneider Electric, 2026).
The numbers behind the constraint are striking. A single AI query can use on the order of a thousand times the electricity of a traditional web search, and modern AI facilities demand anywhere from 100 to 750 megawatts each (EnkiAI, 2026; TechPlusTrends, 2026). The International Energy Agency projects that data center electricity consumption could roughly double globally by 2030, with AI the majority of the new demand, and Gartner projects that 40 percent of AI data centers will be power-constrained — physically unable to draw the electricity their hardware could use — by 2027 (IEA, 2025; Gartner, via Scienceshot, 2026). Grid interconnection alone now takes 24 to 36 months in major U.S. markets, and in some regions 4 to 10 years — against a data center that can be designed and built in two to three (Spheron, 2026; World Economic Forum, 2026). The queue, as one analyst put it, is the queue: you cannot clear a grid backlog by spending more money at the same spot.
The breakthroughs we are betting on run on compute, compute runs on power, and power is now the wall. A community that can help solve the power problem is not doing AI a favor — it is claiming a seat at the center of the decade’s defining buildout.
This reframes the entire municipal conversation. A data center is not, at bottom, a warehouse full of computers a town reluctantly tolerates. It is the physical substrate of the technology we have staked our optimism on — and the single hardest input to secure is exactly the thing a forward-thinking local government is positioned to help provide responsibly: power, water stewardship, and a place to build. That is enormous leverage. The bottleneck is real, it is physical, and it is local. Which means the community holds more cards than a moratorium mindset assumes — and the right move is to play them, not fold them.
Across the country, communities facing data center proposals are reaching for pause buttons. The impulse is defensible — the concerns are real, and a bad deal is worse than no deal. But a moratorium has three quiet costs. It forfeits leverage, because the community never states what a good project would look like and so never gets one. It cedes the narrative, casting local government as reactive and obstructionist rather than strategic. And it leaves money on the table — the tax base, the infrastructure, and the clean-energy contributions that a well-structured project could deliver simply go to a different town that did the work. The alternative is not "approve everything." It is to decide, in advance and in writing, the terms on which the community will say yes — and then hold firm to them. A developer who walks away from fair, transparent, enforceable terms was never bringing a good deal. A developer who accepts them has just been turned into a municipal asset.
A moratorium is the community saying "we don’t know how to handle this." An ordinance is the community saying "here is exactly how."
Water is the concern that sinks the most data center proposals, and for good reason: a large evaporatively-cooled facility can consume enormous volumes, historically tens to hundreds of millions of gallons a year for a hyperscale site, drawn in many cases straight from the municipal drinking-water supply (CAES Field Report, 2026; Tom’s Hardware, 2026). But this is a solved engineering problem, and a city can simply require the solution. Closed-loop and direct-to-chip liquid cooling systems are filled once and then recirculate the same water in a sealed system, with no ongoing evaporation and near-zero day-to-day water consumption — a Water Usage Effectiveness approaching zero, against an industry average near 1.8 liters per kilowatt-hour (Introl, 2026; Vantage Data Centers, 2026). Microsoft now reports next-generation designs that consume, over an entire year, roughly the water of a single restaurant; Oracle reports effectively zero ongoing community water use for its closed-loop AI facilities (Microsoft, 2024; Oracle, 2026).
The honest caveat — and municipal leaders should state it plainly rather than have a developer’s lawyer raise it — is that there is a real water-for-energy tradeoff. Eliminating evaporative cooling can raise electricity consumption, because evaporation is a thermodynamically cheap way to shed heat; during peak summer conditions, evaporative systems can use 10–35 percent less electricity than fully dry alternatives (arXiv, 2026). This is precisely why the water pillar and the energy pillar below must be written together: a city that mandates closed-loop cooling should expect a somewhat higher power draw, and should meet it by requiring the facility to bring its own clean generation — which turns the tradeoff into an advantage rather than a hidden cost.
Here is where a city moves from harm-reduction to genuine community benefit. The most forward-looking condition a municipality can attach is a requirement that a data center bring its own on-site clean generation — solar, wind, storage, and where appropriate other low-carbon sources — sized meaningfully above the facility’s own projected consumption, so that the site is a net contributor of clean energy rather than a net drain on the grid. This is not utopian. Utilities and market operators are already moving in exactly this direction: some markets now condition new data center approvals on the operator demonstrating an independent power-generation plan, a "bring your own power" requirement driven by the simple fact that the grid cannot absorb the new load fast enough otherwise (datacenterHawk, 2026). Operators are increasingly required to share power back to the grid as a condition of interconnection (Data Center Knowledge, 2026). The top five hyperscalers have already contracted over 40 gigawatts of U.S. renewable energy (Milken Institute, 2026). A city requiring over-provisioned on-site generation is pushing on an open door.
The oversizing target — a figure such as generation capacity set well above estimated peak need, with the surplus made available to the community — is the mechanism that flips the script. Instead of a data center that strains local power, the community hosts one that helps power it. But this is also where a city must understand the limits of its own authority, and this is the single most important legal point in this briefing.
It is tempting to write an ordinance simply mandating that a data center pump its surplus electricity onto the local grid "at no charge to the city." A municipal attorney will stop you, and it is better to understand why now than in litigation. Selling or transferring electricity to the grid is a regulated transaction. Under long-settled law, the Federal Energy Regulatory Commission (FERC) regulates wholesale sales and transmission of electricity in interstate commerce, while state public utility commissions regulate retail sales, local distribution, and utility service obligations (Pillsbury, 2026). A city zoning authority sits beneath both. When power flows onto the grid, it enters a wholesale market a municipality does not control, governed by interconnection rules, cost-allocation fights, and market-based-rate requirements that are the active subject of FERC proceedings right now (Orrick, 2025; Utility Dive, 2026). A flat local "give the grid free power" mandate would likely be preempted.
This is a reason to be precise, not a reason to abandon the goal. A city cannot set the wholesale price of exported power, but it retains substantial tools to achieve the underlying aim of a net-positive-energy facility:
The principle survives the legal constraint. The community can still insist on a facility that generates more clean power than it consumes. It simply has to route the "how" through the correct legal channels rather than a single well-meaning sentence in a zoning code.
Every commitment above is only as good as the community’s ability to verify it. The most durable protection a city can write is not a single technical requirement but a standing obligation to report — publicly, on a fixed schedule, in plain terms. An ordinance can require quarterly public disclosure of water withdrawal and consumption, energy consumed versus energy generated on-site, the share of consumption met by clean sources, and other agreed metrics such as noise levels and local hiring. This directly answers one of the best-documented failures in current practice: data center deals are frequently struck under non-disclosure agreements, without economic-impact assessments, with residents learning the terms only after approval (Food & Water Watch, 2026; America’s Plan, 2026). Transparency is the cheapest enforcement mechanism a city has. It costs the operator little, it builds the public trust that makes the whole project politically durable, and it converts vague promises into a public record the community can hold up against reality.
Reporting without consequences is theater. The final pillar is an enforceable remediation ladder, written into the agreement, that triggers automatically when a reported metric falls below an agreed threshold. A workable structure: if a site misses a threshold, it has 30 days to submit a remediation plan and a further 60 days to implement it, with defined escalating consequences — financial penalties, then suspension of permits or incentives — for failure to cure. This mirrors the clawback provisions that community groups have successfully introduced elsewhere when job-creation or mitigation targets went unmet (America’s Plan, 2026). The specific windows are a policy choice, and counsel should set them against local administrative-procedure timelines. But the structure — objective threshold, short deadline to plan, defined deadline to fix, real penalty for failure — is what separates an enforceable agreement from a press release. It also protects the good-faith operator, who gets a clear, predictable process rather than the arbitrary risk of a future political backlash.
None of this discipline is worth much if the underlying project does not benefit the community — so does it? The evidence says it can, under the right terms, and disappoints under the wrong ones. On the encouraging side, Loudoun County, Virginia — the "data center capital of the world" — now derives roughly 38 percent of its general fund revenue and nearly half its property tax revenue from data centers, easing the tax burden on residents and funding schools and services (JLL, 2025). Metro Phoenix data center development generated $863 million in state and local tax revenue in a single year (JLL, 2025). Construction phases employ hundreds of workers, and industry analyses cite large ancillary-employment multipliers (CBRE, via WYEDC).
But the sophisticated leader must weigh the other half of the record, because an AI-optimist scrutinizing this briefing certainly will. Permanent on-site jobs at a data center are few — often dozens, not hundreds — and aggressive tax abatements have in some cases produced eye-watering costs per job: one Ohio facility received a $54.3 million property tax abatement associated with roughly 20 jobs (Good Jobs First, 2025). Independent research finds that data centers boost local growth and jobs mainly where the surrounding economy is already urban and developed, with rural host communities feeling far less benefit (Fortune / The Conversation, 2026). The lesson is not "data centers are good" or "data centers are bad." It is that the terms determine the outcome. A project that pays full freight (or a fair negotiated PILOT), reports transparently, brings its own clean power, and protects the water can be a substantial net positive. A project handed a 100-percent abatement under an NDA in exchange for vague promises is how communities end up subsidizing their own strain. The difference is entirely in the architecture — which is the city’s to write.
The communities that will look wisest in ten years are not the ones that moved fastest to block data centers, nor the ones that rolled out the reddest carpet. They are the ones that treated a data center proposal as an engineering-and-governance problem to be solved on the community’s terms. The cooling technology to protect the water exists. The interconnection trend toward operator-supplied clean power is already running in the city’s favor. The legal instruments — zoning conditions, development agreements, transparency mandates, remediation ladders — are ones municipalities use every day. What is scarce is not capability but nerve and imagination: the willingness to sit down and write the terms of a good deal instead of reaching for a pause button. Delay is a community admitting it can only fear the future. Architecture is a community deciding to shape it. For a technology that is not going away, the second posture is the only one that ages well.
The same discipline that lets a city architect a good data center deal — knowing its own exposure, its fiscal base, and its leverage before the developer walks in — is what StrataHelm builds for local leaders. See where your community stands →
America’s Plan (2026). "Tax Abatements and Data Centers: What Communities Give Up." americasplan.org.
arXiv (2026). "Small Bottle, Big Pipe: Quantifying and Addressing the Impact of Data Centers on Public Water Systems"; and "AI Data Centers and the Water Use Feedback Loop." arxiv.org.
CAES Field Report (2026). "Understanding How Data Centers Impact Surface and Ground Waters." University of Georgia. fieldreport.caes.uga.edu.
datacenterHawk (2026). "Behind-the-Meter Power Solutions: The Data Center Industry’s New Reality." datacenterhawk.com.
EnkiAI (2026). "AI Data Center Grid Strain: Power Halts Growth in 2026." enkiai.com.
International Energy Agency (2025). "Energy and AI." iea.org.
Schneider Electric (2026). "AI Power Infrastructure: The Hidden Bottleneck Behind AI" (Jensen Huang, GTC 2026). blog.se.com.
Scienceshot (2026). "AI Data Center Power Bottleneck: Grid Beats GPUs" (Gartner power-constraint projection). scienceshot.com.
Spheron (2026). "Power-Bound, Not GPU-Bound: AI Data Center Power Constraints Are the Real 2026 Bottleneck." spheron.network.
TechPlusTrends (2026). "AI Data Center Power Requirements 2026: The Grid-to-Chip Guide." techplustrends.com.
World Economic Forum (2026). "Is Power Grid Connectivity the Strategic Bottleneck for AI?" weforum.org.
Data Center Knowledge (2026). "Why Data Centers Are Turning to Behind-the-Meter Power." datacenterknowledge.com.
Food & Water Watch (2026). "What Do We Lose When Our Leaders Give Handouts to Data Centers?" foodandwaterwatch.org.
Fortune / The Conversation (2026). "Data Centers Boost Jobs 4% in Cities and Rural Economies Barely Feel a Dent"; "Why Better-Off Cities and Towns See More Benefits from Data Centers Than Rural Regions." fortune.com; theconversation.com.
Good Jobs First (2025). "Property Tax Breaks for Data Centers — Let Us Count the Ways"; and tax-abatement cost-per-job analyses. goodjobsfirst.org.
Introl (2026). "Water Usage Efficiency: AI Data Center Cooling Without Crisis." introl.com.
JLL (2025). "How Data Centers Transform and Engage with Local Communities" (Loudoun County and Metro Phoenix figures). jll.com.
Microsoft Cloud Blog (2024). "Sustainable by Design: Next-Generation Datacenters Consume Zero Water for Cooling." microsoft.com.
Milken Institute (2026). "The Deal Behind-the-Meter: Vertical Integration Reaches the Grid." milkeninstitute.org.
Oracle (2026). "Closed-Loop Cooling in Oracle AI Data Centers." oracle.com.
Orrick (2025). "Powering Data Centers," Megawatts to Megabytes guide (FERC interconnection and behind-the-meter classification). orrick.com.
Pillsbury (2026). "The Emerging Regulatory Framework to Power Data Centers" (FERC vs. state jurisdiction). pillsburylaw.com.
Tom’s Hardware (2026). "Microsoft CEO Says New AI Data Centers Use as Little Water Annually as a Restaurant." tomshardware.com.
Utility Dive (2026). "PJM Proposes Behind-the-Meter Reforms in Data Center Colocation Effort." utilitydive.com.
Vantage Data Centers (2026). "Cooling Without the Drain: How Closed-Loop Systems Cut Day-to-Day Water Use." blog.vantage-dc.com.