At every public hearing about a data center, the same moment eventually arrives. A resident stands up and asks: “How do we know you’ll actually do what you’re saying?”
The developer, usually sincerely, answers: “We’re committed to this community.”
It is the least informative sentence in the entire process — not because the developer is lying, but because commitment isn’t a legal category. What protects a community isn’t sincerity. It’s structure: a written agreement that converts every promise in the PowerPoint into an obligation with consequences.
That’s the community benefit agreement — or its cousins, the development agreement, PILOT agreement, and incentive package. Done well, it’s the single most important document in a data center project. Done badly, it’s a press release with a signature page.
Here’s how to tell the difference, clause by clause.
The spine: taxes and the PILOT
Every data center deal runs on taxes. These projects carry enormous assessed value — hundreds of millions in equipment — and enormous leverage, because they can go to the county next door. So they negotiate.
The common tools: tax abatements (we’ll reduce your property taxes), fee waivers (we’ll skip the permitting fees), and PILOTs — payments in lieu of taxes, where the company pays a negotiated amount instead of the standard bill. These aren’t inherently bad. A negotiated deal that lands a project can be worth far more than sticker-price taxes on a project that never arrives. The question is whether the community priced the trade with open eyes.
The red flags: a proposed agreement published too late for anyone to read it before the vote — this happens constantly, and it’s the single most reliable tell of a bad deal. Long, flat abatements with no performance conditions. And net calculations that quietly ignore costs — new roads, fire apparatus, water capacity — that the county will carry for decades.
What good looks like: Fairfax County, Virginia. Facing a proposed 600,000-square-foot data center near a school and neighborhood, the Board of Supervisors approved it with a PILOT requiring the developer to pay $35 million over ten years — roughly $3.5 million a year — starting within thirty days of the special exception, whether or not construction has begun. The payments scale down if the project’s construction or power draw gets delayed. That’s a community benefit you can cash: not tied to projections, not contingent on the developer’s success, running like an annuity from day one.
The organs: water, noise, jobs
Three promises die in boilerplate more than any others.
Water. Every developer says “minimal water use.” Sometimes that’s true — a closed-loop air-cooled design. Sometimes “minimal” means a million gallons a day on a strained aquifer. The good version of a water clause names the numbers: a hard cap on withdrawals, the source identified, cooling technology specified, the right to monitor, and a required response if usage exceeds the cap. “We will use water-efficient cooling” is a vibe. “We will withdraw no more than X gallons per day from source Y, monitored quarterly, with curtailment obligations if exceeded” is a clause.
The fastest route to a number like that usually isn’t argument — it’s joint fact-finding: both sides agreeing in advance on who does the study, what question it answers, and that they’ll live with the result. It costs a few weeks and it buys a figure nobody has to relitigate at the microphone for the next six months.
Noise. The surprise villain of data center opposition. The hum of cooling systems has driven entire neighborhood revolts — people who supported a project on paper couldn’t live next to it at 2 a.m. Good clauses specify decibel limits at the property line, measured by a third party with real baseline measurements taken before construction, plus specific mitigation commitments and a complaint process with deadlines. And here’s a negotiation note most people miss: a developer who commits to an industry-acknowledged best practice, like keeping noise at or under 55 decibels at the property line, can use that concrete commitment as a credibility asset. Specificity helps the honest developer as much as it protects the community.
Jobs. The most inflated promise in the industry. Construction jobs are real but temporary. Permanent jobs are real but modest — data centers are efficient buildings, not factories. The red flag is the number that conflates the two: “800 jobs” that means 750 construction workers for eighteen months and 50 technicians forever. There’s nothing wrong with 50 good permanent jobs. There is something wrong with marketing them as 800.
The strongest structure I’ve seen for making jobs commitments enforceable: Kentucky’s Job Development Assessment Fee, used for projects like the Meta data center in Gallatin County. The company signs an agreement committing to specific job numbers and wages. If it hits them, it captures a share of employee state payroll withholding. If it falls short, employees keep their tax break — and the state claws back the difference directly from the company. The incentive only exists if the jobs do. That’s the design principle in one sentence.
The teeth: enforcement
Here is what “worth the paper it’s printed on” actually means, in five questions:
- Is there a performance bond or security? If the developer fails to deliver — on construction, on a payment, on a mitigation obligation — does money already exist to make the community whole? Or would the town have to sue and hope?
- Are there cure periods and default triggers? What exactly counts as a breach, how long does the developer have to fix it, and what happens if they don’t?
- Can the community terminate — and what survives termination? The best agreements specify which obligations (payments already owed, restoration requirements) outlive the deal itself.
- Who monitors, and who pays for it? An agreement nobody audits is a napkin. Good deals fund independent monitoring — often at the developer’s expense.
- What happens when the project sells? This is the one that blindsides communities. Data centers get sold — to other operators, to REITs, to infrastructure funds. If the agreement doesn’t bind successors and assigns, your deal died in the transaction and nobody told you.
Also worth knowing: as a general matter, a tax abatement — a reduced bill — is easier to defend in court if challenged than a PILOT, which involves the community collecting money it then handles. Your counsel will know your state’s law; the broader point is that structure isn’t pedantry. Structure is what survives.
The paper itself
An agreement with teeth has five properties, every time:
It’s written, and public, before the vote — not summarized in a staff presentation.
It’s specific: numbers, dates, decibels, gallons, dollars. Vague language is where promises go to die.
It’s conditional: the benefits to the developer — permits, abatements, approvals — vest as the obligations to the community are performed.
It’s secured: bonds, guarantees, escrow, or collateral behind the big promises.
And it survives: successors bound, payments that persist, obligations that outlast the ribbon-cutting.
If your agreement has all five, you have a contract. If it’s missing two or more, you have a press release.
For the community side of the table
If you’re a town staring at a draft agreement: you do not have to sign what you’re handed. The first draft is an opening position, and developers expect negotiation — the ones who don’t are telling you something. Bring your own expert. (Data center development agreements are not a general-practice matter; the towns that get good deals hire counsel who’s done one.) Slow the vote down. A deal that can’t survive thirty days of public reading probably can’t survive thirty years of operation.
And ask who is running the room. If the firm facilitating your public meetings is also on the developer’s payroll, or bills more when the project clears, you are negotiating against your own moderator. What I run at CommonWatts is deliberately boring by comparison: one process serving both sides, fees fixed and disclosed to everyone before the first session, nothing contingent on the vote. Apply the same test to anyone who offers to help you that you’d apply to the agreement itself — ask for their terms in writing.
For the developer side of the table
If you draft agreements: the teeth are not against your interests — they’re your best argument. “Here is the agreement, here is the bond, here is the clause that binds whoever buys this site from us” ends more public hearings than any slide deck ever built. A community that trusts your agreement trusts your project. And an agreement you’d be embarrassed to publish is a confession, discoverable later at considerable expense.
I’ve spent years in rooms where the “how do we know?” question was answered with commitment, and rooms where it was answered with structure. The difference between those rooms is the difference between projects that get built with their communities and projects that die at them.
Sincerity is nice. Structure is better. Put it in writing — with teeth — or don’t put up the slide.