Data Center Ballot Measures: Why Harwood, North Dakota Voted the Other Way

Data center ballot measures delivered a clear message on Tuesday: in 44 of 45 local propositions tracked nationally, voters landed on the anti-data-center side. The exception was Harwood, North Dakota, a town of about 940 people roughly 10 miles from Fargo. There, voters rejected a charter amendment that would have required public hearings and independent assessments for large projects. The headline reads as a rare pro-data-center result. The details are more complicated, and they show how local governments handle fast-moving infrastructure deals.

What Happened in Harwood

According to the North Dakota Secretary of State’s office, 325 voters opposed the measure and 105 supported it. That is about a 76 to 24 split, not a narrow loss. The tally comes from Redbud Consulting, a ballot tracking firm, which counted 45 data-center-related propositions nationwide, with Harwood the only one on the pro side. I could not verify that Redbud Consulting has a Wikipedia article, so that anchor should be treated as plain text in the published version.

The context matters. A $3 billion AI facility, described as equal in size to 17 football fields, is already under construction in town. Its developer, Applied Digital, hopes to employ 200 people, a figure close to a fifth of the town’s population. The project is due for completion next year, and the rejected measure would not have touched it.

Reading the Result Correctly

A Vote on Process, Not on the Facility

The most important analytical point is that Harwood voters did not hold a referendum on whether the data center should exist. Mayor Blake Hankey said so directly: the proposition was not a moratorium and was not aimed solely at data centers. It was an amendment to the town’s home rule charter that would have applied to future projects above a certain scale. A gas station would not have qualified. A large data center would have triggered independent assessments and mandatory public hearings.

Treating the 325-to-105 tally as an endorsement of AI infrastructure would be a mistake. A voter could reasonably oppose this particular amendment while still disliking data centers. That ambiguity is the main weakness of the “lone pro-data-center town” framing.

Why Voters Rejected It

Reporting points to two explanations, and they are not mutually exclusive:

• Drafting concerns. The mayor called the language imprecise and said some stipulations “would have been financially burdensome to a city our size.”
• Scope confusion. Supporters Mike Swartz and Mark Coolidge said some residents disagreed over whether the rules would also hit smaller buildings and businesses, not just large ones.

A third factor is plausible but unproven: the prospect of local property tax and sales tax revenue may have made voters wary of anything that could slow future development. The reporting does not confirm this, so it should be treated as interpretation.

The Speed Problem: 21 Days From Announcement to Construction

The case for the measure rested on timing. Swartz said residents were “blindsided,” noting that only 21 days passed between the announcement and the start of construction. Many in town, he said, did not know what a data center was.

That compressed schedule is the strongest argument for the amendment, and the measure’s defeat does not refute it. Fast approvals are common when developers are choosing between competing sites. In Harwood’s case, the competing site was Fargo, a short drive away.

The Role of the Non-Disclosure Agreement

Hankey said he signed a non-disclosure agreement with Applied Digital and does not regret it. His reasoning: without the NDA, the project would “almost certainly” be in Fargo, with no sales tax or property tax benefit for Harwood.

This is a genuine trade-off and deserves a critical look. NDAs let local officials negotiate against rival jurisdictions, but they also remove the public from the earliest and most consequential stage of a decision. The mayor’s argument is coherent from a competitive standpoint, and it is also exactly the dynamic that residents like Swartz say they want to constrain. Neither side is simply wrong. They are optimizing for different things: capturing the investment versus legitimizing it.

The National Backdrop: Resistance Is Growing

Harwood sits inside a much larger pattern. Major developers and technology companies, including SpaceX, Microsoft, Google, Amazon and OpenAI, are seeking new sites for the data center capacity that supports artificial intelligence products. They have met stiff resistance at town halls and city councils.

The reported concerns fall into four groups:

1. Environmental impact, including land use and energy demand.
2. Electricity bills and who bears the cost of grid upgrades.
3. Water supply, particularly for cooling.
4. General distrust of AI, which colors local debates regardless of technical merits.

The 44-to-1 national tally is striking, but it needs a caveat. Ballot propositions that reach voters are often put there by organized opponents, so the sample is self-selected. It measures the intensity of organized opposition more than the average opinion of all communities that host or court data centers. It is still a meaningful signal for developers.

Why Harwood Is an Outlier, Not a Counterexample

Several features set Harwood apart from the places where measures went the other way:

• The facility already existed. The measure was prospective, so the vote carried no power to stop the visible project.
• The measure was broad. Because it covered any large project, it was easier to criticize on cost and clarity.
• The town is small. A city of 940 has limited staff to run independent assessments and hearings, which gave the mayor’s cost objection real weight.
• Economic stakes were concentrated. A single project promising jobs equal to roughly a fifth of the population shapes how residents weigh procedure against benefit.

The vote was also the end of a fractious process. After contentious council meetings, including one where protesters were ejected by sheriffs, Swartz gathered enough signatures to place the question on the ballot. That is the citizen initiative process working as designed. Whatever the outcome, residents used it to force a public debate that did not happen before construction began.

Practical Lessons for Local Governments and Developers

For Municipal Officials

• Separate confidentiality from consultation. If an NDA is unavoidable, schedule public disclosure and hearings immediately after, not after construction starts.
• Define thresholds precisely. Rules triggered by capital cost, power load or acreage are easier to defend than vague references to “large projects.”
• Budget the process. Small towns can partner with counties or use pooled assessments to keep review costs manageable.
• Check existing tools. Zoning rules and development agreements may already allow targeted hearings without a charter amendment.

For Residents and Advocates

• Draft narrowly. Harwood’s measure failed partly because voters disagreed about its reach. Narrow, well-defined language reduces that risk.
• Separate the issues. A transparency rule and a moratorium are different proposals, and combining them invites confusion.
• Engage early. Pressure on councils before votes on agreements has more leverage than pressure after.

For Developers

Speed wins sites, but secrecy builds opposition. Harwood shows that a developer can win the local vote and still leave residents unconvinced. Coolidge’s remark that “it’s all about money” is a trust problem, not a legal one, and it can resurface when the next project comes up. Applied Digital did not respond to a request for comment.

Open Questions

Reporting leaves several things unanswered. We do not know how many of the 325 “no” voters supported the data center, how many objected to drafting, and how many simply prefer fewer rules. We also do not know what the final tax revenue will be, or whether the 200 jobs will be filled by local residents. Until those numbers appear, claims that Harwood represents a pro-data-center community are speculation.

What the Harwood Vote Actually Signals

The most important insight is that Harwood did not reject public participation in principle. It rejected one poorly scoped version of it, after a project had already been approved in 21 days. The more likely lesson for the next wave of data center ballot measures is that proposals which are narrow, cheap to administer and clearly separated from a ban will fare better than broad charter amendments.

Watch three things next. First, whether Harwood’s council adopts a smaller, more precise disclosure rule on its own. Second, whether the 44-to-1 national pattern persists as more measures reach ballots, particularly in towns where a facility is already built. Third, whether states begin to regulate early-stage secrecy around large projects. The unresolved question is whether communities will gain a real voice before the shovels arrive, or only after the decisions that matter have been made.

FAQ

Did Harwood vote in favor of data centers?

Not directly. Voters rejected a charter amendment that would have required public hearings and independent assessments for large projects. It was a vote on a procedural rule, not on whether data centers should be allowed.

How many data center ballot measures were there?

Redbud Consulting counted 45 data-center-related propositions on the ballot nationwide. Forty-four went the anti-data-center way, and Harwood’s was the exception.

Would the Harwood measure have stopped the current data center?

No. The existing facility, which is due for completion next year, would not have been affected. The measure would only have applied to future projects above a defined scale.

Why did the mayor oppose the proposition?

Mayor Blake Hankey said the language was imprecise, that some provisions would be financially burdensome for a city of Harwood’s size, and that it applied to any large business, not just data centers.