Harwood, North Dakota: The One Town That Voted to Keep Data Centers Coming

Inside Harwood's Lone Pro-Data-Center Vote

On Tuesday, voters in 45 US communities weighed in on data centers — and in 44 of them, the message was the same: slow down, or at least give us a say. Harwood, North Dakota, a town of roughly 940 people, was the exception. Its residents rejected a ballot measure that would have required independent assessments and mandatory public hearings before large projects, such as AI data centers, could be built. The margin was decisive: 325 against, 105 in favor, according to the North Dakota Secretary of State's office.

The proposition would not have halted the $3 billion AI data center already under construction in Harwood's backyard, a facility the size of 17 football fields that is set to be completed next year and is expected to employ about 200 people — roughly 20% of the town's population. Instead, it would have applied to future projects meeting a scale threshold. A new gas station would not have qualified; a new large data center would have faced independent assessments and public hearings.

The vote exposed a split that has become familiar across the country. Residents like Mike Swartz and Mark Coolidge argued the measure was needed because the town was 'blindsided' by how quickly the current project moved — 21 days from announcement to construction. Mayor Blake Hankey opposed it, saying the language was imprecise, that it would have applied to any business coming to town, and that its requirements 'would have been financially burdensome to a city our size.' He also disclosed that he signed a non-disclosure agreement with Applied Digital, the developer, and said that without moving quickly, the data center 'would almost certainly be in the city of Fargo right now, without any kind of sales tax or property tax for our community.'

Applied Digital did not respond to a request for comment. The result leaves Harwood as a national outlier — a small town that, when given the chance to impose new guardrails on large-scale development, chose not to.

Why Harwood Broke With 44 Other Communities

What Harwood Actually Voted On

The proposition was framed by its opponents as broader than a data-center referendum. Mayor Hankey said the home rule charter amendment 'wasn't strictly aimed at data centers' and would have applied to 'any kind of business coming to town.' That framing appears to have mattered: Swartz and Coolidge attributed the defeat partly to voters disagreeing over whether the measure would also capture smaller buildings and businesses. In a town of 940, where a single large employer can reshape the tax base, the question of scope was not abstract.

The 21-Day Approval and the NDA

The speed of the original approval is central to the dispute. Swartz said it took 21 days from announcement to construction, and that many residents 'didn't even know what that was or meant.' Hankey's decision to sign an NDA with Applied Digital — and his statement that he does not regret it — illustrates the leverage dynamics at play. For context, NDAs between municipalities and developers are common in competitive site-selection processes, but they can limit the public information available before a project is approved, which is precisely the grievance the ballot measure sought to address.

Why Harwood Broke With 44 Other Communities

Redbud Consulting counted 44 of 45 data-center-related propositions nationwide as falling on the anti-data-center side. Harwood's outlier status is best explained by what residents stood to lose, not by warmer feelings toward AI. The town already has the project; the measure would not have affected it. What was at stake was future revenue and the town's reputation as a place where large projects can be approved quickly. Hankey's Fargo comparison makes the competitive logic explicit: if Harwood had hesitated, a neighboring city would have captured the sales and property tax base.

The National Pattern Behind One Small Vote

The article notes that major developers — including SpaceX, Microsoft, Google, Amazon and OpenAI — are seeking new AI data centers and encountering stiff resistance at town halls and city councils, with Americans worried about environmental impact, electric bills, water supply, and in some cases AI itself. Harwood is a data point in that trend, but an inverted one: it shows that when a community has already secured the jobs and tax revenue, the political calculus can flip. The contentious meetings, including one where anti-data-center protesters were ejected by sheriffs, suggest the underlying tension in Harwood has not been resolved by the vote.

What This Means for Applied Digital and Its Peers

For Applied Digital, the outcome removes a near-term obstacle to future expansion in Harwood and preserves a favorable local posture. For other developers, the lesson is more nuanced: community opposition is real and increasingly organized, but it can be defused — or at least outvoted — when a project's economic benefits are already visible and the proposed restrictions are framed as costly or overly broad. The risk is that this model depends on speed and information asymmetry, both of which are drawing scrutiny.

What Harwood's Vote Signals for Developers and Towns

The Harwood result shows that data-center opposition is not uniform — it can be neutralized when a community already has the jobs and tax revenue, and when restrictions are framed as costly or overly broad.

For developers and local officials navigating similar votes:

• Quantify the local fiscal benefit early. Hankey's Fargo comparison — sales tax and property tax that would otherwise go to a neighboring city — was a concrete argument that resonated. Developers should expect to supply comparable figures.
• Expect scope language to be the battleground. The measure failed partly because voters disagreed over whether it applied only to large data centers or to any business. Drafting precise thresholds matters for both sides.
• NDAs carry political risk. Hankey defended signing one, but the 21-day approval timeline and residents' claim of being 'blindsided' became the core grievance. Communities that move fast should anticipate a backlash and plan for public engagement even when confidentiality is required.
• Watch whether Harwood's vote becomes a template or an outlier. With 44 of 45 measures going the other way, the national trend remains restrictive. Developers should track whether other towns with existing data centers follow Harwood's lead in future election cycles.

Metrics worth watching: Applied Digital's construction timeline for the Harwood facility (completion expected next year), whether the company announces further expansion in the area, and the count of data-center ballot measures in the next election cycle.