AI infrastructure is moving into rural towns that need investment, but the jobs case is much thinner than the construction cranes suggest.
Rural America is being offered a new version of an old promise: let the next big industry build here, give it room, power and maybe some tax relief, and the jobs will follow. The problem is that data centers do not behave like mills, factories or warehouses. They are enormous capital projects with relatively small permanent workforces.
That is why Jay, Maine, matters beyond Maine. The former Androscoggin paper mill, once a pillar of the town’s economy, is now the proposed home of a $550 million neocloud data center led by developer Tony McDonald and JGT2 Redevelopment, with Sentinel Data Centers as a partner. The Verge reported on May 13 that the 1.4 million-square-foot site has become a flashpoint in the national debate over whether AI infrastructure can really revive rural communities.
The appeal is easy to understand. Jay lost hundreds of jobs after a 2020 explosion and the later closure of the mill following Pixelle Specialty Solutions’ departure in 2023. Gov. Janet Mills said in her April 24 veto message that the mill had once accounted for about 22 percent of the town’s tax base. When a site like that sits underused, a data center sounds less like a tech project and more like a lifeline.
But lifelines come with terms. Maine lawmakers had passed what would have been the country’s first statewide moratorium on large data centers, pausing permits for projects of 20 megawatts or more until November 1, 2027. Mills vetoed it because it did not carve out the Jay project. Her office said the redevelopment was expected to create more than 800 construction jobs and at least 100 permanent jobs, while The Verge cited a broader 125 to 150 permanent-job estimate used in the local debate.
Construction work is real work. It fills hotels, keeps trades busy and brings money into a town for a period of time. But it is not the same as replacing a paper mill that employed generations of families. Once a data center is built, the operating model changes. These facilities are designed to run with a small number of highly specialized employees and a lot of expensive machinery.
That distinction matters because public incentives are usually justified with permanent local benefit. Ball State economist Michael Hicks studied data center openings across Texas counties and found effectively no net long-term job creation. The headline number can look impressive during construction, but the durable employment effect may disappear once contractors leave and the facility settles into routine operations.
Neocloud data centers make the question sharper. These facilities are built to serve GPU-heavy AI and machine-learning workloads. They need dense power, advanced cooling and industrial-scale reliability. They can also be more valuable to the companies using them than to the towns hosting them, unless the local deal is structured around the one benefit that data centers can reliably offer: taxable property value.
This is where Jay’s old mill site becomes a real test. A $550 million facility in a small town could transform the tax base if assessed and taxed like other commercial property. It could support schools, roads, public buildings and basic services. But if the project is paired with aggressive local abatements, the town could absorb grid, water and land-use tradeoffs while giving away the very upside that made the deal attractive.
The rural shift is already happening
Jay is not an isolated case. Pew Research Center data published in April found that 67 percent of planned U.S. data centers are in rural areas, while 39 percent are planned for counties that do not currently have one. Existing data centers are still mostly urban, but the next wave is moving toward cheaper land, available industrial sites and communities hungry for investment.
That shift changes the negotiating balance. Big technology and infrastructure companies often arrive with lawyers, consultants and long-term power strategies. Small towns may have part-time boards, limited planning staff and little experience evaluating megawatt-scale projects. It is not that rural officials cannot make good deals. It is that the other side has done this many times before.
The infrastructure burden also does not stay abstract. Data centers consume large amounts of electricity, and some cooling systems put pressure on water resources. Mills herself acknowledged in her veto message that a moratorium was appropriate given concerns about the environment and electricity rates, even as she argued Jay deserved an exception. That is the tension in one sentence: the state sees the risk, while the town sees the empty mill.
For AI companies, this should be a warning. The next phase of AI will not be judged only by model performance, chip supply or cloud margins. It will be judged by whether the physical infrastructure behind it creates deals that local communities still defend ten years later. A rural town does not need a data center to be a factory. It needs the company behind it to be honest about jobs, transparent about utilities and realistic about incentives.
The practical takeaway is simple. If rural America is going to host the AI economy, towns should bargain for tax base first, workforce promises second and public subsidies last. The data center boom may bring investment to places that badly need it, but investment alone is not renewal. The communities that benefit will be the ones that understand exactly what they are selling before the servers arrive.
Also read: Tencent’s revenue miss puts its AI payoff under pressure • Metis TechBio’s Hong Kong debut shows AI biotech is back in demand • Anduril’s $5 billion raise turns defense AI into venture’s new prize
