At its peak, the Androscoggin paper mill in Jay, Maine, a rural town about 67 miles northwest of Portland, employed about 1,500 people — until a pulp digester exploded in 2020, forcing the mill to close permanently.

In 2023, the 1.4 million-square-foot facility was purchased through a joint venture by JGT2 Redevelopment and a number of other holding and capital companies. The project is led by developer Tony McDonald. Over the next three years, McDonald and his team broke down the mill’s machinery and shipped it to Pakistan, and worked to clean up the industrial site for resale. That resale agreement was finalized earlier this year, according to McDonald — turning Jay into the latest flashpoint over giant data centers in America.

Maine is particularly appealing for data center developers for its relatively cool year-round temperatures, lax land-use statutes, and 54 percent renewable energy mix, the eighth highest in the nation. There is a handful of planned data centers around the state, which recently prompted the state legislature to pass a bill ordering an 18-month moratorium on permits and building of any proposed data center that consumes more than 20 megawatts of power. Lawmakers wanted to pause construction in order to study data centers’ impact on local economies, the power grid, and the environment.

But that bill, which would have been the country’s first, was vetoed by Maine Gov. Janet Mills last month. In her veto, she cited one overriding reason: jobs. A $550 million facility proposed for the shuttered paper mill in Jay, she argued, would create 125 to 150 permanent, high-paying positions in a town that had watched its largest employer close.

From mill towns in Maine to farm counties in Indiana to desert plots outside Abilene, Texas, data center developers are telling local governments: Bring us in, give us what we need, add some tax breaks, and the jobs will follow. More than 35 states have responded by offering incentives and more to attract the industry.

There’s little research into whether massive industrial sites actually deliver the long-term economic gains they promise, but early reports suggest otherwise. Experts say that rural communities often lack the governmental expertise to properly assess how data centers might impact an area. According to recent Pew Research Center data, 67 percent of planned data centers in the US are headed to rural areas, and 39 percent are going to counties that currently have none. As data center development scales rapidly, it’s becoming clear that what rural communities around the country are actually getting isn’t jobs, but a power- and water-hungry industrial facility that temporarily employs about as many people as a midsize restaurant.

The data center fight in Maine

Originally, Tony McDonald had planned to sell the mill to an oriented strand board company called Godfrey Forest Products, which would have employed approximately 150 people, he said. When federal tariffs killed the financial backing for that project, McDonald pivoted to an idea he’d been getting pitches about.

“Most of the people that were contacting us, you know, they were all hat and no cattle,” McDonald said. He fielded multiple calls from what he terms “data center cowboys” who claimed to have one of the seven big tech companies as a client and were looking for a place to build a new data center. When he’d dig deeper on the caller, he’d find that they didn’t actually have the backing they claimed.

After a few conversations, he began pursuing a data center partnership with Sentinel Data Centers, a New York-based company that specializes in data centers serving the healthcare, financial, and hyperscale industries, according to its website. Sentinel did not respond to multiple requests for comment for this story.

McDonald says that, as he understands it, the project in Jay will be a neocloud data center, a specialized facility built to deliver high-performance GPU computing for AI and machine learning workloads. If a neocloud data center is going into the old mill, it will require more than 100kW of energy per rack, according to industry standards, and will need either direct-to-chip or immersion cooling, both of which require ample space and water resources.

While McDonald has repeatedly said that he is not interested in tax breaks for the project, he did want to leave the door open for potential tax benefits down the road if the town determines that it’s worth it. Just before Mills vetoed the moratorium, the Maine state legislature passed a law that excludes data centers from some of Maine’s tax breaks for businesses, but it leaves the door open for local municipalities to offer tax break agreements and other municipal incentives. That could spell trouble for small communities like Jay, hoping for a tax windfall to help them invest in schools, community buildings, and roadworks.

Maine state Rep. Melanie Sachs, the sponsor of the Maine moratorium bill, claims that McDonald did not inform the Jay Select Board of the new plan to turn the mill into a data center until late February 2026, just days before her moratorium bill was scheduled for a floor vote. Her bill was first introduced to committee on January 30th.

The Select Board heard McDonald’s presentation in March and voted 4-0 in support, according to The Maine Monitor. The moratorium bill passed both the state House of Representatives and Senate on April 14th, and Mills vetoed it on April 24th, citing the job creation in Jay as her reason.

Sachs, who chairs Maine’s House Energy, Utilities and Technology Committee, said her legislation was never about banning data centers. “This bill was about creating the playbook,” she said. “And we were told, ‘Don’t worry your pretty little heads about it, data centers are not coming to Maine anytime soon.’ They came anyway, and without a framework in place, towns have no mechanism to evaluate the claims developers are making.”

“Even if it’s 30 jobs, that means a lot to Jay, then, okay, but you’ve swept away protections for 1.4 million Mainers for 30 jobs,” Sachs said.

The economics of data centers

Michael Hicks, the director of the Center for Business and Economic Research at Ball State University, is an economist and a professor who last November published one of the first causal analyses of data center employment effects in the United States. He studied data center openings across 254 Texas counties and measured their actual effect on local long-term employment.

He found that net job creation was effectively zero. Whatever long-term jobs existed at data centers were being offset by losses elsewhere in the same sector.

”As you drive by a data center, you see people working on it,” Hicks said. “You see construction workers. The hotels locally are packed. But there’s no net pulse of that. A lot of these workers are there for three weeks to do their part of it, and then they’re gone … The real question is whether there are permanent jobs associated with data centers, and in Texas, the answer is no.”

Texas serves as an ideal test bed because of its isolated grid and a mix of large, fast-growing metros and tech hubs alongside rural, remote towns, which mirrors the rest of the country. As Hicks puts it, it’s essentially a mini-United States with its variety of regions, and the economic lessons learned in Texas can be widely applied across the country.

Rural towns are often “outgunned” when trying to negotiate deals with large data center builders, said Anthony Elmo, public education funding defender at Good Jobs First, a nonprofit research center focused on corporate and government accountability in economic development.

“They don’t have the resources to negotiate,” said Elmo. “They don’t know what to ask for. They don’t have the legal expertise, and they don’t feel like they have the leverage, which I think is part of the issue.”

On the national level, the math isn’t much better, either. Business Insider recently reported that the national subsidy for data centers exceeded $2 million in costs per permanent job, and in some cases, like one in New York, companies received nearly $77 million in tax breaks for a facility that created exactly one permanent position.

Microsoft’s Quincy, Washington, facility, which is roughly comparable to the initial plan in Jay, employed as many as 500 workers during construction but now operates with just 50 full-time employees. The type of data center being built determines where on that spectrum of long-term employment a community lands. Neocloud data centers, like the one coming to Jay, can require 30 to 50 full-time staff, depending on size.

“Of those, say, 50 jobs, a chunk of them are maintenance, a chunk of them are technicians in charge of backup generators. The high-tech jobs make up maybe 10 percent of the facility,” Elmo points out. And many times, data center companies will count remote workers in other states as employees of the state in which the data center is located. “We may get a little bit of an economic effect from that, but it isn’t nearly as much as if it were a physical person in Maine buying goods and adding to the local economy,” he said.

Most of the lobbying around data centers focuses on job creation and “upskilling,” or training workers for new or better-paying jobs. But according to researchers, even the retraining argument holds little water.

Just under 30 percent of Jay’s population has a bachelor’s degree or higher, while 90 percent have a high school diploma, according to recent census data. And, as Ball State University’s Hicks says, education matters for upskilling.

In the 1800s, when farm laborers were sent into the workforce in the Midwest, they had a basic education that made them trainable for factory work. Following World War II, men and women returned from the factories and the battlefield having learned many new skills, and in the post-Civil Rights South, more Black men and women became educated and entered the workforce in ways they had been unable to before.

As Hicks points out, each of these three industrial revolutions only worked because of the big, new supply of educated people ready to move into those jobs. By contrast, the US currently has negative net immigration, low birth rates, and consistently underfunded education; there is no equivalent human-capital “wave” to support a similar jobs boom in data centers and AI, he says.

“The waves of industrialization accompanied waves of human capital into the United States,” Hicks said. “So, where do we think there’s this surge of employment surrounding data centers that can mimic those three events?”

”The big tech companies are investing in upskilling the construction trades,” Elmo points out, citing OpenAI’s recent agreement with NABTU and noting that in fully developed data center markets, electricians and HVAC maintenance workers float from project to project as contractors. “For states that don’t have developed data center infrastructure, like Maine, it’s not a permanent job. It’s an 18-month job. That’s it.”

The jobs promise, Hicks argues, distracts from the one benefit a data center can reliably provide to a rural community: tax revenue.

Using the Jay data center as an example, Hicks says that a $550 million data center in a town of 4,620 people, where the median home value is around $215,000, would carry an assessed value exceeding the combined worth of every home and every business in the town. The former mill had a tax abatement but generated roughly $1.8 million in tax revenue for Jay in its last year of operation, according to the Livermore Falls Advertiser. Taxed at the same rate as any other commercial property, that revenue could fund schools, rebuild infrastructure, and attract residents for generations.

”You could make that town into a Hallmark Channel town with those sorts of tax dollars, and then jobs would follow,” Hicks said. But that all depends on whether or not the town decides to grant the future project special tax breaks.

It’s not clear precisely how much tax revenue the data center could generate, as of this reporting, because there are still many unknowns about the project, including who Sentinel’s clients might be, the type of data center that will be built, what kind of tax incentives the town Select Board might offer, and even how many jobs the data center might bring to Jay.

McDonald’s job estimates have been all over the map, ranging from 100 to 150 as the project has progressed. He says the numbers he gave the Select Board and the state legislature are based on what Sentinel has told him, but warned that he’s “not a data center guy.”

What data centers actually mean for jobs

Ultimately, this is a tech ouroboros. The same data center infrastructure, subsidized as a jobs program, is purpose-built to reduce human labor, and the AI it powers is explicitly designed to automate work. Communities are being asked to trade tax revenue and grid capacity for jobs in an industry whose core product is labor replacement.

“It’s the biggest capital expenditure since the Manhattan Project, and it isn’t going to create tens of thousands of jobs in the long term,” Elmo said. “It’s not some economic boom. Meta, Amazon, OpenAI, Oracle, they’re shedding jobs in real time while spending billions on data centers so that other organizations can shed additional jobs through AI. At some point, people need to ask more critical questions about this.”

As Hicks points out, rural America is being sold the same bill of goods it has been sold for 50 years. “Civic leaders are living 75 to 150 years in the past,” and framing these data center deals in the old industrial-boom mindset without the educational and demographic conditions that made previous booms possible.

The mill didn’t save rural towns, nor did the manufacturing center, the call center, or the Amazon warehouse. Based on the economic data, data centers will not save them either, and the one thing that could genuinely help — treating the facility’s tax base as a community windfall rather than a negotiating chip — is precisely what most states are legislating away.