UADA researcher examines economic prospects of AI data centers in rural areas

By RYAN McGEENEY |University of Arkansas Division of Agriculture

Across the country, the explosive growth of massive data centers, many of them meant to power the growth of artificial intelligence, has drawn both criticism and support from communities in both rural and urban areas.

As more land is annexed for future data center development, some have questioned the economic costs and benefits of those decisions. Frank Seo, assistant professor of rural development for the University of Arkansas Division of Agriculture, recently examined the possible economic ramifications of AI data centers for Arkansas, at both the state and county level.

In “Data Centers vs. Factories: Do AI Facilities Truly Benefit Rural Economies? Evidence from an Arkansas Simulation,” first published in Southern Ag Today, Seo compared three industry scenarios based on the creation of 100 jobs in an AI data center, auto manufacturing, and food manufacturing. Based on existing data — national average productivity and work hours for each scenario — Seo simulated the total economic impact of each industry in the state of Arkansas and in four rural counties: Calhoun, Dallas, Newton, and Woodruff, each of which has a low population density according to U.S. Census data.

“I became interested in this issue because there are strong opinions on both sides, but relatively little objective evidence at the county level showing how the economic impacts of data centers compare with those of other industries that could potentially locate in rural communities,” Seo said. “I think this is particularly important right now because communities and policymakers are making decisions about data center development, incentives, infrastructure, and land use.”

The economic scope 

The article focuses specifically on the economic impacts of one year of operation and does not address other concerns, including electricity and water consumption or environmental factors such as air and noise pollution.

Seo examined three factors for each business model: the output multiplier — the total economic output generated per dollar of direct output; the employment multiplier — total jobs supported per direct job; and the tax impact — tax revenue generated per dollar of output.

“When we think about the economic impact of an industry, there are three parts to consider: direct, indirect and induced impacts,” Seo said. “The key point is that the economic benefits of a data center do not stop with the people directly employed by the facility. The activity can also create additional benefits through local suppliers and household spending.

“For example, suppose a company builds a data center in County A. The construction itself creates economic activity by hiring construction workers and contractors. That is the direct impact.

“The construction company also needs materials and services such as concrete, steel, fuel, electrical equipment and engineering services,” he said. “If those goods and services are purchased from businesses within the county, those businesses also benefit from the project. They may increase their own employment or purchase additional supplies from other local businesses. That is the indirect impact.

“There is also a household spending effect,” Seo said. “Workers who earn income from the construction project and from businesses supporting the project spend some of that income locally, for example, at grocery stores, restaurants, retail stores, housing and health care providers. That spending generates additional economic activity. That is the induced impact.

“The important question for a rural community is how much of this activity can actually take place locally,” he said. “If a rural county has relatively few suppliers, the construction company may have to purchase many of its materials and services from outside the county. In that case, some of the economic benefits leave the community rather than circulating locally. The same thing can happen with household spending. If workers have limited opportunities to spend their income locally, they may spend it in neighboring communities or through businesses located outside the county, including online retailers.”

The good, the bad and the context

In his findings, Seo said that data centers compare favorably with auto and food manufacturing facilities in terms of overall economic output, although the magnitude and composition of their economic impacts vary by region.

“The differences are largely driven by the economic linkages associated with data centers,” Seo said. “At the statewide level, these linkages are spread across a much larger and more diversified economy. Data centers purchase specialized goods and services from businesses across the state, while employee spending generates additional activity across a broader labor market. In rural areas, smaller labor pools, fewer local suppliers and a less diversified economy mean that a greater share of these indirect and induced effects can occur outside the local area.”

Seo said, however, that data centers generally generate fewer employment spillovers than traditional manufacturing across regions. He explained that counties with smaller regional labor pools, particularly for the specialized technical workers required by data centers, may need to recruit workers from outside the region.

“Data centers often rely on technically specialized workers, hardware, and external services, so some of the spending flows out to other regions,” Seo said. “Conversely, the simulation results show that those two manufacturing industries are more likely to hire locally and purchase inputs from nearby businesses, which helps keep dollars circulating within the community.”

Seo found that data centers’ top advantage over more traditional manufacturing was in tax impact.

“Across all regions, data centers generate more tax impact per dollar of output than the other sectors,” Seo wrote. “So even if not all the economic activity stays local, there can still be meaningful benefits for public finances.”

However, he cautioned that this effect should be considered carefully, because data center developers often negotiate tax abatements or utility waivers before construction.

“Tax incentives must be considered carefully, especially for rural counties that host AI data centers, particularly when those counties have limited supporting industries and limited opportunities for local household spending,” Seo said. “When a county cannot capture as much of the economic activity through local businesses and employment, the tax revenue generated by the data center becomes an important part of evaluating its overall local impact.

“In 38 states, data centers receive some form of dedicated tax incentive, including sales and use tax exemptions, property tax incentives, and other investment incentives,” he said. “The key policy question is not simply whether a data center generates economic activity, but how much of that activity and the associated tax revenue the local community can retain after considering the incentives provided to the facility.”

While there are no proposed data centers in any of the four counties Seo chose for his model, he said they were selected because of their general similarity to many rural counties in Arkansas.

“The counties were selected because they represent highly rural economies and provide examples of how different industries might perform in rural settings,” Seo said. “They were not selected based on the presence of a particular proposed data center.

“My goal with this article was not to argue that data centers are either good or bad. Instead, I wanted to provide an evidence-based way to answer a more specific question: How much economic activity does a data center generate locally, and how does that compare with other industries that could potentially locate in rural areas?”

To learn about extension programs in Arkansas, contact your local Cooperative Extension Service agent or visit uaex.uada.edu. Follow us on Facebook and Instagram. To learn more about the Division of Agriculture, visit uada.edu. To learn more about ag and food research in Arkansas, visit the Arkansas Agricultural Experiment Station at aaes.uada.edu.