Key findings:
- The expansion of AI and data centers is at present almost the sole driver of US economic growth, as measured by GDP.
- Because data centers are highly capital-intensive but not very labor-intensive, they have the potential to worsen income inequality and geographic differences in prosperity.
- Data centers also contribute to an increasing divergence between traditional measures of economic development, such as GDP growth, and effects that are actually felt by locals , including job and income growth.
- Policymakers and economic development practitioners must recognize that some kinds of investment and GDP growth, despite being large, will contribute little to the economic wellbeing of host communities.
The conventional formula for successful economic development goes like this: Investment leads to increased output, which produces growth in jobs and prosperity. In other words, if you can attract more private investment, jobs and other good things will follow. That’s why, when reporting on new industrial and commercial developments, reporters reflexively cite the amount of money to be invested. They, along with policymakers and readers, assume that the size of the capital investment is indicative of the project’s economic impact for residents.
It doesn’t always work out that way and increasingly the returns on this approach to economic development are shrinking. That’s because, at present, over 90% of the nation’s economic growth as measured by gross domestic product (GDP) is attributable to information processing and software, largely driven by capital investments in artificial intelligence (AI). This includes the data centers and the energy resources required to deploy them.
The problem with relying on the AI boom to drive local economic development is that it has a problematic characteristic: it’s highly capital-intensive and not very labor-intensive. In other words, very little of the investment that projects command or of the revenue they earn is used to pay workers, especially local workers. The vast majority of the investment and income that data centers command and generate accrues to far-away investors, creditors, and shareholders.
This cohort of AI investors, creditors, and shareholders is made up of a very narrow sliver of well-off Americans, few of whom live in the communities where the data centers will be built and operate. And, because data centers employ few workers relative to the amount of output they generate, little of the income they earn will be injected into the economies of communities where they are located. On top of that, many states and communities mistakenly see data centers as economic development “opportunities” and offer sweetheart development packages, often waiving sales taxes on equipment and abating property taxes in efforts to attract them. This further reduces data centers’ already meager potential contributions to local economies. Despite their immense size and earning potential for investors, data centers will, in most cases, prove to be of small economic value to their host communities and local residents
Adverse effects from AI and data center development aren’t confined to communities that host data centers. When industrial sectors that are highly capital-intensive and non-labor-intensive dominate economic growth and remaining sectors are stagnant, as is the case at present, disconnects can emerge between generalized and largely abstract measures of prosperity, such as GDP, and more specific and tangible measures, such as jobs and income. The result can be an economy whose topline results appear to be spectacular, but in which, below the surface, large and destructive redistributions of wealth and earning power are taking place.
A 30-county region of northern Appalachia with a population of almost two million people has actually experienced this phenomenon. Between 2008 and the present, this region (referred to in the charts that follow as “Frackalachia”), experienced a huge burst of productivity and GDP growth when the application of hydraulic fracturing technology (“fracking”) turned the thirty counties into one of the most prolific producers of natural gas in the world.
However, because fracking for natural gas is one of the economy’s most capital-intensive and least labor-intensive activities and because much of the specialized labor and material inputs fracking requires were sourced from established providers that were located elsewhere, very little of the investment that was made to produce gas or of the revenue generated by the sale of natural gas ever entered Frackalachia’s economy. Consequently, while the region’s GDP grew much faster than that of the nation, the thirty counties suffered from negative job and population growth and personal income grew at only two-thirds the national rate.
Source: Ohio River Valley Institute
It’s true that the economies in the 30 Frackalachian counties were generally weak before the gas boom. But the point is that the boom, as large as it was, did almost nothing to alleviate the region’s overall anemia. And, in some ways, it may even be contributing to it by crowding out other kinds of investment.
The phenomenon of a highly capital-intensive, non-labor-intensive industry suddenly becoming almost the sole source of economic growth, even on an immense scale, caused the region’s ratio of personal income-to-GDP to drop from 100% to 64% as the region’s Mining sector (which includes natural gas extraction in the BEA data), grew to command more than a third of GDP and nearly all of the income it produced went to capital investors, not local workers.
Source: Ohio River Valley Institute
It doesn’t take much imagination to recognize the potential similarities between the rise of natural gas production in Appalachia and the current datacenter boom. In fact, when adjusted for population, the scale of the Appalachian natural gas boom, as measured by dollars invested and as a share of GDP, is probably greater than even the immense sums that companies are pouring into AI and data centers.
In the case of the AI-driven data center boom as with the Appalachian natural gas boom, we’re dealing with a highly capital-intensive, non-labor-intensive industry that is dominating economic growth at a time when other economic sectors are notably weak. When Harvard economist, Jason Furman, recently pointed out that data center-related activity was responsible for 92% of the US economy’s GDP growth in the first half of 2025, he also observed that the rest of the economy grew at an annual rate of just 0.1%.
A similar trend is likely to appear in future economic data in regions where data centers pop up. This phenomenon is likely to be exacerbated by the fact that other major industrial sectors which participate in the data center value chain – natural gas and utility-scale power generation – are also highly capital-intensive and not very labor–intensive. In short, no element of the data center value chain is likely to offer much in the way of economic benefit to the communities that host them.
Even as data centers proliferate, and the important functions they perform become a greater part of the nation’s economy, their economic development shortcomings may show up in national statistics in a disturbing way. They will contribute to what has become a long-term decline in the share of income that the economy allocates to labor.
Since 2001, labor’s share of the income earned by US businesses has declined from 68% to 58% while the portion allocated to capital has grown from 32% to 42%. This has contributed greatly to income inequality. If AI and data centers continue to dominate economic growth, the problem will worsen. Income allocated to labor will stagnate, as it did in Frackalachia.
Nationally, labor’s decline as a contributor to economic output is already posing serious challenges to the economic well-being of people whose incomes depend on their labor. It’s a problem that the unfortunate residents of Frackalachia know all too well as they are forced to accept either lower standards of living or the uprooting of their families in order to find employment and economic opportunity elsewhere.
It’s not inevitable that AI and data centers will continue to dominate the nation’s economic growth. AI’s economic impacts and the role played by data centers may turn out to be smaller than expected or the rest of the nation’s economy may just be in a lull and will soon bounce back, reducing the share of growth attributable to AI and data centers. And perhaps that growth, if it happens, will be more labor-intensive and beneficial to workers and communities whose economic prosperity is more tied to labor than it is to capital.
But, if current trends continue and economic growth remains largely confined to the capital-intensive, non-labor-intensive sectors that make up the AI and data center economies, workers in more regions of the country and in other industrial sectors will begin to find themselves in situations like the one experienced in Frackalachia. The US economy as a whole could become less labor-intensive, with knock-on effects for how income and wealth are shared.
Already, we are hearing about the “K-shaped economy” in which wealthier households are prospering and working class households struggle to make ends meet. Fed chairman, Jerome Powell recently described it this way:
“If you listen to the earnings calls or the reports of big, public, consumer-facing companies, many, many of them are saying that there’s a bifurcated economy there and that consumers at the lower end are struggling and buying less and shifting to lower cost products, but that at the top, people are spending at the higher income and wealth.”
Indeed, news stories suggest that retailers and service providers are responding to pricing signals by shifting marketing strategies in ways that cater to high-end customers who have a greater willingness and ability to spend. Correspondingly, consumers at the lower end of the income scale may be left with product mixes in which there are fewer choices and reduced quality.
The effects of growing income inequality as well as increases in inequality between regions could be socially and politically corrosive and may ultimately result in governments having to intervene to modify the ways in which constituents are taxed and the ways in which industry and markets allocate income in order to adequately and fairly meet the needs of workers and their families.
The most effective way to avoid such future crises is for policymakers who are currently involved in economic development to recognize that, in a changing economy, capital-intensive and non-labor-intensive enterprises, such as data centers, may be ascendant, but that, if they are lightly taxed and poorly regulated, they are unlikely to have much economic or financial impact for local residents.
The alternative for local policymakers is to concentrate economic development efforts and financial resources on businesses and industries that are labor-intensive, locally owned, and that are deeply interwoven into existing local economies. The Ohio River Valley Institute in its report on the economic rebirth of Centralia, Washington, an old coal town in Washington state, described how investments in these types of activities can actually revive failing economies and have disproportionately beneficial economic effects.
Meanwhile, the federal government must start contemplating the economic and social implications of an economy increasingly made up of industries that employ fewer and fewer workers and therefore provide little in the way of labor-based income.