Data Centers and Local Economies in the Age of AI: A Shift–Share Approach

Data centers are the physical infrastructure behind cloud computing, artificial intelligence, and enterprise software. The rapid diffusion of artificial intelligence is increasing demand for computing capacity, accelerating investment in data centers, and raising concerns about their local economic and environmental effects. We assemble a facility-level panel of U.S. data centers and link it to county-level measures of employment, establishments, payroll, income, house prices, electricity prices, and public-supply water withdrawals. To address endogenous site selection, we develop a shift-share instrument motivated by two complementary requirements of fiber backbone connectivity: a feasible fiber route and a point of access into that fiber. Because installing long-haul fiber across many separate parcels is costly and time-consuming, providers have often followed continuous transportation corridors, such as existing railroad lines, where routes are already assembled and access or rights-of-way are less burdensome. At the same time, a data center must be able to enter the backbone through a splice or access point. We construct two historical exposure measures. The first identifies locations where terrain made long-distance routes relatively inexpensive to build, while the second identifies access points into the fiber backbone. We interact these historical shares with growth in scale-oriented facilities outside the U.S. The resulting instrument does not systematically predict county outcomes before the modern expansion of data centers. After that expansion begins, the estimates show positive effects on total employment, construction employment, establishments, payroll, tax returns, adjusted gross income, and wages. We complement the shift-share design with dynamic long-difference specifications that use the same exogenous shares. The estimates show persistent gains in employment and establishments, construction effects that are stronger at shorter horizons, sustained increases in house prices and public-supply withdrawals, and positive effects on electricity prices that emerge later. Overall, this pattern is consistent with the hypothesis that data centers increase economic activity, both directly and indirectly, which puts pressure on house and electricity prices and increases water demand.