Just Passing Through?

The Wrong Question

The question that most communities are asking about data centers is some version of, “Are they good or bad for us?” The question matters, but skips a distinction worth making first. Before you can evaluate whether a major capital investment is good or bad for a community, you need to know what kind of investment it is in the first place.

A core distinction is whether an investment is the siting of infrastructure or the seeding of industry. Infrastructure and industry are both valuable and needed. But they are different economic development mechanisms, deliver different kinds of benefits, and should be evaluated against different expectations. When a community expects infrastructure to be industry — expecting it to generate new firms, thicken labor markets, and transform the local economy — it sets itself up for disappointment. And when civic attention is consumed by an infrastructure project that has been misclassified as an industrial one, the work of building real industry can get lost or neglected.

Data Centers Through the Anchor Lens

The investments that actually transform a local economy are defined not by their size but by what they cause to grow nearby. The economic development literature identifies four mechanisms that separate anchors from large capital expenditures. The brief applies each one to data centers:

Labor markets — A typical hyperscale facility employs 50 to 150 permanent workers. The jobs are specialized but narrow; they build a workforce, not a labor market. A region doesn’t become more attractive to manufacturers or technology firms because it has 50 data center technicians. It only becomes more attractive to more data centers.

Supply chains — Servers from Dell or HPE, GPUs from NVIDIA, cooling systems from Vertiv or Schneider Electric. Procurement is corporate, global, and standardized — designed for scale, not for the kind of iterative purchasing that builds local supplier ecosystems.

Knowledge spillovers — In mature data center markets like Northern Virginia, the primary spillover has been the attraction of more data centers, a clustering effect driven by shared infrastructure rather than knowledge exchange. There is little evidence that proximity to a data center makes it more likely that a new technology company will start.

Firm formation — Absent. Even Loudoun County, after a quarter century and 300 facilities, developed clustering within the infrastructure service layer — not a jump from infrastructure to industry.

The Landscape Is Moving

The governance question would be straightforward if data centers were a stable, well-understood asset class with predictable economics. They are not. Communities are negotiating 15- to 20-year agreements against an investment whose technology, economics, and even physical form factor could look fundamentally different within a few years. The brief identifies eight dimensions of active change, including:

Construction is automating. Prefabrication and robotics are projected to cut on-site labor — the one local benefit everyone agrees on. One drilling-robot pilot saved eight weeks per data center across ten hyperscaler projects.

The grid connection is severing. At least 46 facilities, representing roughly 30 percent of all planned U.S. capacity, are building their own dedicated power plants rather than relying on local utilities.

Overbuild risk is real. U.S. utilities are tracking more than twice the committed capacity that third-party demand estimates project by 2030.

The grid is vulnerable from the load side. In Virginia, clusters of data centers simultaneously dropping off the grid forced the operator to scramble to avoid damaging power plants. NERC’s chief engineer called it “one of our most important emerging risks.”

The counterparty may not be who you think it is. Hyperscalers increasingly build through special-purpose vehicles backed by private equity. The check doesn’t stop because the tech company goes bankrupt. It stops because the financing vehicle was never as durable as either side assumed.

Ratepayers may bear undisclosed costs. In Virginia, Dominion Energy projects residential costs could increase by $14 to $37 per month by 2040, driven primarily by data center load — costs borne by households that had no seat at the table.

Model efficiency may reduce demand. If AI models become more compute-efficient, or if demand consolidates to fewer sites, communities built around data center revenue could face fiscal cliffs with obligations constructed on revenue that is no longer there.

Companies are even evaluating space-based and underwater compute infrastructure — not because it’s cheaper, but because terrestrial capacity may be approaching the limits of what can be built fast enough. The fact that companies are willing to consider far more expensive options tells you how little the asset’s value depends on being embedded in any particular place.

Clear Terms, Not Closed Doors

The infrastructure-vs-industry distinction isn’t a reason to say no to data centers. It’s a way to say yes with clear terms and to have a plan for what you do with the proceeds.

A community that uses data center fiscal revenue to fund a workforce development program, hire a full-time economic development director, or invest in a community college technical program is doing something logical. It is using infrastructure revenue to build the institutional capacity for industrial strategy. But that only works if the community actually captures the revenue, invests it deliberately, and doesn’t mistake the fiscal stabilization for the industrial transformation itself.

Civic attention is one of the scarcest resources a community has. Every hour that a city council, a planning commission, or an economic development team spends debating a data center proposal is an hour not spent on the investments that do have the potential to generate industry dynamics. Infrastructure governance should be competent and proportionate. It should not dominate the economic development agenda.

Some capital arrives in a place and embeds, creating workforce depth, supply chain relationships, knowledge that accumulates, and firms that build on each other’s work. Some capital arrives and passes through, generating fiscal events along the way but leaving the local economy structurally unchanged. Knowing which kind you are hosting is the first step toward making better decisions.