Most new data centers in the U.S. are coming to rural areas

The U.S. has more than 3,000 operational data centers, with over 1,500 more in development, implying a substantial expansion ahead. Planned facilities are concentrated in the South and Midwest, and 67% of planned data centers are in rural areas versus 87% of existing ones in urban areas. The article is primarily a structural overview of data-center geography and household proximity, with limited immediate market impact.

Analysis

The investable implication is not the headline growth in computing capacity; it is the relocation of load to cheaper land but more constrained infrastructure. Rural siting raises the probability that the binding constraint shifts from real estate to power interconnects, transmission upgrades, and water rights, which tends to favor firms with utility-scale development expertise over pure-play colocators. The second-order winner is the electrical supply chain: transformers, switchgear, conductors, cooling systems, and grid software should see a multi-year order book extension as projects move from urban infill to greenfield builds.

The market is likely underestimating the duration mismatch between announced projects and commissioned capacity. Even when financing is available, grid connection queues and local permitting can stretch monetization by 12-36 months, so the revenue ramp for beneficiaries should be modeled with a wide timing band. That argues for preferring names with backlog already tied to hyperscaler capex rather than those dependent on a speculative future buildout.

A key contrarian point is that broad public concern may not translate into political delay at first, because local opposition is diluted when projects are dispersed across counties and rural jurisdictions. The real policy risk is later: once load growth forces rate hikes or reliability issues, states may impose stricter water, zoning, or tariff conditions that compress returns. If AI capex slows, the planned pipeline can quickly become an inventory overhang for land, power equipment, and speculative shell development.

The clearest loser is the low-quality greenfield developer without secured power or anchor tenant commitments; these assets are vulnerable to repricing if financing costs remain high or if utility interconnection timelines slip. In contrast, operators with existing campus density near major grids and fiber routes should gain pricing power as customers prioritize time-to-power over headline land cost. That supports a long-quality / short-speculation framework within the data-center ecosystem rather than a blunt thematic long.

Key Decisions for Investors

• Long VRT on a 6-18 month view; thesis is rising grid bottlenecks and mission-critical infrastructure spend, with downside limited by backlog visibility. Best entry on any AI capex pullback or post-earnings multiple compression.
• Long ETN or HUBB versus short an equal-dollar basket of speculative data-center developers / REIT adjacent names; capture the second-order beneficiary trade where equipment demand monetizes sooner than real estate development.
• Long AMT or CCI selectively only if paired with a monitor on fiber-led site selection; their edge is network adjacency, but the risk/reward is weaker than power/electrical names because tower/fiber demand is more indirect.
• Avoid or short high-leverage rural land/power-option plays that trade on announced capacity rather than contracted load; use a 12-24 month horizon because project cancellations and interconnect delays typically show up with a lag.
• For a cleaner expression, buy call spreads on grid/electrical-capex beneficiaries into any broad tech selloff; the setup offers convexity if backlog converts over the next 2-3 quarters while capping premium burn if AI spending pauses.