Why Data Centers Are Seeking New Locations Now

The hyperscalers face a pressing challenge. They need land, power, and fiber — and they need it fast. But the places that used to check all three boxes are running out of room.

Northern Virginia hosts roughly 70% of the world's internet traffic and has become the canonical example of what happens when data center demand outpaces regional infrastructure. Power queues stretching years into the future, moratoriums on new construction in Loudoun County, and municipalities pushing back on what amounts to industrial-scale electricity consumption in residential corridors illustrate this issue. What was once the gold standard of data center site selection has become a cautionary tale about concentration risk.

So the industry is moving — not retreating, but actively scouting for the next generation of viable locations. The factors driving those decisions reveal a lot about where digital infrastructure is headed.

The Demand Side Has Changed Shape

The surge in AI workloads is doing something the previous cloud boom didn't: it's compressing timelines and inflating power requirements simultaneously. A standard hyperscale data center might consume 20–40 MW, while AI training clusters routinely demand 100 MW or more on a single campus. Microsoft, Google, and Amazon have collectively announced hundreds of billions in data center investment over the next several years.

That's not simply "more of the same." It's a fundamentally different scale requirement — one that blows past the capacity of markets built for the previous generation of compute. When a single tenant needs as much electricity as a mid-sized city, you can't just bolt another building onto an existing campus.

The volume of data being generated, stored, and processed is growing in ways that compound this pressure. Edge computing, autonomous systems, real-time AI inference, and video streaming at higher resolutions all feed back into the demand for compute infrastructure, distributed across more locations, not fewer.

What Actually Drives Site Selection

Power availability is the dominant variable right now — not theoretically, but operationally. Developers are asking utility partners one question above all others: how many megawatts can you deliver, and by when? A site with excellent fiber, cheap land, and favorable taxes is worthless if the local grid can't support the load or if interconnection studies will take four years.

Connectivity matters enormously too, though it's more solvable than power. Submarine cable landing stations, major internet exchange points, and proximity to existing fiber corridors all factor into latency-sensitive site decisions. For latency-tolerant applications — deep storage, AI training, batch processing — connectivity requirements relax somewhat, which opens up geographies that would have been non-starters a decade ago.

Water access is the underappreciated third rail of data center site selection. Cooling accounts for roughly 40% of a data center's energy consumption, and most large facilities still rely on water-intensive cooling systems. Regions experiencing drought conditions or tightening water rights — a growing list that includes parts of the American Southwest — are seeing developers walk away from otherwise attractive sites.

The insider reality is that power, water, and connectivity form a three-legged stool, and most "ideal" locations are missing at least one leg.

The Cost Equation Is More Complicated Than Real Estate

Surface-level real estate economics still matter — land in rural Ohio is cheaper than land in the San Jose suburbs — but sophisticated operators have learned that minimizing acquisition cost can mean maximizing operating cost over a 20-year asset life.

Electricity rates are the long lever. A data center running at 100 MW for two decades at a rate that's $0.01/kWh higher than an alternative location represents roughly $17.5 million in additional annual electricity spend. Over the life of the asset, that gap compounds into a number that makes even significant real estate premiums look trivial.

Labor markets matter more than the industry sometimes admits. Operating a hyperscale facility requires skilled technicians, security personnel, and network engineers. Rural sites offer cheaper real estate and power but thin talent pools. Hybrid approaches — campuses sited for power and land availability within reasonable commuting distance of a mid-sized metro — are increasingly the operational sweet spot.

Operational efficiency over the asset lifecycle, not initial build cost, is what separates disciplined infrastructure developers from those who chase cheap land and then regret it.

Regulations, Incentives, and the Policy Landscape

State and local governments have figured out that data centers generate significant tax revenue while requiring relatively little municipal service. That recognition has driven a wave of incentive packages — property tax abatements, sales tax exemptions on equipment purchases, and, in some cases, direct grants tied to job creation or renewable energy commitments.

Virginia created the problem it now has, in part, because its incentive structure was aggressively attractive for decades. Other states have watched and calibrated. Georgia, Texas, Ohio, and the Carolinas have all positioned themselves as alternatives, with varying success.

The green energy angle is increasingly central to the incentive conversation. Major hyperscalers have made public commitments to 24/7 carbon-free energy, which means they're not just looking for cheap power — they're looking for renewable power that can be matched hour-by-hour, not just annually averaged. States with strong wind or solar resources, combined with favorable interconnection policies, are winning deals that would have gone elsewhere five years ago.

Zoning and permitting timelines are the hidden variable that can make or break a site. A location with every other advantage becomes noncompetitive if permits take 36 months. Markets that have streamlined permitting for large industrial users — treating data centers the way they once treated manufacturing plants — are capturing investment at the expense of more bureaucratically complex jurisdictions.

Future-Proofing in a Moving Target Environment

The criteria for an ideal data center location in 2030 may look different from 2024's criteria in meaningful ways. Small modular reactors — still pre-commercial but advancing — could dramatically alter the power availability calculus by enabling on-site or near-site generation that bypasses grid interconnection entirely. Microsoft's recent agreement with Constellation Energy to restart a nuclear unit at Three Mile Island is a preview of where the industry's power procurement strategy is heading.

Advanced cooling technologies, including liquid cooling and direct-chip cooling, are reducing water consumption for the highest-density AI compute environments. As these technologies mature and become standard, some of the geographic constraints around water access may loosen — slightly.

The distributed intelligence model favored by edge computing and real-time AI inference is pushing some deployment to locations closer to population centers, which runs counter to the "go where power is cheap and land is vast" logic that governs large campus decisions. The industry will need both models simultaneously: massive centralized training infrastructure and distributed inference infrastructure near end users.

The developers who are winning long-term are the ones underwriting sites on 20-year assumptions about grid evolution, not just where interconnection queues stand today.

For anyone in infrastructure development, land brokerage, or clean energy — this moment represents exactly the kind of structural dislocation that creates opportunity. Markets that once seemed too remote, too rural, or too underdeveloped for data center consideration are getting a second look. The question isn't whether the industry will expand into new geographies; it's which sites will be ready when the capital arrives.

Explore the InfraSale Marketplace for your next opportunity!

[INTERNAL LINK: data center site selection]

[INTERNAL LINK: infrastructure investment trends]

[INTERNAL LINK: renewable energy incentives]