Georgia, Virginia and Arizona: America's Real AI Infrastructure Belt

Three American states host a disproportionate share of the physical infrastructure artificial intelligence depends on, and each arrived there differently.

Northern Virginia:

Loudoun County and its surroundings host the densest concentration of data centres on earth, a position built over decades from the region's early role in internet routing and its proximity to federal government demand.

An enormous share of global internet traffic passes through northern Virginia, and the cluster's existence is self-reinforcing: new facilities locate there because the network connectivity, fibre routes and operational expertise already exist.

AI has intensified pressure on a system already at capacity. Electricity demand from data centres has grown to a scale that dominates the local utility's planning, with projections requiring generation and transmission investment that will take years. Residents face the prospect of higher bills and new transmission lines, and local politics have become genuinely contentious, with land use fights over specific projects now routine.

Virginia's position remains dominant, and the constraint has shifted from land to power.

Georgia:

Georgia has become one of the fastest-growing data centre markets in the country, attracting major investment around Atlanta and beyond.

The attractions are familiar: available land, comparatively affordable and available electricity, tax incentives, and a location with good connectivity to the eastern United States. Atlanta also has a substantial technology employment base and strong engineering education at Georgia Tech, which supplies talent.

Growth has been rapid enough to prompt debate about tax incentives, electricity capacity and water use, mirroring conversations in Virginia a decade earlier. The state has also seen substantial investment in electric vehicle and battery manufacturing, which competes for the same power.

Arizona:

Arizona's position rests on semiconductor manufacturing. Major fabrication investment, including facilities from the leading foundry and established American chipmakers, has made the state a centre of American advanced chip production, supported by federal incentives intended to reduce dependence on East Asian manufacturing.

The attractions include established semiconductor operations dating back decades, a workforce with relevant experience, land availability and low natural disaster risk, an underrated factor for facilities costing tens of billions.

Water is the obvious question. Arizona faces severe long-term water constraints, and semiconductor fabrication uses substantial quantities. The industry's response has been aggressive recycling, with facilities reclaiming most process water, and the state's water debates are dominated more by agriculture than by chips, though the political salience of new industrial water demand is high.

Arizona also hosts significant data centre capacity and has been a testing ground for autonomous vehicle deployment, with commercial robotaxi services operating in Phoenix.

The common pattern:

All three states illustrate the same shift: AI's physical requirements are relocating economic activity toward places with power, land and permissive development policy, and away from the coastal metros where the software is written.

They also illustrate the same emerging conflict. Data centres and fabrication plants generate tax revenue and construction employment but relatively few permanent jobs, while consuming electricity and water that residents also need. That trade-off is now the central local political question in every American region attracting AI infrastructure.