A new survey from TRG DataCenters ranks Virginia as the top-positioned US state for AI data center growth, based on an analysis of power capacity, grid carbon intensity, and water stress across 24 states. The August 2026 report suggests that as AI infrastructure demand grows, smaller markets with the right mix of power and water resources could increasingly compete with established hubs.
Virginia’s Lead, With a Caveat
Virginia scored 100 on TRG’s readiness index, ahead of Ohio (88.2), Tennessee (83.7), Georgia (81.8), and Texas (80.8). TRG mapped seven large AI sites in Virginia totaling 1,085 MW of capacity, less than Ohio’s 1,410 MW or Tennessee’s 1,373 MW. Virginia’s edge comes largely from a lower-carbon grid: TRG CEO Chris Hinkle said Virginia’s mapped facilities are associated with around 2.3 million tons of annual CO2 emissions, compared to roughly 4.5 million tons in Ohio and 4.3 million in Tennessee, despite Ohio’s larger mapped capacity.
Water remains a concern for Virginia, though. About one-third of its mapped data centers sit in areas of high water stress, compared to none in Ohio and just 7% in Tennessee. Hinkle said site selection around water availability is something Virginia will need to keep managing as the market grows.
Smaller States Punch Above Their Weight
South Carolina placed sixth with a score of 80.5, nearly matching Texas despite having just one large AI site and 142 MW of mapped capacity, versus Texas’s 13 sites. Hinkle attributed this to South Carolina’s nuclear-heavy power supply and its lack of facilities in high water stress zones. He also flagged New York, Iowa, and Mississippi as markets worth watching, citing New York’s low-carbon grid and Mississippi’s existing AI capacity of more than 500 MW.
Texas’s Siting Problem
Texas had the most mapped AI sites in the study (13, totaling 1,301 MW), but 58.8% sit in areas of high water stress. Hinkle expects this to push developers toward more sophisticated siting and cooling strategies, including closed-loop or air-based cooling and recycled water, rather than slowing growth outright.
Readiness Isn’t the Whole Picture
TRG cautions that its water-stress measure describes facility locations rather than actual water consumption, and that statewide advantages don’t guarantee a specific site has adequate grid connections, transmission capacity, fiber, or long-term power availability. Hinkle said those project-level factors will ultimately determine where development lands, potentially producing a more geographically diverse AI infrastructure market as developers match projects to the best-suited locations.