Which US State Has the Most Data Centers ?

Virginia's 291 buildings are followed by Texas at 120, California at 109, Oregon at 106 and Ohio at 86. Washington (73), Arizona (61), Iowa (58), New Jersey (51) and Illinois (46) complete the top ten, which together hold 1,001 of the 1,378 mapped buildings, or 73%. Virginia on its own accounts for 21%.

Below that the map thins quickly: the median state has 10 mapped data centers, 33 of 51 areas have 15 or fewer and only nine states reach 50. Oregon in fourth is the surprise. It sits ahead of Ohio and Washington and is outside the top ten of the tracker Electric Choice, though Pew's chart, drawn from Data Center Map listings, has it ninth.

The Atlas stores some data centers as points, some as building outlines and some as campus boundaries, and a volunteer can draw a hyperscale campus as one polygon or a dozen buildings. This map counts one per mapped building, adding a campus or point only where no building sits inside it: 1,378, from 1,472 raw objects.

Change the rule and the leader changes. Merge everything within 1 km into one site and the national count drops to 563: Texas 53, California 42, Virginia 40, Arizona 38, Ohio 29. Virginia's buildings shrink 7.3-fold because Loudoun County alone has 189 outlines packed together; Texas's shrink only 2.3-fold, which points to more separate locations.

Iowa shows the gap: 58 buildings but 14 sites. Wyoming has 17 and 4, South Carolina 11 and 3. The Atlas does not say why, but hyperscale campuses mapped building by building would produce that pattern.

Loudoun County's economic development director, Buddy Rizer, traces the cluster to the late 1990s, when AOL moved its headquarters there and "everybody started running their fiber out to AOL." An internet exchange point, MAE-East, followed. Equinix's Jon Lin put the next step simply: "Once all the connectivity was there, all the content providers seemed to be there."

The county lists proximity to Washington and a major airport, an established workforce, a reliable power grid and a business-friendly permitting process. As of 1 March 2026 Loudoun had about 233 existing data center buildings and about 56.5 million square feet, and data centers produced $1.2 billion in property tax revenue in fiscal 2026, 39% of the county budget.

State policy helped. Virginia's sales tax exemption, dating from 2010, lets qualifying data centers buy equipment tax-free if they create 50 jobs, pay 150% of the local prevailing wage and invest $150 million. Legislative auditors at JLARC found it saved $928.6 million in fiscal 2023, the state's largest economic development incentive, used by 90% of the industry by megawatts. Industry says it is a key factor in where it builds, which is a claim, not proof Virginia would lack the cluster without it.

At the top, yes. Pew Research's April 2026 chart, built on Data Center Map listings, puts Virginia, Texas and California first, second and third among operating facilities, at 398, 296 and 277. Electric Choice, as of 29 September 2026, has 488, 389 and 230, with Georgia level with California.

Beyond that the sources split. Aterio's March 2026 pipeline count, which adds projects under construction and announced, has Texas ahead of Virginia 962 to 954 and Georgia third, with California outside its top five. Pew counts listings, Electric Choice tracked facilities, Aterio pipeline sites and this map mapped buildings, so none of their numbers is offered here as a state total.

Against Electric Choice, coverage swings from near-complete to almost empty. Ohio's 86 is 83% of its 104, Arizona's 61 is 73% of 83 and Virginia's 291 is 60% of 488. California sits at 47%, New Jersey 40% and Texas 31%. Then Illinois 21%, Georgia 15%, New York 11% (20 of 189) and Florida 9% (15 of 171).

Pew's operating counts agree on the weakest cases: New York 20 against 148, Florida 15 against 120, Pennsylvania 9 against 78. Census data rank Pennsylvania fifth for data processing and hosting establishments in 2023, with 937. These are large markets, so thin mapping is likelier than thin supply.

The map can also run close to commercial counts: Oregon's 106 against 115 operating in Pew's data, Iowa's 58 against 64. The units differ, buildings against facilities, so that does not prove the same sites are counted. The Atlas's authors call validation of the OpenStreetMap data "an ongoing development".

Five states have no mapped data centers, and DC, Louisiana and West Virginia have one each. In OpenStreetMap that more likely means no volunteer has tagged one than that none exist. The Census Bureau's 2023 County Business Patterns counted 206 data processing and hosting establishments in Delaware, 59 in Hawaii, 46 in Rhode Island, 41 in Vermont and 33 in Alaska.

Those are business locations, not buildings, and a hosting firm need not run a data center, so this is an inference, not a verified undercount. But the same file counts 119 such establishments in DC, 130 in Louisiana and 45 in West Virginia, which makes one mapped building each look like a gap.

The snapshot matters too: PNNL warns later Atlas versions may not contain every location in earlier ones, so an empty state may fill in without a single new building.

It settles where mapped data centers cluster: Virginia, Texas and California lead this map, Pew's operating counts and Electric Choice, and the top ten states hold 73% of mapped buildings. It also settles that buildings and campuses are different questions, with Virginia first on 291 buildings and Texas first on 53 sites.

It does not settle how many data centers any state has. The 1,378 figure counts the map, not the industry, and the gap to Electric Choice's 3,300+ is widest in New York, Florida and Georgia. For a number to quote, use a tracker's own figure with its own definition.

The map will move as volunteers add buildings. If a later Atlas version shows New York or Florida jumping, that will mostly measure new mapping, not new construction.