Image: dotshock/Envato
Texas is fast becoming America’s AI power base. As gigawatt-scale data centers chase cheap energy and quick permits, the Lone Star State gains on Virginia’s long-held lead.
Data center site selection used to be simple: choose a location near a major metro area with strong fiber connectivity and preferably outside a disaster zone.
Those factors still matter — but a new one now dominates the decision-making process: power availability.
The latest generation of data centers that are being planned or built to host vast arrays of Graphics Processing Units (GPUs) demand power capacities of a gigawatt (GW) or more. These AI factories are gravitating to non-traditional locations such as West Virginia, South Dakota, and Alberta, Canada.
But the biggest beneficiary of these new data centers? Texas.
According to DCByte, Texas is beginning to show signs that it might eventually challenge Virginia as the top state for data center capacity.
Virginia leads with 24% of all US data center capacity. Texas is second with 15% and rising fast. It leads Virginia in both new construction and planned construction. As much as 16 GW of new data center capacity is expected to be added in the Lone Star State between 2026 and 2030.
Further expansion is far more difficult for Virginia. It may be strategically located near Washington, DC, with access to government markets that have deep pockets, but it faces severe grid constraints.
The local utility famously wrote a letter to data center developers a couple of years ago, saying they had no more power available. Additionally, several local communities are opposed to the construction of more data centers in their areas, citing concerns about electricity prices and water usage. Texas is far more welcoming.
“Texas’s fast approvals and low-cost gas win when it comes to site selection,” said David Bell, VP of Utility and Microgrid Development at VoltaGrid, told TechRepublic at Data Center World Power. “Texas is the best place for access to firm, redundant power fueled by natural gas and renewables.”
Rather than waiting for utilities to deliver more power, many data centers are finding ways to generate their own on-site power. They not only control their own energy. They can make it available faster. In some cases, developers are even bringing in large numbers of natural gas generators onsite to provide power during the construction phase.
“50% to 65% of all data centers will be running onsite power within 10 years,” said Bell.
The Lone Star State’s proximity to massive natural gas fields translates into cheap, abundant, and reliable power. In addition, the state benefits from a favorable regulatory framework that Bell said enables permitting and approvals to be processed three times faster than in other states.
The Texas grid authority (the Electric Reliability Council of Texas, or ERCOT) is independent of other US regional grids. Thus, it is not subject to an additional layer of oversight, bureaucracy, and approvals at the federal level.
“The ERCOT grid is preconditioned for change; as we control our own grid, we can be more agile,” Woody Rickerson, senior VP and COO of ERCOT, told TechRepublic. “Data centers can’t be treated the same as powering grandma’s house as they are asking for GW-sized loads. We are in a race right now to power AI and data centers: you don’t enter a race unless you are willing to run.”
Compass Datacenters, for example, has projects operating or under development all over the US. Company CEO Chris Crosby favors Texas, where it has a 320 MW site in Meridian and another 360 MW project at its Red Oak campus in the Dallas-Fort Worth area.
“Gaining new grid capacity can be difficult,” Crosby told TechRepublic. “There is a smart mentality towards growth in Texas that doesn’t exist in some other areas.”
Virginia may currently house the largest concentration of data centers in the world, located in what is known as “Data Center Alley.” But Texas has more data center development planned by the end of the decade than the following six states combined (including Virginia).
Between now and 2035, Texas is likely to add at least four times as much new data center capacity as Virginia – much of it large hyperscaler AI factories.
And don’t discount the importance of the cost of energy. According to Bell, the average cost of gas in Texas is about 25% cheaper than in Virginia. Such a gap amounts to millions of dollars in energy savings alone for a 1 GW AI data center, he added.
When paired with abundant fuel supplies and a rapid-approval environment, those energy savings tilt the economics sharply in Texas’s favor. And as demand for power-hungry AI infrastructure continues to surge, states that can deliver cheap, fast, and reliable electricity will shape where the next generation of digital infrastructure takes root — with Texas quickly emerging as the benchmark.
And if you’re tracking where hyperscalers are planting their next mega-sites, take a look at our report on OpenAI’s $7 billion Michigan data center campus — what it means for power procurement, jobs, and regional grid planning: OpenAI’s $7B Michigan Data Center, a Major Stargate Push.
Originally from Scotland, Drew Robb has been a full-time writer for 25 years. He lives in Florida and specializes in IT, engineering, and business. As well as TechRepublic, he writes for a wide range of magazines including Gas Turbine World, SDxCentral, HR Magazine, and eWeek.
