Voters back data center limits as Texas result raises stakes for AI expansion

The same report cited research firm SemiAnalysis’s estimate that Texas accounts for 41% of new U.S. data center capacity in 2028. It said a hold lasting more than six months would bring extended delays there. That is a share of anticipated new capacity in one year, rather than the country’s entire installed data center base.

The timing limits the immediate implications. Most U.S. data centers expected in 2027 are already under construction, the report said. It also reported Democratic victories in Ohio, Georgia and Arizona—the states following Texas in new capacity—with the winners in Ohio and Georgia favoring restrictions whose details remained unclear.

The report’s broader assessment that nearly two-thirds of 2028 capacity could face delays is consequently a scenario of political exposure, rather than a tally of canceled projects. It also noted that most political analysts did not expect Democrats to block construction permanently. Winning office or a ballot measure does not by itself specify which facilities must stop work.

Berkeley Lab estimates U.S. data center electricity use at 58 TWh in 2014 and 176 TWh in 2023, according to DOE’s December 2024 summary. Source: U.S. Department of Energy — data center electricity demand.

Electricity demand does not settle who pays

The power challenge predates the election. The U.S. Department of Energy’s December 2024 summary of research by Lawrence Berkeley National Laboratory put U.S. data center electricity use at 58 terawatt-hours in 2014 and 176 terawatt-hours in 2023—roughly a tripling over that period. The laboratory’s report, funded through the department’s Industrial Efficiency and Decarbonization Office, covered energy use from 2014 through projected demand in 2028.

It estimated that data centers consumed about 4.4% of U.S. electricity in 2023 and projected 325–580 terawatt-hours in 2028, approximately 6.7–12% of national electricity use. Those figures cover data centers collectively, rather than AI alone; the 2028 range is a forecast made in 2024, rather than a measurement of demand following this election.

National consumption estimates do not establish the effect of a particular facility on household bills. The Energy Department’s Office of Policy outlined the relevant financial questions in a January 2025 technical brief on rates for large electricity customers: how to allocate system costs without unfairly shifting them to other customers, protect against underused utility investments and manage the risk that demand exceeds supply.

Those issues leave room for responses besides restrictions on construction. The brief drew on pending and approved rate structures, agreements and special contracts, while recognizing data centers’ role in innovation and economic growth. The department’s 2024 release also identified onsite generation and storage, reuse of infrastructure at retired coal facilities and innovative electricity rates as strategies for accommodating demand.

These approaches address different questions. A development restriction determines whether or when a facility may proceed. A rate agreement addresses its electricity costs and the financial risk borne by utilities and other customers. Neither the election tally nor the national demand forecast establishes what any individual community will pay or save.

Berkeley Lab’s January 2025 technical brief lists large-load tariff examples and their regulatory status as of January 1, 2025. Source: U.S. Department of Energy — electricity rate design.

The next test is which projects the rules reach

In Texas, the next consequential decisions concern the promised moratorium and the legislation Hinojosa wants before it ends. Their scope and timing will determine how the policy affects projects scheduled for 2028, alongside facilities already being built. In Ohio and Georgia, the reported preference for restrictions still needs to become a defined policy before its construction effects can be assessed.

For local ballot measures, final returns, operative text, effective dates, duration and treatment of existing approvals are needed to distinguish temporary pauses from lasting prohibitions. The evidence needed to turn an electoral result into a capacity estimate is a project-by-project accounting of those rules. The votes show support for limits; how much computing infrastructure is delayed, and who bears its electricity costs, remains unresolved.