States Gain Leverage Over Big Tech's AI Buildout

Democratic victories in Nevada, Georgia, Ohio and Texas could give state and local officials more power to slow data-center construction, putting projects and the artificial-intelligence investment boom at risk. Decisions on permits, tax breaks and electricity service often sit far from the communities expected to host the facilities, but local governments may still influence whether some projects move forward.

Who Controls the Buildout

States controlled by Democrats who support expanding or extending AI construction moratoriums will account for 45% of planned U.S. data-center capacity come January, Goldman Sachs estimates. Texas alone is expected to host a quarter of the planned AI infrastructure buildout. That concentration gives governors and local officials consequential choices as companies race to build.

Governors-elect have proposed ending tax breaks, tightening climate and electricity restrictions, and allowing local governments to impose construction moratoriums. Texas Republican Governor Greg Abbott had already imposed a data-center moratorium, and Governor-elect Gina Hinojosa campaigned on extending it until lawmakers establish stronger AI regulations. Donald Trump has opposed attempts to curb AI growth and spending, citing economic and national security concerns, especially competition with China. Many Republican politicians who might otherwise have backed AI rules, including rules for data-center construction, stayed silent.

The authority at issue is practical as well as political. State and local officials can affect whether companies finance projects, said Mona Dajani, co-chair of Cooley's infrastructure, energy and real estate practice. Hundreds of developments awaiting permits, utility service or tax incentives could be particularly exposed, she said. Projects already under construction could face risks, too.

Texas law grandfathers in permits for some projects, protecting them from later changes to local permitting requirements. But that doesn't address changes to state permitting or electrical-service guarantees that Governor-elect Gina Hinojosa could vacate.

The Costs of a Race to Build

Delays carry a bill. Interest, equipment commitments and utility expenses keep accruing while a project sits on hold, Dajani said, and existing projects could eventually become financially unviable. AI companies already struggle to build data centers fast enough to meet anticipated demand. They could then miss their spending targets.

The boom has grown large enough to shape the wider economy. ING economists say it has accounted for as much as a quarter of U.S. economic growth this year. AI infrastructure spending is expected to approach $1 trillion a year for the foreseeable future, said Stijn Van Nieuwerburgh, a Columbia Business School economist. As a share of gross domestic product, that would exceed every other investment boom in modern U.S. history, including railroad construction.

The construction rush has also left money tied up in projects that may not work. Data centers already face $200 billion in project delays, said Mike Mathews, global digital infrastructure practice leader at risk adviser Marsh. Stanford University's AI Index Report counted 5,427 data centers in the United States at the end of last year. AI companies have announced plans for 3,969 new U.S. data centers, according to Aterio, a data-center research company, but only about 800 are under construction. AI experts have questioned whether many announced projects are viable.

CoreWeave, a neocloud business that rents computing capacity, recently projected extensive losses, and its bond yields surged to 15%. Its expected return on investment was about 9%. Amazon, Google, Meta and Microsoft don't face the same problem, but their stocks have fallen in recent weeks as investors questioned spending that may not generate profits for years.

Permission, Not Just Permits

The source describes no organized community response or mutual-aid effort. It does point to a possible shift toward community-based development: AI companies could build capacity while seeking public and political permission to operate facilities over the long term. In the previous decade, Google and other companies promoted sustainable construction using renewable energy for data centers built during the cloud-computing expansion, even before it was cost-effective.

A slowdown could curb growth that has contributed to rebounding inflation and public antipathy toward AI and data centers. It could also give companies time to plan more sustainably, rather than pour money into unworkable projects. Wells Fargo estimates that each 10% reduction in planned Texas data-center capacity would reduce Big Tech companies' AI infrastructure spending by 1.6%.

The economic consequences wouldn't automatically mean a recession, said Joe Brusuelas, chief economist at RSM US. He expects growth to ease from last quarter's 3.4% pace to closer to 2.4% if AI investment slows, while remaining solid over the long term. “It's slowing, but it doesn't mean we're going to have a recession,” Brusuelas said.

“This is a rare case where state and local leaders have influence of geopolitical-level import,” said Harry Krejsa, director of studies at the Carnegie Mellon Institute for Strategy & Technology. He added: “If you squander political license to operate at home, that'll be a bigger impediment to remaining competitive with China.”