OpenAI's decision to start calling its data centers "AI compute campuses" landed the same weekend that candidates backing data center moratoriums appeared competitive in up to 15 of 36 U.S. governors' races4,1. The rebrand is a messaging exercise. The political exposure is structural. If moratorium-friendly candidates win even a fraction of those races, the permitting environment for the 2028 buildout changes materially—though how much depends on whether new governors can actually push moratoriums through their legislatures.
The thesis: political risk to AI infrastructure is crossing from local nuisance to systemic variable. It may not block the buildout outright—the empirical record of local moratoriums suggests most capacity survives—but it now sits squarely in the path of the industry's most capital-intensive year, and the range of outcomes Tuesday will narrow is wide enough to matter.
SemiAnalysis models 62.2 GW of U.S. net data center additions in 2028, most of which has not started construction2. The 2027 pipeline is largely insulated: roughly three-quarters of the 33.8 GW due across the top 20 states is already under construction. But the 2028 tranche is different. It still needs approvals, power interconnections, and construction schedules that will be set under whichever governors take office in January. The 13 states with the largest 2028 net additions account for 43.5 GW, or 70% of the national total, and 10 of those 13 elect a governor on Tuesday.
The Information's analysis found that candidates backing a data center moratorium could win in up to 15 of 36 governors' races, including in Texas, Ohio, Michigan, and New York. If the pro-moratorium candidate won in every state where one is running, those states would together contain roughly 30 GW of planned 2028 capacity. That is a ceiling, not a forecast, and the headline should be read accordingly. SemiAnalysis noted that if the less restrictive candidates prevail instead, the exposed capacity falls to 5.7 GW. The gap between those two figures spans the full range of plausible outcomes, and the realistic landing zone depends on which specific races break which way—capital planners should weight the competitive subset, not the ceiling.
Texas alone illustrates the stakes. The state adds close to 30% of U.S. gigawatts in 2028. Governor Greg Abbott, facing Democratic challenger Gina Hinojosa, has veered from his typical pro-business stance in response to community outrage over data center development7. Hinojosa would impose a moratorium until legislation gives communities a voice, protects household utility bills, and ends tax incentives for data centers. The regular Texas legislative session runs from January 12 to May 31, 2027, and the Electric Reliability Council of Texas is targeting December 10 to file a report on grid impacts. The Texas legislature is expected to remain Republican-controlled; a Hinojosa moratorium would require either bipartisan buy-in or executive action whose legal durability is untested. Even so, the overlap between a new governor's inauguration and the legislative calendar means permitting uncertainty—whether from a formal moratorium or from the political signal that one is being pursued—could stretch well into 2027, compressing the construction window for 2028 capacity.
At a Bloomberg Invest panel in Dallas, Crow Holdings' Michael Hyun characterized a crackdown on data center development as "economic suicide". But he also acknowledged the double-edged nature of the boom: "You're competing for many of the same contractors to develop your site and guess who can pay more?". The frenzy has drawn real estate's finite resources toward data centers, creating cost pressure across the Texas construction market.
The backlash is bipartisan. Bloomberg Government identified nine AI battleground races that could swing the next Congress5. Christiana Lee Armstrong, senior AI and government reporter at Bloomberg Government, described the dynamic: "This is likely the first federal election that we've seen, where candidates across the country are really having to stake out positions on issues related to eye safety, concerns about jobs, about data centers". In Ohio, former Senator Sherrod Brown has attacked Senator John Hughes over past support for data center-friendly policies. Democrats have used data centers as a vulnerability against Republicans, but Republicans have also tried to outflank Democrats on the issue.
New York offers a preview of what moratoriums look like in practice. The state enacted the first statewide moratorium on data center development, but nearly four months in, organizer Adrienne Yocina told Wired she is "getting nervous" that Governor Kathy Hochul will not sign stronger legislation. "I feel like she's just trying to shut us up," Yocina said of Hochul8. New York's experience suggests that even where moratoriums pass, the regulatory follow-through is uncertain, creating a different kind of delay risk: not outright prohibition but prolonged limbo. That limbo may be the more realistic threat than outright cancellation—and it is the threat that the empirical record of local moratoriums does not measure.
Into this environment, OpenAI is preparing a post-midterms policy and PR campaign. Chris Lehane, the company's chief global affairs officer, told Axios that the midterms will be "a bit of a movie trailer for what you re going to see in 2028," when he expects AI to be "the issue" in the presidential election. The company's proposals include developing AI "speed limits," pushing the U.S. to take a role in developing global standards, and rebranding data centers as "AI compute campuses". An OpenAI spokesperson confirmed to Business Insider that "compute campuses" is the company's new preferred term, saying they "better describes the scale and role of the infrastructure we re building for AI"6.
Lehane described the conversation around data centers as "good and healthy" and said, "Communities should hold companies responsible and accountable so they really and truly get a transformative benefit from these projects". The company is simultaneously scaling its infrastructure ambitions—and the capital stack behind them—while trying to soften the political ground beneath those ambitions. The broader talent churn across hyperscalers, including departures at AWS and Microsoft, and Google's hire of Barret Zoph as VP of research, suggests the infrastructure leadership market is repricing in tandem with the buildout.
SemiAnalysis argued in September that datacenter moratoriums were not killing the U.S. buildout: of roughly 20 GW inside local restrictions, only 1.5 GW faced actual delay. That track record is the strongest counterargument to this piece's thesis, and it held through October. But the September analysis predated the current wave of gubernatorial campaigns making moratoriums a central promise. The distinction is not that statewide moratoriums will necessarily bite harder—New York's four-month moratorium has produced activist frustration, not enforcement—but that they change the mechanism: a governor with a legislative mandate can freeze permitting across an entire state pipeline simultaneously, whereas local moratoriums pick off individual projects. Whether that theoretical leverage translates into actual delay depends on a variable the local-moratorium track record does not capture: legislative control.
A Pew survey over the summer found that only 16% of Americans have a positive view of AI's future role in society. That number is the demand signal politicians are reading. The rebrand from "data center" to "compute campus" is a response to a sentiment problem that runs deeper than nomenclature—and deeper than the "speed limits" framework Lehane is floating. Adding political risk to the financial complexity already embedded in late-stage pipeline financing compounds the discount rate on capacity that has not yet broken ground.
What to watch: Tuesday's results in Texas, Ohio, Michigan, and New York will determine whether the moratorium-exposed capacity lands closer to 30 GW or 5.7 GW. In Texas specifically, the December 10 ERCOT filing will reveal how the grid operator frames data center load growth for the incoming administration. OpenAI's post-election policy push, which Lehane signaled would begin immediately after the midterms, will test whether "speed limits" and "compute campuses" gain traction with a potentially reshaped Congress3. Most governors cannot impose statewide moratoriums unilaterally; they need legislative majorities, and in states like Texas and Ohio, pro-business legislatures may block or water down moratorium proposals regardless of the governor's campaign promises. New York's four-month experience suggests the gap between campaign promise and regulatory action is where the real delay risk lives.