Everything About AI
NextFin News - Ernest Popescu had agreed to build a data center in Nashville, Tennessee, this summer when the Metro Council passed a moratorium that effectively banned new projects through December. The delay was long enough that the chief executive of developer Metrobloks withdrew and shifted his focus elsewhere. Across the country, the same calculation is now playing out on ballots, in governor's mansions, and in Senate races: the 2026 midterm elections will decide whether the artificial-intelligence infrastructure boom keeps building, moves to friendlier ground, or slows down for good.
The political fight is no longer confined to planning boards. Voters in eight states will decide at least 39 local ballot measures related to data centers in 2026, according to the elections tracker maintained by Ballotpedia, with 35 of them — 90% — falling on the November 3 general election. Ohio alone accounts for more than half, with 20 measures across 17 jurisdictions. But the real test runs through four battlegrounds where the outcome will set the template for the rest of the country: Nashville, Tennessee; Pataskala and the Ohio local-measure wave; Ohio's statewide Senate and governor races; and Michigan, where a new OpenAI-Oracle campus became a campaign backdrop.
The central question is not whether Americans like data centers — polls say they do not. It is whether opposition has hardened into a durable regime that raises the cost and the time of building, everywhere. The evidence points to structural change, not a campaign-season blip. That distinction matters for every investor in the AI buildout, from data-center REITs to power-equipment makers, because a structural shift in the social license to build does not revert when the campaign ads stop running.
The Four Battlegrounds
1. Nashville: Where a Moratorium Became a Model
Nashville is the clearest case of local action turning into a national template. The Metro Council enacted a countywide freeze on new data-center permits through December 1, 2026, while permanent zoning rules are written, and passed separate ordinances imposing distance buffers between facilities and schools, churches, daycares, and zoos — 100 feet for small centers, 2,640 feet for large ones — along with a ban on the largest facilities. Mayor Freddie O'Connell also signed an executive order targeting large-scale data-center development.
The practical effect is visible in one deal. Metrobloks' Ernest Popescu had committed to a Nashville project; the moratorium's delay window pushed him to walk away. Since then, he has watched financing partners grow more anxious about moratoriums spreading after the election. The pattern is self-reinforcing: a developer who cannot underwrite permitting risk stops bidding, which lowers land prices and tax-revenue projections, which makes the next council member more willing to vote no.
The backlash has legal contours too. DC Blox sued Metro Nashville in federal court in August 2026 over the moratorium and the hold on its permits for a site next to the Nashville Zoo; a federal judge dismissed the lawsuit in October but gave the company 21 days to file a revised complaint. The litigation is a warning shot: cities are learning what they can do, and developers are learning what they can challenge.
2. Pataskala, Ohio: The 25-Megawatt Line
In Pataskala, Ohio — a city of 18,500 — the most common lawn signs in the campaign's final weeks support neither a candidate nor a party. They read: VOTE YES TO BAN DATA CENTERS. The November 3 charter amendment would prohibit construction of data centers with an aggregate monthly demand or peak load greater than 25 megawatts.
That 25-megawatt threshold is the political unit of the movement. It is deliberately drawn below the size of the hyperscale AI facilities that drive the current buildout, so it blocks the projects that actually matter without touching the server rooms local officials barely notice. A statewide version of the same amendment — a ban on data centers requiring at least 25 MW — missed the 2026 ballot after failing to gather the 413,488 valid signatures needed by July 1, but it is aimed at 2027.
Pataskala is not an outlier. Ohio has 20 data-center measures across 17 jurisdictions, more than every other state combined. Nine would prohibit facilities above specified electrical-demand thresholds; three jurisdictions are weighing outright bans; others would require financial guarantees from developers seeking special tax incentives or force voter approval for large developments. The measures span bans, temporary moratoriums, zoning changes, project-specific approvals, and requirements for future voter approval.
3. Ohio's Statewide Races: The Sleeper Issue
The local measures are the ground game; the statewide races are the air war. In Ohio's U.S. Senate contest, Democrat Sherrod Brown has called Republican incumbent Jon Husted "the face of data centers in Ohio" and blamed state policy for higher energy bills. A National Republican Senatorial Committee memo warned in August that Brown's attacks could make data centers a "sleeper issue" in a race forecasters call a toss-up.
The governor's race has moved even further. Democrat Amy Acton has proposed a "conditional moratorium" on data centers. Republican nominee Vivek Ramaswamy, who had championed bringing the industry to Ohio, released an "Ohioans-first data center pledge" calling for legislation and, until it passes, a moratorium on data-center construction in the state. He has also said he would sign an executive order on his first day halting new project announcements until that law is in place, and wants tech companies to provide free electricity to the communities that host them.
When both nominees in a swing-state governor's race are competing to be tougher on the same industry, the political center has moved. The question is whether that movement survives inauguration.
4. Michigan: The OpenAI-Oracle Backdrop
Two months after OpenAI and Oracle broke ground on a massive data-center campus in Saline Township, Michigan, Democratic U.S. Senate candidate Abdul El-Sayed held a rally in front of the site, using the controversial project as a backdrop for his insurgent campaign.
I stand with local and state elected officials saying that we cannot approve any more of these until we have federal-level guardrails.El-Sayed narrowly won his primary.
Michigan's version of the fight is less about ballot measures — the state has only one tracked measure — and more about whether a flagship project becomes a symbol. The Saline Township campus gives opponents a physical place to point to, the way a closed factory once gave labor campaigns a backdrop. That symbolism is what travels across state lines.
The Numbers Behind the Backlash
The political revolt is running against an unprecedented buildout. U.S. data-center power demand is forecast to more than double to 66 gigawatts in 2027 from 31 gigawatts in 2025, according to Goldman Sachs Commodities Research, climbing through 41 GW in 2026. Scheduled capacity additions are set to reach 13.6 GW in 2026 and 36.3 GW in 2027, compared with realized additions of 6.4 GW in 2024 and 8.5 GW in 2025. The Energy Information Administration projects total U.S. electricity demand will hit record highs of 4,195 billion kilowatt-hours in 2025, 4,270 billion kWh in 2026, and 4,349 billion kWh in 2027.
Only about half to 60% of the data-center capacity scheduled for the next one to two years is expected to come online on time, the same Goldman Sachs analysis found, because of delays and cancellations. That is the vulnerability: a buildout already running behind schedule cannot absorb new political friction without projects slipping further or relocating.
Public opinion gives the movement its fuel. A Gallup poll earlier this year found 7 in 10 Americans oppose the construction of an AI data center in their area. A separate national survey found 64% of respondents did not believe rapidly building data centers was a good thing, 77% worried the facilities could raise their electricity rates, and only 14% said they would be willing to live near one. The concerns cut across party lines, which is why candidates from both parties are now triangulating against the industry.
Why This Is Structural, Not Cyclical
The first-order read of the midterms is that a few races turn on a few ballot measures. The second-order question is whether the outcome changes the cost of capital for the entire AI buildout. Here the distinction between cyclical and structural matters, and the evidence favors structural.
Three tests separate a cycle from a regime change. First, who is driving it? In 2026, 82% of the data-center measures qualified for the ballot through citizen-initiated petition processes, not referrals from local governing bodies. Grassroots petitions are slower to build but harder to reverse than official decrees, because they embed the opposition in the electorate rather than in a single administration.
Second, is the response temporary or permanent? The Urban Institute reported in September that 313 moratoriums had already been enacted across 44 states and more than 200 cities, counties, and other localities. States are not just pausing; they are rewriting the rulebook. New York suspended discretionary environmental permits for new hyperscale facilities for up to a year while agencies study energy demand, water use, air quality, and local impacts, and proposed making large facilities pay more for energy or provide their own supply. Pennsylvania now requires developers seeking state permits to make legally binding commitments to cover project-related power and grid costs, obtain local approval, and provide energy and water data. Virginia created an energy-consumption tax projected to generate $600 million a year and urged regulators to assign data-center-driven transmission costs to the facilities themselves. Texas ordered a halt to new data-center permits while state agencies audit electricity and water use.
Third, does the driver self-correct? A cyclical backlash fades when the perceived harm fades. But the harm here is structural to the technology: AI training and inference workloads are power-hungry by design, and the facilities that host them are large, noisy, water-intensive, and visibly industrial. The gap between what the facilities promise — a handful of permanent jobs and some construction work — and what they consume — tens of megawatts and millions of gallons — does not close with better messaging.
The counter-thesis is real and deserves weight. The industry argues that strict rules simply redirect investment to friendlier jurisdictions rather than destroying it, and that the economic prize — construction jobs, corporate tax revenue, a place in the AI economy — still wins most fights. Dan Diorio of the Data Center Coalition said New York's moratorium would send jobs and investment elsewhere. States are also learning to negotiate: some developers have pledged higher utility payments, lower water use, and community contributions. The Urban Institute found those commitments often lack an external enforcement mechanism, but the pledges themselves show the industry is adapting, not surrendering.
There is also a capacity argument: compute demand continues to far outstrip available supply, so even delayed projects eventually get built somewhere. Apollo's chief economist Torsten Slok has argued that a slower-than-expected data-center buildout would reallocate gains rather than erase them — owners of existing physical capacity would benefit first, while AI growth stories tied to future capacity would be punished. That is a meaningful hedge against the bear case, but it is a redistribution, not a refutation: someone still loses when the buildout slows.
The falsifying signal is concrete. If fewer than half of the 35 November ballot measures pass, or if the winners of Ohio's and Texas's governor's races reverse course in 2027 and actively re-court unrestricted buildout, the structural thesis weakens toward a cyclical campaign blip. Either outcome would say the opposition was broad but shallow — loud at the ballot box, quiet once the incentives arrive.
Who Benefits, Who Is Exposed
The mechanism cashes out unevenly across the AI infrastructure complex. The first-order losers are developers and landlords with projects midway through permitting in states where the measures are concentrated — Ohio, Illinois, Wisconsin, Kansas, California, Michigan, Nevada, and Florida — and the heavy debt that leaves little room for delay. A charter amendment that blocks a 25-megawatt facility does not wait for a financing model to adjust.
The first-order beneficiaries are owners of already-permitted, already-powered capacity. When new supply slows, occupancy and pricing power for existing campuses improve — that is the Slok trade in practice. Power-equipment and grid-construction companies — the Vertivs, Quanta Services, and GE Vernovas of the market — face a more mixed picture: their order books are tied to announced projects, and announcements are precisely what moratoriums chill in the short term, even if the long-term demand case is intact.
Utilities sit in the middle. Data-center opposition is, at its core, a bill-allocation fight: residents do not want to pay for the transmission and generation that hyperscale loads require. States that force developers to bring their own power or pay for grid upgrades — Pennsylvania, Texas, Virginia — protect ratepayers but also create a moat: only the deepest-pocketed tenants can clear the bar, which concentrates demand among the largest operators and the utilities that serve them.
What to Watch After November 3
The short-term read is volatility. The 35 November measures will produce a patchwork of outcomes, and markets will react to the headline pass rate before the fine print is digested. A high pass rate would pressure data-center developers and REITs with concentrated exposure to the eight ballot states; a low pass rate would be read as a reprieve for the buildout.
The medium-term read is policy. The measures that pass will be tested in court — Nashville's litigation is the preview — and rewritten in state legislatures when they convene in January 2027. Watch Ohio's governor and legislature first: with half the nation's measures and both parties already competing on the issue, whatever Ohio codifies will be copied or fought in every other swing state.
The long-term read is the cost of capital. If the structural thesis holds, underwriters will start pricing permitting risk into data-center debt the way they price construction risk today, and the hurdle rate for new projects will rise permanently. That is the real fate of data centers being decided in November — not whether a few projects get built, but whether the entire asset class is re-priced for political risk.
The midterms will not stop the AI buildout. What they can do is make it more expensive, slower, and more concentrated — and reward the players who already own the power, not the ones still asking for permission.
Explore more exclusive insights at nextfin.ai.
