Voters Approve 38 of 44 Ballot Measures Restricting Data Centers

Voters in ten US states approved 38 of 44 ballot measures restricting data centers in yesterday's elections, an outcome that now threatens the pace of the country's rapid AI buildout. Redbud Consulting, a polling firm that tracked data center initiatives on the ballot through its datacentersontheballot.com project, counted the results across the ten states. The approved measures include dozens of moratoriums and outright bans on new facilities.

What do the results cover?

The 44 measures appeared on ballots in ten states, and 38 have passed so far, according to Redbud Consulting's count. The restrictions target new data center construction rather than the existing fleet. Analysts following the results suggested the measures will not affect data centers already under construction or approved through 2027. Instead, they will largely hit projects planned for 2028 and beyond.

That timeline matters for the AI industry. Companies have committed to a fast expansion of computing capacity, and the pipeline of approved sites gives them roughly two years of headroom. Anything planned for later in the decade now faces a patchwork of local bans and moratoriums.

Only one measure on the ballot allowed a data center to proceed. Even that single approval, in Harwood, North Dakota, came with strings attached: developers must pay for independent reviews and hold public hearings when they expand existing sites or propose new projects. In plain terms, even the industry's one win at the ballot box carried built-in friction.

How did candidates fare?

The backlash extended beyond ballot measures. Voters also elected officials who campaigned on promises to crack down on AI and data centers. Several of those victories came in key states that are building out the most data center capacity in the United States, which places critics of the buildout in positions where they can influence permitting, zoning and utility decisions.

The combination is unusual. Local governments have typically competed for data center projects with tax breaks and cheap power. This week's results show voters in some of the most active construction states pushing in the opposite direction.

Why does Big Tech read this as politics, not policy?

The industry's response has been to frame the defeats as partisan rather than substantive. Big Tech companies claim the voter backlash against AI is "anti-Trump" rather than "anti-tech," tying the local results to opposition to the president's push for a large, rapid American AI buildout.

That framing carries commercial weight. If the restrictions reflect general dissatisfaction with the administration, companies can argue the mood may shift with the next election cycle. If the restrictions reflect durable local concerns — power costs, water use, land, noise — then the approvals problem will persist regardless of who holds office.

The evidence available so far does not settle that argument. What it does show is a broad pattern: 38 restrictions passed across ten states, one approval passed, and that approval still imposed new obligations on developers.

What happens to the buildout now?

The immediate effect on capacity is limited. Projects under construction and approvals already granted through 2027 remain intact under the analysts' reading of the measures. The pressure lands on the planning stage, where developers must now assess which jurisdictions will accept new sites after 2028 and at what cost.

For the AI sector, the risk is not a sudden halt but a narrowing map. Every additional moratorium or ban removes a candidate location, and the elected officials who campaigned on crackdowns will control many of the decisions that determine whether future projects clear local review.

The industry's claim that this is anti-Trump sentiment rather than a rejection of its technology will be tested at the next round of elections — and in the meantime, the 2028 pipeline just became harder to build.