Voters across 40 localities in ten states considered about 45 measures on data center construction, with the results leaning heavily toward tighter limits on building. Some results were still up in the air. The votes added another hurdle for data centers, which hold the computing equipment used for AI and need networks to connect that equipment with users. Network operators may now have a harder time knowing where the next round of AI capacity is going to be built, and they may also have a harder time knowing when it will be ready for them to use.
Key Points
- In Ohio, 21 measures concerned data centers.
- In Harwood, in North Dakota, voters rejected a proposal that would have required reviews for industrial developments, data centers among them.
- A tracking poll cited by New Street Research found that opposition went from 42% in August 2025 to 75% in August 2026.
- SpaceX hopes that deployment of its AI computing facilities in orbit can start in late 2027.
Voters Back Broader Restrictions on Building Data Centers
The results in so far showed restrictions winning in nearly every place, but several votes were still undecided. Separately, New Street Research analyst David Barden reported that there were 45 votes for Nov. 3, and restrictions had either passed or were ahead in 39. There is still some uncertainty about the final count, but both accounts point to broad opposition to more building. Harwood was the one place where the reported result went the other way and favored development.
Voters were able to slow down or stop building in a few different ways on their ballots:
- Some measures banned data centers outright.
- Some measures set limits on how much more could be built.
- Some asked voters to turn down proposals that would have helped development.
Voters backed restrictions in Ohio and Wyoming, and also in California, Illinois, Florida, Maine, Kansas, Nevada and New Jersey. Some states were already putting projects on hold or adopting moratoriums before voters went to the polls, which means the results add to difficulties that companies were facing as they tried to build additional computing capacity. For networks serving AI facilities, there is also the question of planning ahead, and whether a proposed site will actually get built. Plans for connections and capacity at that site may depend on construction getting the go-ahead in the first place. The results reported so far, though, do not give any figures for what the votes might mean for spending on networks or for traffic.
AI Spending Runs Into Local Opposition
Barden put spending on AI capital projects at $690-725B for 2026, and Barden’s analysis said roughly $198B worth of U.S. data center developments had been stopped or delayed by local resistance during the first six months of 2026. Those amounts put a value on the projects facing opposition. They do not show how much capacity might end up being canceled or how long projects facing delays could have to wait before construction starts.
The poll findings that New Street Research cited suggest the political pressure comes from both parties, not just one.
| Party | Share opposed to data centers | Analysis citing the polling |
|---|---|---|
| Democrats | 69% | New Street Research |
| Republicans | 54% | New Street Research |
For Independents, the opposition figure was 53%. More than 60% were opposed in each type of community—cities, suburbs and rural areas—suggesting that moving a project outside city limits would not necessarily help builders avoid resistance.
There is also the question of how people feel about AI itself. According to figures Barden cited, 73% of Americans were worried about greater AI use, up from 68% in 2023. When asked what effect AI had on society, 39% said it was negative, while 11% said it was positive. In all 25 countries surveyed, another response was more common than excitement about AI. Those numbers are about how people feel about AI in general, and they are not about the way any one person voted on a data center measure.
What the Limits Could Mean for Prices, and Musk’s Orbit Idea
The industry should do a better job of explaining how people already benefit from AI, Barden said. In Barden’s view, people want what comes out of data centers, so demand is coming from consumers, and he says that limiting construction could lead to higher bills for customers and possibly more money for the operators that are already running facilities.
That is Barden’s assessment, though. No numbers were provided to show how far prices might rise, or what changes there might be in demand for networks. Elon Musk, who heads SpaceX as CEO, returned to his proposal for computing facilities in orbit after voters went to the polls. SpaceX hopes deployment can begin in late 2027, and that would mean putting AI equipment in a very different place. Deployment has not begun, however, and putting those facilities into orbit is still something SpaceX plans to do rather than something already happening.
The questions facing people managing networks and working in IT are more immediate, and remain about facilities on the ground. Forecasts suggest AI spending will be high, but those forecasts do not identify the places where facilities will go, and they do not give dates for when those facilities can be used. There is also no answer in the forecasts about how much capacity delayed projects will ultimately make available.