Ohio’s data-centre ballots turn AI’s power demands into a test of local consent
On 3 November 2026, voters in 18 Ohio municipalities considered 21 questions about data centres. That was more than half of the 41 data-centre-related measures on ballots in ten states, according to Reuters, Reason and Fortune. By the following morning, Fortune reported that at least 14 Ohio municipalities had passed bans or other restrictions, based on early unofficial results.
Ohio has more than 200 operational data centres and ranks fifth among US states by number of data-centre projects, Fortune reported. State data cited by the publication points to as much as $40 billion in planned AI infrastructure investment by 2030. Fortune also reported that 125 local data-centre moratoriums were already in place, a figure from its own tally that has not been independently verified.
The votes are a test of whether AI infrastructure can be sited through ordinary local politics, one charter amendment or zoning question at a time. The question is not simply whether Ohioans like data centres. It is how local ballots became a test of public consent for AI infrastructure, and what the early results can actually change.
A patchwork, not a referendum
Fortune reported that 13 municipalities voted on bans or heavy restrictions, mostly aimed at large campuses using 25 megawatts or more. Two townships, Monroe and Sprigg, voted on zoning restrictions intended to preserve agricultural land. Wilmington held four referendums on whether to uphold rezonings for a data-centre project. Nationally, The Times of India, citing Ballotpedia, reported that 28 of the 41 measures would limit development if approved while 13 would facilitate it, and that some questions were advisory. A vote could therefore be a binding charter ban, a zoning decision, a temporary pause or an expression of preference with no direct legal effect.
Fortune reported that in Sunbury residents appeared to have passed a charter amendment banning data centres using 25 megawatts or more, with 100 per cent of the unofficial vote counted. The measure was aimed at a possible hyperscale Amazon project, but the project’s fate remained unclear. In Trenton, a ban passed 80 per cent to 20 per cent. Organisers hoped it would keep an 18-building Amazon Web Services data centre out of the city, although Fortune reported, citing local reporting, that it likely would not affect a four-building AWS campus already under construction nearby. Pataskala voters approved a charter ban on large data centres.
In Perry, more than 87 per cent of voters approved a requirement that developers seeking tax breaks offer the village financial guarantees. In Wilmington, around 80 per cent rejected rezoning across four referendums meant to accommodate a data-centre project. Ashville residents voted to block a data-centre campus already under construction. Reason reported that voters in Upper Sandusky, Conneaut, Trenton, Pataskala and Piqua added data-centre bans to their town charters, while the Ohio towns of Oregon and Sunbury limited their bans to hyperscalers. That distinction, Reason said, preserves development of smaller enterprise data centres.
Fortune’s tally covered at least 14 municipalities. Cleveland.com reported that the 11 proposed charter amendments it tracked were all on course to pass. The Times of India, attributing a count to Conserve Ohio, said 10 local measures the group helped draft had passed. The figures may cover different categories, but the reports do not reconcile them, and the supplied reporting included no certified canvass. The early evidence supports a narrower conclusion: many local measures appeared to pass, not that one final number is settled.
Why the votes happened
Reuters reported before the election that opposition crossed party lines. Jessica Baker of Conserve Ohio, an advocacy group, said concerns about data centres’ electricity and water demands were driving the pushback, including worries about health risks to children even where developers generate their own power. That is an organiser’s characterisation, not a measured finding, but it points to the issues that recur in the reporting: power demand, water use, tax breaks, public oversight and the terms on which communities are asked to accept large facilities.
Fortune reported that a Gallup poll published in May found about 70 per cent of Americans opposed data-centre construction in their local areas. It also reported that an ICAPP and SurveyUSA poll of 1,650 Ohio adults found 55 per cent opposed data-centre construction in the state, 65 per cent opposed tax breaks and 57 per cent said they would vote in favour of a proposed state constitutional amendment banning new large-scale data centres above 25 megawatts. Those figures were reported by Fortune and have not been independently verified. Fortune also reported that Keith Durkin of Ohio Northern University attributed negative sentiment to fear of environmental and economic risks and to limited public knowledge about data centres. His comment is an interpretation, not a measured cause.
The industry counterweight
The early results do not show that local opposition has stopped the AI buildout. Reuters reported that US data-centre capacity under construction rose 48 per cent from December 2025 to June 2026, measured in gigawatts, according to Cushman & Wakefield. That figure measures construction, not completed capacity, and it predates the November votes. Reuters also reported that Baron Fung of Dell’Oro Group expected developers to steer projects elsewhere rather than abandon them if Ohio voters approved restrictions. The industry already faced years-long waits for grid connections and rising construction costs.
That expectation points to local bans redistributing AI infrastructure rather than reducing it, although the early results do not show whether developers will follow that path. Campaign spending is another reported detail that does not settle causation. Cleveland.com reported that technology companies and business groups spent more than $400,000 trying to defeat local restrictions. The figure comes from a newsletter summary and does not establish that spending caused or failed to cause any particular result.
What the votes can change
Cleveland.com reported that the local results were already fuelling calls for a statewide moratorium on new data-centre development. The reporting described calls, not an enacted policy. The Times of India also reported that Conserve Ohio hoped to put a statewide proposal before Ohio voters next year; that is an advocacy group’s stated aim, not an adopted policy. The immediate questions are local and legal. A charter ban, a zoning refusal or a financial guarantee requirement can shape what a planning authority may approve. It does not by itself halt every project, and it does not settle how officials will apply the new rules.
The early results do not resolve the legal effect of each measure, how officials will apply them or what will happen to specific projects. Some results remain uncertified, and the fate of the Sunbury project remains unclear. The measures can change local rules and local politics; they cannot, on the early evidence, be treated as a completed national verdict on AI infrastructure.
That is why Ohio is a useful test case rather than a simple scoreboard. It shows that data-centre siting has become a question of public consent, argued through ratepayer concerns, land use, public oversight and grid capacity. It also shows the limits of that consent. The next stage is less dramatic: certification, legal interpretation, zoning administration, utility negotiations and developer choices. A local majority can reject a project in principle. Whether it can govern the infrastructure that follows is the harder question, and the one the early results leave open.
A local charter ban can change what a planning authority may approve, yet it does not by itself explain how the computing capacity behind AI services is divided, accessed or secured. That distinction between a physical site and a hosted service is where cloud security becomes useful: it examines how responsibility for infrastructure and data is shared between a provider and its customers, and why a stated capability is not the same as a tested one. If that wider picture is what you want to understand, The Hack Academy’s online cybersecurity training programme lists cloud security among its subject areas, with self-paced learning and practical exercises alongside the theory.