The everyday costs of AI infrastructure revealed by votes in 10 US states
In the November 2026 US midterm elections, voters directly restricted data centers, which are considered essential infrastructure for artificial intelligence. Of 45 data center measures on ballots in 10 states, 33 restrictions had passed as of the count on the morning of November 4, with another six considered possible to pass. The election results showed that voters placed electricity bills, water use and neighborhood conditions ahead of arguments about technology industry growth.
In the author's assessment, these results represent less an eruption of anti-technology sentiment than the emergence of politics focused on everyday concerns about who should bear the costs of AI infrastructure. The central issue in these votes was not simply opposition to construction.
Even as residents used AI services, they questioned arrangements under which the electricity grids, cooling facilities, transmission infrastructure and tax breaks supporting those services imposed costs on their communities. The measures approved included outright bans, temporary moratoriums, energy-use limits and requirements for public hearings or resident consent before project approval.
This signals that the administrative practice of treating data centers as ordinary industrial facilities, like factories or warehouses, is no longer adequate. The strongest opposition came from Ohio. According to reports published by Engadget, The Associated Press and others on November 4, 2026, Ohio currently has 257 data centers, and all 20 related measures in the election had either passed or were considered likely to pass.
In Urbana and Hubbard, outright bans on data centers passed with support of 76% and 77%, respectively. Support of 76% to 77% is difficult to achieve with the supporters of a single political party alone.
This can be interpreted as indicating that everyday costs and pressure on local resources have become issues that cut across conservative and progressive divisions. Politicians were also affected by this trend.
Democrat Sherrod Brown made opposition to data centers a major campaign pledge in the US Senate election and won after attacking Republican Senator Jon Husted over links to the data center industry.
According to figures cited in the same reporting from an NBC exit poll, 56% of voters said opposition to AI and data centers had also influenced their other electoral choices. The 56% figure indicates that AI infrastructure issues have moved beyond local government permitting disputes to become a consideration in national elections. When industrial policy affects residents' living costs, data centers become an issue not just of buildings housing servers, but of electricity bills and water supplies.
The reasons for this response are relatively clear. Data centers draw large, steady supplies of electricity and use cooling equipment to remove heat from servers. For local residents, the costs of grid expansion, concerns about water shortages, noise and traffic changes, and land-use disputes are more immediately apparent than the jobs companies promise.
The source material identified high electricity bills, water shortages and concerns about community impacts as reasons for public opposition. Although the available figures do not show the extent of rate increases in every area, voters used the ballot to express concerns that data center expansion could strain essential infrastructure.
Concerns about electricity bills and water shortages become new factors in technology policy
This exposes a longstanding weakness in the AI industry. Users experience search, translation, image generation and work automation as convenient services on a screen, but the computation behind them depends on electricity, water and land somewhere.
Technology companies have described cloud computing and AI models as abstract services, but the actual infrastructure occupies substations, cooling towers and transmission lines in specific counties and cities. The costs remain physically concentrated in local communities, while the benefits spread nationally or globally.
These US votes brought that imbalance into the electoral process. There are counterarguments, too.
One is that blocking data centers could slow AI development and cause communities to lose investment, tax revenue and some employment opportunities.
Given that the Trump administration is promoting AI development as a major growth strategy, successive restrictions by local governments could also weigh on national competitiveness. This counterargument has some validity from an industrial strategy perspective. To be persuasive, however, it must not conceal the costs residents would bear, and must provide quantified plans for electricity procurement, water use, electricity supply priorities during emergencies, and waste heat and noise management.
The approach of securing approval first and explaining later has already failed in this election. As experts cited in the source material forecast, construction is unlikely to stop completely.
These votes did not provide evidence that AI demand has weakened. Rather, the data center industry now faces pressure to adopt on-site power, renewable energy power purchase agreements, cooling methods that reduce water use and strategies that distribute facilities across multiple sites.
If US local governments make resident consent a prerequisite for permits, companies will have to move beyond siting strategies focused solely on grid connection availability and account for community acceptance as part of project costs. AI service prices that do not reflect those costs are unlikely to be sustainable. This issue is relevant to readers in South Korea as well.
In South Korea, increased use of AI services will inevitably increase demand for server capacity and electricity. The questions are where to build data centers, which power sources to use and whose consent to obtain. The combination of concentration in the capital region, available grid capacity, cooling water supplies and procedures for informing residents could lead to conflicts similar to those in the United States.
According to this publication's analysis, South Korea's policy priority is not competition to attract data centers, but establishing rules in advance for allocating the costs when data centers enter residential communities.
Resident consent and energy planning are also central to South Korean data center policy
The author argues that the course for the central and local governments is clear. First, the permitting process for large data centers should disclose projected electricity consumption, water-use methods and emergency power measures in language residents can understand.
Second, developers should be required to explain in advance how they will assume responsibility for local grid expansion costs and cooling water demands. Third, public hearings should not be treated as formalities, but designed as meaningful mechanisms through which residents can change project conditions.
The 45 votes across 10 US states showed that AI infrastructure policies lacking these three conditions encounter political resistance. Companies also face responsibilities.
Data center developers can no longer easily choose locations based solely on rapid permitting and low electricity prices. Communities are asking not whether advanced technology is necessary, but how it will change household electricity bills and water supplies. Without answers, AI companies may lead in model performance but still face infrastructure constraints at their sites.
Data center construction can generate less political conflict when technology companies treat residents as stakeholders sharing the burdens, rather than merely as people to persuade. This election was not a declaration that progress in the AI era should slow. It was a demand to change the criteria used to judge that progress.
Expanding server capacity and improving models alone cannot create a sustainable technology ecosystem. Electricity grids, water, land and resident consent must be considered together within the same plan. If South Korea dismisses the US results as someone else's conflict, it is likely to experience similar disputes later.
Accepting AI as infrastructure in everyday life first requires answers about who receives the benefits and who bears the costs. This article was prepared with reference to reporting published by Engadget, Tom's Hardware and The Associated Press on November 4, 2026.
Individual article titles and reporter names could not be identified in the source material.
FAQ
Q. Do US votes opposing data centers directly affect electricity bills in South Korea?
A. These votes do not immediately change electricity bills in South Korea. However, the same relationship between the spread of AI services and growing data center electricity demand also applies in South Korea. If large data centers become concentrated in particular areas, discussion of transmission grid reinforcement, power supply reliability and electricity tariff structures could also intensify. Ordinary users should recognize that even if AI service prices appear low, the underlying electricity costs may be transferred to society as a whole. The US case is the first example of this structural transfer of costs becoming an election issue.
Q. Would blocking all data centers weaken the AI industry's competitiveness?
A. Banning all data center construction could put pressure on AI research and cloud service operations. However, the message of these US votes is closer to a demand for procedures that explain local costs and assign responsibility for them than to an unconditional ban. Projects with arrangements for electricity procurement, cooling methods, resident consent and benefits for local communities are likely to continue. Experts also expected a shift toward on-site power, renewable energy contracts and water-saving cooling methods rather than a complete halt to construction. Competitiveness comes not only from rapid construction but also from designing infrastructure that can operate for a long time without conflict.
Q. What should residents examine when reviewing a data center plan?
A. Residents should first examine the developer's projected electricity consumption, water-use methods, cooling equipment and emergency generation plans. If a public hearing is held, they should seek answers about grid expansion costs, noise and traffic changes, and water-use priorities during droughts, rather than focusing on job numbers. If official documents lack figures or fail to identify who is responsible, residents should follow procedures to request that the local government disclose supplementary information. Because data centers remain in communities for many years, conditions set during the initial permitting process determine their future effects on daily life. As in the US cases, once construction begins without clear conditions, the options residents can exercise later diminish sharply.
Original Korean article: youthassembly.kr
English brief by Korea Wind Intelligence, koreawindintel.com
