Senator Jon Husted
Chips, models and the technological race with China have long been seen as the biggest risks to America’s AI push. Another one is now emerging: voters themselves. In an internal memo, the Senate Republicans’ campaign arm warns leading AI companies that growing opposition to data centers could cost the party a crucial Senate seat in Ohio. What may look like a local campaign issue touches one of the fundamental requirements of the entire AI boom. Without new data centers, there can be no ever more powerful models. But their demand for electricity, water and land is becoming increasingly tangible for local communities.
An internal memo sounds the alarm
The National Republican Senatorial Committee, or NRSC, has issued an unusually direct appeal to leading AI companies.
At the center of the concern is the Senate race in Ohio, where Republican incumbent Jon Husted is running against former Democratic Senator Sherrod Brown.
Brown has made data centers a major campaign issue, attacking Husted over his earlier support for such projects and the tax incentives granted to them.
Republican campaign strategists are increasingly concerned that the strategy may be working.
According to the memo published by Axios, if Husted loses and data centers are seen as a reason, politicians across the country could become more reluctant to support similar projects. The issue, the memo argues, has developed into an underestimated factor in the campaign.
The nervousness is understandable.
A Fox News poll from August 13 currently puts Brown ahead of Husted by 53 to 45 percent. A single poll neither proves that data centers are responsible for the gap nor predicts the election result. Republicans are nevertheless concerned that the issue could make an already difficult race even harder.
From tech enthusiasm to an acceptance problem
Ohio is far from an isolated case. A nationally representative survey by the Annenberg Public Policy Center at the University of Pennsylvania shows how quickly public sentiment has shifted.
Sixty-one percent of Americans now oppose the construction of new data centers in their own communities. In the spring, the figure was 49 percent. Opposition has therefore risen by twelve percentage points in just four months.
The political distribution is particularly striking.
Sixty-nine percent of Democratic respondents oppose new data centers near where they live, but so do 54 percent of Republicans and 53 percent of independents.
This is not following America’s usual left-right divide.
Resistance is strongest among younger people: 70 percent of those under 30 oppose new data centers in their area.
For an industry that often presents its technology as a defining project for the next generation, that is a notable signal.
AI suddenly has a physical footprint
Part of the problem may be that artificial intelligence long appeared largely intangible to consumers.
ChatGPT lives in a browser. An AI assistant writes text on a smartphone. An image generator does not need a visible smokestack.
Yet behind every one of these applications is a physical infrastructure of servers, chips, power lines, cooling systems and buildings.
As models become more powerful, that infrastructure is expanding rapidly.
The Lawrence Berkeley National Laboratory estimates in its US Data Center Energy Usage Report, published in June, that data centers could account for around 11.8 percent of total US electricity consumption by 2030 under its central scenario.
Depending on how the market develops, the estimate ranges from 9.5 to 15.3 percent.
That turns the digital AI boom into an energy-policy issue of considerable scale.
And as soon as a data center is planned in a specific community, the perspective changes.
Residents are no longer debating large language models. They are debating new power lines, generation capacity, land use, noise, water consumption and their own electricity bills.
Electricity bills are particularly dangerous politically
The US administration has already recognized the risk.
In March, President Donald Trump announced the so-called Ratepayer Protection Pledge. Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI committed to financing the additional power supply and grid connections required by their new data centers themselves.
The costs are not supposed to be passed on to ordinary electricity customers.
In July, the initiative was expanded to utilities, project developers, cooperatives and states. According to the White House, the participating organizations now account for around 80 percent of the electricity supplied to US households and businesses.
That alone is a political signal.
When a government explicitly asks some of the world’s largest technology companies to guarantee that their data centers will not raise household electricity bills, it shows how sensitive the issue has become.
Whether voluntary commitments can fully prevent such effects over the long term is another question. New power plants and grids still have to be planned, approved and built. Data centers also compete with households, industry and electrification for available capacity.
The promise of jobs is not enough for everyone
Supporters point to the economic benefits. Data centers bring billions in investment to local regions, create construction work, can increase municipal tax revenues and may support new energy and grid infrastructure.
For the United States, they also have strategic importance. Washington increasingly views the expansion of AI infrastructure as part of its economic and geopolitical competition with China.
But the calculation looks different from the perspective of a local community than it does from Washington.
Building a data center may create a significant number of jobs in the short term. Once the facility is completed, however, permanent staffing requirements are relatively modest compared with the size of the investment.
The energy demand remains.
That explains part of the political problem: the benefits of an AI data center may be substantial for the country as a whole or for the company building it, while some of the burdens remain local.
Large infrastructure projects have faced the same asymmetry for decades.
It is now reaching an industry that until recently was often perceived as almost infinitely scalable.
Texas shows how quickly the mood can turn
The shift is particularly visible in Texas.
Governor Greg Abbott was long one of the most enthusiastic supporters of major technology investments and celebrated Google’s planned investment of around $40 billion in the state as recently as late 2025.
He has since become sharply critical of the data-center industry. Abbott said over the weekend that the sector had partly created its own political problems through the way it dealt with local communities.
That puts him at odds with Trump.
The president continues to defend the expansion of data centers and has described communities that reject them as shortsighted. From his perspective, too much is at stake: without additional computing capacity, the United States may struggle to maintain its lead in artificial intelligence.
The dispute shows how far the issue has moved beyond conventional party lines.
It is increasingly becoming a conflict between national growth strategy and local acceptance.
Local opposition can become an investment risk
For the technology industry, this creates a new economic risk.
An AI company may have sufficient capital, access to the necessary chips and the technical ability to build a data center.
If it cannot secure political approval, none of that is enough.
Data centers require large sites, lengthy grid connections and, in some cases, new generation capacity. A project cannot simply be moved elsewhere at short notice.
As political opposition grows, approval processes take longer. If conditions become stricter, costs rise. If projects are cancelled, planned computing capacity disappears.
In the end, that could even constrain the development and use of new AI models.
Growing public opposition is therefore already being described as a potentially serious threat to the pace of America’s AI expansion.
It does not have to come to that.
But the risk has changed.
For a long time, the central question was whether America could secure enough chips and electricity.
Now another one has emerged: will the industry actually be allowed to build the infrastructure it needs where it needs it?
Ohio could become a political test case
That gives the Senate race significance far beyond one state.
If Husted loses, it will not automatically prove that data centers decided the election.
Inflation and the cost of living remain major concerns for Ohio voters. Candidates, party identification and the wider national political environment matter as well.
Political strategists, however, do not operate only on neatly isolated causal evidence.
If campaigning against data centers appears to work, others will copy it.
That is the fear behind the Republican memo.
A lost Senate seat could signal to other candidates that opposition to AI infrastructure can win elections.
And that dynamic has already started.
In Pennsylvania, Nevada, Wisconsin and other states, the treatment of large data centers is becoming a campaign issue. Democratic and Republican politicians alike are demanding tougher conditions or distancing themselves from projects they previously supported.
Individual local protests could therefore develop into a national political movement.
The AI industry may lack an answer to a very simple question
Technology companies frequently talk about productivity, medical research, innovation and strategic competition with China.
For someone facing a new data center near their home, the calculation is different:
What do I get out of it? What happens to my electricity bill? How much water will it use? How many permanent jobs will it actually create? And why should a multibillion-dollar company receive tax incentives?
As long as the industry answers these questions primarily by stressing the importance of artificial intelligence to America’s future, the two sides may simply be talking past each other.
Opposition to data centers is not necessarily opposition to technology.
The Annenberg survey provides an interesting indication of this distinction: while resistance to new data centers rose sharply within a few months, Americans’ broader attitudes toward artificial intelligence remained largely unchanged.
Americans have not suddenly changed their minds about AI as a whole.
They are increasingly questioning how much they should have to pay for its physical infrastructure in their own communities.
AI’s next bottleneck could be democratic
America has capital. It has some of the world’s leading technology companies. It is building new semiconductor factories and data centers, and it is trying to secure additional power for the next generation of artificial intelligence.
None of that can expand indefinitely against public opposition.
There is an irony in the current AI race.
A technology whose development is frequently described as exponential is running into an extremely analog constraint: people have to accept the buildings required to power it in their own communities.
Opposition in Ohio does not have to stop America’s AI boom.
But it is reminding the technology industry of a limit that cannot be removed with more computing power or a better model.
Growth needs more than capital and electricity.
It needs public acceptance.
SK
