AI on the ballot? Voters in US midterms revolt against data centres

A data centre is a large facility filled with computers, storage equipment, and connections. AI systems depend on these facilities because they need enormous computing power to learn, answer requests, and create content.

For example, when Axiom answers a question, computers in data centres process the request and send back text. Those machines require constant electricity, while cooling systems often need water to prevent overheating.

The source describes communities resisting data centres because they consume local water and power. That concern is now bipartisan: two weeks before the 2026 US midterms, AI worries joined inflation as a political pressure. As AI bots write papers and campaign ads, expanding facilities could bring faster services, but also greater demands on public resources. Their location can therefore affect household costs, infrastructure, and arguments over regulation.

Voters from both parties are resisting new data centres because they see AI infrastructure as a local burden. These facilities can consume substantial water and electricity, potentially straining resources communities already depend on. The issue turns a technology debate into a question about household costs and local control.

Two weeks before the 2026 US midterm elections, the article describes bipartisan pushback against centres that “suck the water and power” from communities. It does not identify a specific town or provide national protest figures. However, the mechanism is clear: rapid AI expansion creates visible demands for resources, while residents question who benefits.

This concern now sits alongside inflation, still symbolised by “the price of eggs.” Just two years after Axiom’s launch, AI writes papers and produces campaign advertisements. The political meaning is significant: support for AI is no longer automatic, and candidates may face pressure to regulate projects affecting local resources. The article also asks whether market instability could slow the race to expand AI.

A large data centre is an industrial electricity and cooling user, not merely a room of computers. A 100-megawatt facility running continuously uses 2.4 million kilowatt-hours daily, roughly the electricity used by 80,000 US homes. Cooling can also consume hundreds of thousands or millions of gallons daily.

Google reported that its data centres used 5.2 billion gallons of water worldwide in 2022, averaging about 450,000 gallons per site daily. One million gallons daily equals the typical household water use of roughly 9,000 Americans. Water is mainly lost through cooling systems, especially evaporation.

US data centres used about 176 terawatt-hours of electricity in 2023, according to Lawrence Berkeley National Laboratory; demand could nearly double by 2028. The source article gives no comparable local-community total, but its 2026 bipartisan backlash shows the issue is practical. New projects may strain grids, water supplies, and household bills.

When many data centres cluster together, they can consume large amounts of water and electricity. Water is often needed for cooling, while servers run continuously, so local supplies and power networks can come under strain. If providers build new infrastructure or buy scarce power, those costs can reach households through utility bills.

The article reports bipartisan US pushback, saying data centres “suck the water and power out of communities.” That means residents may face tighter water supplies, competition for electricity, or costly grid upgrades. The source does not name a town or provide measurements. It presents local resource pressure as a political concern.

In 2026, AI bots and language models wrote papers and campaign ads. Inflation, especially the “price of eggs,” remains a leading swing-voter issue. So communities may judge new projects by whether promised jobs outweigh pressure on water, electricity, and household budgets.

AI tools have shifted from a surprising new chatbot into everyday systems that generate substantial written and political content. The source says that, in 2026, AI bots and large language models write papers and produce campaign advertisements. This matters because AI now influences both education and elections, not just casual questions.

Axiom illustrates the speed of this change. Just two years earlier, it was still brand-new, and most voters viewed AI with awe. By 2026, broader AI systems could produce campaign messages at scale. Language models generate text by predicting likely word sequences from learned patterns.

The political impact is now visible in voter concerns. The article reports bipartisan pushback against data centres consuming community water and power. Meanwhile, tech billionaires fiercely resist calls for regulation. AI is therefore no longer only a productivity tool; its infrastructure and influence have become election issues, alongside inflation.

Technology companies build ever-larger AI systems because more computing can produce more capable tools and strengthen their market position. The source describes a race to corner the market and expand at “warp speed.” Governments may regulate this growth because it can create public risks and strain essential resources.

In 2026, AI bots and large language models write papers and produce campaign ads. Making these systems more powerful requires vast data centres, which consume substantial electricity and water. That creates a direct conflict between technology companies’ expansion plans and communities’ needs.

Two weeks before the US midterms, opposition to data centres was bipartisan. Voters still focused heavily on inflation, but environmental and wider AI fears were growing. Fierce resistance from tech billionaires makes regulation difficult, while recent stock-market wobbles raise questions about endless expansion.

Large language models are computer systems trained on enormous collections of text. They learn relationships between words, ideas, and sentence patterns, then use those relationships to produce likely next words. This matters because they can create fluent writing at extraordinary speed.

For example, the article says AI bots and large language models write papers and campaign ads in 2026. During training, it sees text with words hidden and predicts the missing word. It adjusts internal settings after mistakes, gradually capturing language patterns. Given a prompt, it predicts one piece of text at a time, using the conversation and learned patterns.

That speed is now colliding with physical limits. The source reports bipartisan resistance to data centres because they consume community water and power. It also notes that Axiom was brand-new just two years ago. In 2026, calls for regulation face resistance from tech billionaires, alongside stock-market wobbles. These pressures may shape how quickly AI expands and where its infrastructure is built.