What Ireland and New Zealand show about the backlash against AI data centers

Data centers have become a campaign issue in the United States. Wired reports they could help decide key midterm races in less than two weeks. In an interview with Wired's Molly Taft, the writer Paris Marx argues that none of this is new. Marx is the author of Hyperscale, a book on the tech industry's search for power and the opposition it has met. Marx says other countries went through the same conflicts years before American voters did.

The pattern Marx describes is simple. Data centers need large amounts of electricity, water and land. Those costs land on particular towns and grids. The people who live there usually see few of the benefits and often have little say.

Ireland: when the grid says no

Marx points to Ireland as the clearest warning. Dublin got a moratorium on new data centers in 2021. According to Marx, it did not come from activists. It came from the grid operator, which said there were too many data centers and not enough capacity to connect more.

The industry adapted. Marx says companies that could not get grid connections in Dublin looked at other parts of Ireland. They also began drawing on the gas network to run generators, and eventually built dedicated gas power plants to supply their facilities directly rather than through the grid.

Two consequences followed, according to Marx. First, Ireland's grid is public, and only the public utility can build electricity lines. Marx says data center companies lobbied for legislation that would allow private companies to build electricity cables, which raised concerns about equity in the system. Second, operators obtained emission licenses from the Irish Environmental Protection Agency to burn fossil fuels, and the agency warned that this would make climate targets harder to meet.

Marx sees the same sequence in the US now, where companies are racing to build large private gas plants for data centers. "When I started to see that happen in the US, I wasn't surprised," Marx told Wired. "But the scale of it is staggering, and how quickly it's escalated."

Earlier warnings: water, moratoriums and secrecy

Ireland was not alone. Marx notes that the Netherlands and Singapore had moratoriums on data centers going back to 2019.

In the US, Marx cites The Dalles, Oregon. According to Marx, Google tried to keep its water use secret and got the city council to help. The dispute ran for the better part of a year before residents obtained the figures. Google was using 29 percent of the water in the community.

Marx says the launch of ChatGPT at the end of 2022 made it clear that generative AI would need many more data centers. Places already under strain were a preview of what more communities would face. Wired's newsletter lists more recent cases: a report in The Texas Tribune that the build-out in Texas threatens years of progress on air pollution, and a Guardian report that Google apologized for clearing more than 800 acres of forest in Finland without the necessary environmental permits. Wired reporter Joel Khalil expects data centers could become a political flash point in Europe as well, where national policy, local priorities and climate goals pull in different directions.

What happens if demand falls

Marx believes AI is in a bubble and asks what the infrastructure looks like if it bursts. Shortages of turbines, chips and other components may delay projects, Marx says, so some may not be finished before demand for AI is reassessed.

Marx compares this to the dotcom era. Fiber was overbuilt then, and so were data centers. Many data center companies failed. Survivors such as Equinix bought facilities cheaply and waited for demand to return. Marx sees a difference this time. AI chips are expected to be replaced within a few years, so the most valuable part of an AI data center may be obsolete by the time demand returns.

The power plants are another matter. "Even if the data centers don't come, then those gas plants will exist," Marx says. They would look for new customers to justify their existence. In that case a community could keep the emissions and lose the promised investment.

A smaller model from New Zealand

The book ends in New Zealand, with a community-led project to build a more accurate speech-recognition model for the Māori language. The data is hosted with local cloud providers. Taft calls it a non-extractive way to use AI. It is a small case, but it reverses the usual arrangement: the community that provides the data also controls where it is stored and who benefits.

Marx wants governments to regulate data center development more firmly and to regulate AI products alongside it. Marx argues that the promised benefits of generative AI have not arrived and that harms have appeared that nobody warned about. The questions Marx wants asked are about scale: how much computation is needed, and whether the results justify the resources being spent. Marx's own answer is that they do not, financially, socially or environmentally. That is Marx's view, not a settled finding.

What it means for work

Every AI tool that drafts, summarizes or codes for office workers depends on buildings that draw power and water from specific towns. The cases Marx describes show that the costs of that infrastructure fall mainly on grid operators, utilities, environmental regulators and nearby residents. They also fall on local officials, who end up negotiating over water figures, grid connections and emission licenses. The benefits of the work these facilities do go mostly to companies and users elsewhere. Ireland and Finland show that national ambitions for AI can conflict with local planning rules and climate targets. Projects like the one in New Zealand point to a different arrangement: smaller models, hosted locally, under the control of the community they serve. What remains open is whether demand for AI will justify the scale of the current build-out, and who pays for the gas plants and buildings if it does not.