Paris Marx on what US data center critics can learn from abroad

Paris Marx on what US data center critics can learn from abroad

Wired's Molly Taft interviewed Paris Marx, a longtime critic of Big Tech and host of the Tech Won't Save Us podcast, about his new book Hyperscale: The Ambition and Excess of Big Tech's Data Empires. The book lands as data centers have become one of the issues that could decide key US midterm races in less than two weeks. Marx's book follows the industry's quest for power and the resistance to it, from antiwar protests aimed at American computers during the Vietnam War to recent resistance in Ireland and Chile, and an alternative vision for data centers in New Zealand. The interview was edited and condensed.

Marx says the story began for him well before the current US backlash. The Netherlands and Singapore had data center moratoriums going back to 2019. By 2022 there was growing recognition of how much electricity data centers were taking from Ireland's grid. He also points to The Dalles, Oregon, where he says Google got the city council to try to keep its water use figures secret. That went on for the better part of a year before the community got the numbers and found Google was using 29 percent of the water in the community. ChatGPT arrived at the end of 2022, and Marx says it became clear generative AI would need many more data centers, bringing these problems to far more communities.

Ireland, in his telling, is the early warning. Dublin passed a moratorium in 2021, and it was the grid operator that said there were too many data centers and no grid capacity to connect more. The industry then looked at other parts of Ireland, at tapping the gas network to run generators, and eventually at building dedicated gas power plants instead of connecting to the grid. Ireland's grid is public and only the public utility can build electricity lines, but Marx says data center companies lobbied for legislation that essentially lets private companies build electricity cables, which raises questions about equity. He adds that operators got emission licenses from the Irish Environmental Protection Agency to burn fossil fuels, while the agency warned that climate targets would be a lot harder to achieve. Seeing US companies now race to build massive private gas plants for data centers, he says he wasn't surprised, though the scale and speed are staggering.

Taft asks what happens to the gas plants and data centers if the industry collapses. Marx says he thinks the industry is in an AI bubble. Planned data centers and gas plants face shortages of turbines, chips and other components, and it is an open question whether they will even be built before demand for AI is reassessed. In the dotcom bust a lot of data center companies went under, and Equinix, which survived, bought up many data centers to be ready for when demand returned. This time he sees a problem: if the centers are built for AI and the chips must be cycled out within a few years, what makes a facility valuable may no longer be useful when demand comes back. And even if the data centers don't come, the gas plants will exist and will look for new customers to justify themselves. He is sure nobody is building them intending to shut them down.

On what to do, Marx would like governments to put much firmer regulations on data center development, paired with regulations on the AI products themselves. He says generative AI has not delivered on many promises and has brought a long list of harms nobody warned about, and that governments are buying the narrative Sam Altman sold in early 2023. He says society should ask how much computation it really needs and whether what Silicon Valley builds justifies the resources being sacrificed. In his view an honest assessment would find no way to justify current resource use from a financial, social or environmental perspective. Taft notes the book ends on a community-led effort in New Zealand to build a more accurate speech-recognition model of the Maori language, with local cloud providers hosting the data, which she calls a non-extractive alternative.

The piece closes with a short newsletter section. Wired colleague Joel Khalil covers how the EU's appetite for AI is colliding with European climate goals, and tells Taft he wouldn't be surprised if data centers become a political flash point in Europe as countries go to the polls next year. Links to other outlets follow: The New York Times on the AI boom boosting gray paint as a commodity; The Texas Tribune on the Texas build-out threatening years of progress on cutting air pollution; and The Guardian on Google apologizing for clearing more than 800 acres of forest in Finland without the necessary environmental permits, for part of a €13 million data center project, with backlash that could bleed into elections next spring.

Key facts

  • Paris Marx, author of Hyperscale, says opposition to data centers has a long record abroad: moratoriums in the Netherlands and Singapore from 2019 and in Dublin in 2021.
  • Marx says Ireland's grid operator ran out of capacity, after which the industry turned to gas generators and dedicated gas power plants, a pattern he sees repeating in the US.
  • In The Dalles, Oregon, Marx says Google got the city council to try to keep its water use secret; once the numbers came out, Google was using 29 percent of the community's water.
  • Marx thinks the industry is in an AI bubble and expects stranded gas plants to hunt for new customers if data centers don't materialise; he wants firmer regulation of both data centers and AI products.
  • Wired's newsletter section adds that Google apologized for clearing more than 800 acres of Finnish forest without the necessary permits, and that data centers may become a political issue in Europe next year.

Why it matters

Data centers are one of the issues that could decide key US midterm races in less than two weeks, and many Americans see the fight as a sudden arrival. Marx's argument is that other places lived through it first: Ireland, the Netherlands, Singapore and others offered early warnings. The Irish sequence he describes, from grid limits to gas generators to dedicated private gas plants and private cable building, is the template he says the US is now following at a larger scale.

Who it affects

Communities near planned data centers, which in his account may struggle to get basic facts such as water use. Also governments and regulators weighing moratoriums or firmer rules, grid operators and utilities, and AI and cloud companies that need capacity. Wired's newsletter section points to Europe too, where Joel Khalil expects data centers to become a political issue as countries vote next year, and to Finland, where Google faces backlash over forest clearing.

How to use it

This is an interview to read for framing, not a tool. For anyone following or contesting local projects, the practical points Marx raises are: look at what happened in Ireland and The Dalles, ask for resource figures such as water use, and watch for dedicated gas plants and private grid infrastructure. His policy ask is firmer data center regulation paired with regulation of AI products, plus a public assessment of how much computation is really justified.

How solid is it

It is a Wired interview, edited and condensed, with a critic of Big Tech who is promoting his own book. The factual claims about Google in The Dalles, Irish gas plants, lobbying for private cable rights and emission licenses are Marx's account, not independently verified in the piece. The AI bubble and the claim that current resource use can't be justified are his opinions. The Finland, Texas and gray paint items are Wired's summaries of reporting by The Guardian, The Texas Tribune and The New York Times.

Risks and caveats

Marx's bubble view and his prediction that gas plants will outlive the data centers are forecasts, not facts. The interview is condensed, so wording may differ from the original conversation. The piece does not name which US midterm races data centers could decide, and details of the Chile and Vietnam War episodes are only mentioned in the intro. Marx's call for regulation reflects his own position, and the interviewer's description of the New Zealand project is favourable.

“I certainly didn't predict that it was all going to kind of come to a head as my book was coming out”

— Paris Marx, author of Hyperscale, in Wired