Paris Marx’s recently released volume, *Hyperscale: The Ambition and Excess of Big Tech’s Data Empires*, arrives as data-center sit-ins become a decisive issue in upcoming U.S. midterm contests. Marx traces a lineage of opposition that stretches from anti-war demonstrations targeting American computers during the Vietnam era to contemporary protests in Ireland, Chile and New Zealand. He argues that the current American flare-up is part of a longer, worldwide pattern of community pushback against massive computing farms.
In Europe, the Netherlands and Singapore imposed data-center moratoriums as early as 2019, signaling early regulatory caution. By 2022, Irish officials reported that new facilities were straining the national electricity network, prompting developers to seek alternatives such as gas-fired generators and dedicated power stations. The Irish public utility, which alone can construct transmission lines, faced lobbying pressure to permit private firms to lay their own cables, raising equity concerns across the grid.
Marx cites a prolonged dispute in The Dalles, Oregon, where Google allegedly concealed its water consumption. After a year-long effort by the city council, officials disclosed that the company was drawing roughly 29 percent of the municipality’s total water supply. The revelation sparked local outrage and highlighted how data-center operators can obscure resource use, a tactic that Marx warns may reappear as AI-driven facilities proliferate nationwide.
In response to grid constraints, Irish data-center firms lobbied for legislation that would allow private entities to install their own transmission lines, a move the Irish Environmental Protection Agency warned could jeopardize climate targets. Companies also secured emission permits to burn fossil fuels, and began constructing gas-powered plants dedicated to server farms. Marx notes that U.S. developers are now mirroring this approach, racing to erect private gas facilities to meet anticipated AI workloads.
The current hype echoes the dot-com era, when excess fiber and data-center capacity led to a wave of bankruptcies before survivors like Equinix absorbed assets and waited for demand to rebound. Marx warns that AI’s rapid expansion may follow a similar boom-bust cycle, especially if the industry assumes server hardware will be refreshed every few years. He cautions that even if AI demand wanes, the accompanying gas plants will remain, seeking new customers to justify their existence.
A final chapter set in New Zealand showcases a community-driven effort to train a Māori language speech model using locally hosted cloud services, offering a non-extractive alternative to the dominant AI paradigm. Marx calls for governments to pair stricter data-center regulations with oversight of AI products, arguing that promises made by industry leaders in early 2023 have not materialized and that the social and environmental costs now outweigh any financial gains.