Detrended

Subscribe.

No spam. Unsubscribe anytime.

The anti-data centre hysteria.

Justin Pyvis26 August 2026Artificial Intelligence · Politics

Data centres have become the new target for industrial-scale panic. In the past year alone, support across the US political spectrum has collapsed from slightly in favour to deeply opposed.

The anti-data centre angst has even reached the point where a majority of Americans would rather a coal power plant be built near them than a data centre. That's despite coal plants using more water and generating considerable pollution, literally taking years off of people's lives. Even more curiously, multiple surveys have found that people would rather live next to a nuclear power plant than a data centre to power AI.

That's unusual because while nuclear power is clean and safe, for the past several decades it has been a classic example of emotion overriding basic arithmetic: enough people caught the fear and no amount of facts will change their minds. In Australia nuclear power is still illegal, for goodness sake, and only two years ago a sitting member with an economics PhD was posting memes of a three-eyed koala!

Meme showing Blinky Bill with three eyes.

So, why do people hate data centres so much? I think there are two main reasons. The first is straight up misplaced blame: another recent poll found that 71% of locals opposed building data centres near them, with the top two reasons being energy and water usage.

Clearly they assume data centres will drain local reservoirs and hike power bills, even if that's often not the case.

More to the point, if power and water bills rise when data centres are built, that's a failure of rate design and unresponsive, monopolistic supply, not data centres. Most countries haemorrhage water each year through subsidies and poor pricing policies. In Colorado, alfalfa alone uses more water than all household and industrial uses combined, and produces little economic benefit (likely negative because farmers don't pay market rates for water). In California, almond farms use fifty times as much water as data centres, yet a decade of drought-shaming never turned into a widespread anti-almond movement.

Water and energy are just convenient scapegoats; I think the real driver of anti-data centre hysteria is the AI movement itself. For years now, wannabe trillionaires like Sam Altman (OpenAI) and Dario Amodei (Anthropic) have been telling anyone who'll listen that AI is going to put everyone out of work, and maybe even kill us all.

Is it then any wonder that people would oppose building the very infrastructure that they were told would destroy their world? It doesn't matter that Altman and Amodei are full of shit, and are peddling such nonsense because they want the government to regulate them a moat in the name of "safety". What matters for public opinion is they've been repeating it ad nauseam, and the masses have been listening.

So far, this hysteria has mostly been isolated to the US. China certainly isn't going to stop building data centres (Chinese people tend to be optimistic, not fearful, of AI), and demand for AI or short-form video clearly isn't going anywhere. But to the extent that anti-data centre NIMBYs do succeed in the US (there's no evidence yet that they are on a scale that matters), there could be a decent opportunity for countries like Australia to pick up the slack.

The good news is the Albanese government and the Coalition appear to be in favour of data centres, suggesting the issue hasn't been politicised – yet. The bad news is the Greens are trying to go thermonuclear on data centres, and there's a chance that One Nation follows.

But really, Australian support for data centres is fragile because even the proponents don't really understand them. On one side of politics, you have government ministers obsessing about leveraging data centres to cross-subsidise other projects, like renewables, whose true costs they want to hide from the public balance sheet. That's much less efficient than using taxes and transfers, and is a surefire method of turning what could be a net benefit into a net cost.

Then on the other side, you have Coalition senators warning that if Australia doesn't build data centres, then "Chinese models will be here in Australia hoovering up our data and people will be using Chinese models, which we have no control over".

None of that is true. First, "we" have no control over American models, either. Second, having data centres physically located in Australia gives you precisely zero control over which models Australians actually use: people can query a Chinese model whether the server sits in Sydney or Shanghai. Third, governments can compel data access, but if it's encrypted (as it will be), it won't help them all that much. Really all it gives you is the ability to turn them off (decent geopolitical leverage) or tax them, which is where the real benefits lie.

In the next year or two it's going to become clear which countries have decided to turn data centres into this generation's nuclear power, and which will tolerate or even revere them for the benefits they can bring. Each choice will tell you whether that country is content with stasis or is serious about progress, because if you can't build data centres, it's probably going to be difficult to build a lot of other things, too.


ShareXLinkedIn

About the author

Justin Pyvis

Independent economist and casual techie based in Perth, Western Australia.More →

Subscribe to the newsletter.

Get new essays delivered to your inbox. No spam. Unsubscribe anytime.