Why the “generational” data centre opportunity feels a lot like the coal and gas debates of old

Australians have spent decades fighting over how much multi-national investors should pay for our resources. Now, it seems the debate has shifted from the mines to the machines.

While Anthropic just opened its first Sydney office, Google is currently negotiating a proposed $20 billion data-centre expansion. Just as Jim Chalmers prepares his toughest budget yet, it looks like one of the world’s most profitable multi-nationals of all time is now playing hardball on whether or not it’ll have to pay taxes.

While the investors are technology companies instead of mining giants, the dilemma should feel familiar for anyone playing along at home. These new investors are not aiming to dig up Australia’s resources so much as seeking to tap into them; our “land abounds”, our abundant sun, our political stability, and, in some places, our water.

Should Australia insist that multinational companies pay their fair share on Australia’s revenue opportunity? Or should the government tread carefully, worried the investment might fall instead to Southeast Asia.

Politically, this is an incredibly hot potato for Labor. Pushing back on tens of billions of dollars in investment during an oil and cost-of-living crisis would be like lobbing a full toss to the newly hyped-up opposition, and just waiting for Angus Taylor and Matt Canavan to hit them for six.

Yet rushing into the next mega-infrastructure wave raises deeper questions. What exactly might Australia be signing up for?

Artificial intelligence may feel intangible, but the infrastructure behind it is anything but. While traditional data-centres have demanded electricity similar to that of a few thousand households, hyperscaling data centres, like the ones planned, could need about as much power as pumped out by a typical coal-fired power plant.

When you consider that some of these hyperscale data-centres drink up to 40 million litres per day, you could almost consider a mega data centre the environmental equivalent of a big coal mine. 

The machines behind the large language models also require enormous computing halls, high-capacity grid connections, and vast amounts of electricity. Many also rely on significant volumes of fresh water for cooling.

Co-locating renewable energy, grid capacity and water access is rarely simple. After more than a generation of over-allocations, ecological decline, and political conflict across the Murray-Darling basin, we know just how badly these trade-offs can go wrong.

Then we add in the unfortunate sober reality of Australia’s climate targets. While Australia’s solar rooftops and household batteries have helped to shift the needle so far, total pollution is not declining fast enough to meet even our 2030 targets. Adding tens of gigawatts of data-centre demand, even if they are backed up with new renewable capacity, would likely only make this even harder to balance going forward. 

Then there is the critical question of jobs, and really, the structure of the Australian economy. 

Hyperscale data centres may require new construction workers to begin with, but once the servers are plugged in, the lights go off, and the workers go home.

In China last year, I saw what this “dark factory” future looks like up close, with vast automated facilities generating enormous value, but with very few long-term jobs on offer. With the exception of the odd engineer and a few security guards working 24 hour shifts, there’s almost no-one to head to the local pub after work.

Contrast that with the mining sector, which employs thousands of workers onsite for decades. Regardless of what the mine’s owners might get away with, those workers then pay tax, buy homes, and indirectly drive local economies.

Just as we begin to talk about a Just Transition for our fossil fuel sector, how should we begin to think about the dark-factory futures of data centres? And what else might it be fuelling?

Last week, five hundred Australian employees at Atlassian reportedly lost their jobs amid an AI-driven restructure. Is this a one-off blip? Or would new investments in data-centres across Australia only accelerate this trend across the rest of the economy?

Yes, turning away billions of dollars of new investment ahead of a tight budget would be incredibly hard to justify. But diving in without a clear understanding of the trickle down effects risks more than a budget deficit.

Australia has spent decades learning how hard it is to reclaim control over its resources once a boom begins. The real question is whether it is prepared to ask harder questions before the next one arrives. Failing to do so is not just an economic risk. It is generational.