Australia Doesn’t Want To Be Ai’s Landlord

CANBERRA — Australia is trying to pull off a feat that's eluding national and local governments around the world: Attract AI investment, but on its own terms.

AI suitors including Microsoft, Google and Anthropic now confront a transactional Australian approach. The federal and state governments are welcoming data center investment, but only if developers run them on clean power, account for water use, create local training benefits and give Australian researchers and startups access to computing capacity.

After watching a nasty data center debate unfold in the United States — now home to 4,000 data centers and a vicious backlash against them — Australian officials are determined to forge their own way.

Buoyed by the nation's leading global role in instituting teen social media restrictions, Prime Minister Anthony Albanese — who admits he doesn’t use AI chatbots — has newfound confidence in his tech policy making skills.

He sees the country in a race — not against China or the United States — but against time. “We can set the terms, we can determine AI’s social license. But we have to do it now,” Albanese said in July, setting a goal of finalizing AI legislation in early 2027.

The government’s overarching goals: protect household electricity prices, and integrate Australia into the AI supply chain.

The unspoken part of the deal: A stronger social license for AI should give the government more room to deal with the labor market disruptions that AI use accelerates.

As the world's 12th-largest economy, Australia's 28-million tech-loving citizens and its vast open spaces offer a juicy target for tech companies. They need to train their models and grow their markets amid uncertainty in the U.S. and rigid rules in Europe.

But according to new polling data shared exclusively with POLITICO, two-thirds want Australia to take a cautious approach with AI, even if that means missing some opportunities. The survey of 1230 Australian adults was conducted by JWS Research between Aug. 6-10.

After nearly four years of debate about generative AI, which often hinged on abstract arguments about productivity, or was driven by doomsday predictions by CEOs such as Sam Altman and Elon Musk, Australian voters also sensed a practical risk: higher bills.

Nicolette Boele, an independent MP representing a wealthy part of Sydney, told the House of Representatives she received over 750 responses when surveying local views on AI.

“Over 80 percent were not confident the government will effectively regulate data centers and three-fifths were worried about a repeat of the gas industry experience and being ripped off," she said. The biggest concerns: water and energy impacts.

State leaders are alert to the political danger. South Australia’s Labor premier, Peter Malinauskas, announced on Aug. 10 a royal commission into the impacts of AI.

Andrew Forrest’s Fortescue Mining is building a 2.3GW off-grid green grid around its mining operations in Western Australia, describing it as one of the largest green off-grid networks in the world.

Forrest’s argument is blunt: data centers should stand on their own, and eliminate risks to other consumers by using new green energy.

Andrew Charlton, the Assistant Minister for the Digital Economy tasked by Albanese with fleshing out a national AI strategy, argues that data centers alone will not lift Australian productivity.

If Australia supplies the land, power and buildings while foreign firms retain ownership of the intellectual property, models and customer relationships, much of the value will flow offshore.

As Charlton put it in a speech Aug. 18: “Unless we take timely action, Australia is on course to be a large and permanent importer of intelligence.”

Australia’s decades-long minerals boom carries warnings for AI policy makers.

Australia typically digs up highly-valued critical minerals, but other countries process and build things with them. That lack of innovation and investment leaves the country closer to the bottom of those markets than the top.

Charlton’s lesson from the resources boom is more complicated. “We owned the choke point in the value chain. We owned the ore,” he said.

This delivered Australians good wages, earned income for governments, and was the platform for local mining-services businesses, even as Australia relied on foreign capital and technology.

AI offers no equivalent Australian choke point.

The government intends to do the next best thing: Move the country up the AI value chain.

Charlton argues that access to computing is the key to this, because “without it, the contestable layer of the AI economy is not contestable by us.”

He wants Australia to participate in "80 percent" of the AI supply chain, from data centers through to training models and building AI applications, and described this on a podcast Aug. 17 as “sovereignty across the whole [tech] stack.”

He sees sovereignty as distinct from ownership: the key is having domestic choices to protect national security and guard against economic vulnerability.

Taken together, these policies make data centers an instrument of industrial, security and energy policy, not simply infrastructure deals.

Canberra also wants a national framework to provide consistency and speed for investors, and in July the government established an Office of AI within the Prime Minister’s Office, to smooth the way.

The largest state, New South Wales, has released six principles for data centers and promised speedy approvals for projects that comply.

This is the heart of the Australian approach to AI: trying to move quickly while resisting the idea that speed should settle every question.