Anthony Albanese seated on a stage with an orange wrapping on his shoulders in Fiji

Prime Minister Anthony Albanese at the opening of the Vuvale Skills hub in Suva, Fiji, 6 July 2026 (Leon Lord/AFP via Getty Images)

Australia’s chance to harness data centres for Pacific development

Australia’s AI infrastructure build is a foreign policy question as much as a domestic one – and what the Pacific gets from it isn’t settled.

Geordie Fung

14 July 20263 min read

The Australian Financial Review has reported (Opens in new window) that Anthropic – the tech giant behind the popular Claude AI chatbot – wants to secure 1.4 gigawatts of Australian data centre capacity at a cost of up to US$15 billion. Prime Minister Anthony Albanese is expected this week (Opens in new window) to set out (Opens in new window) his AI vision.

Both data centre capacity and his AI vision are foreign policy questions as much as domestic ones – defining Australia’s Pacific relationships and its position in AI governance for a decade.

Pacific Island nations are not in a position to compete for AI infrastructure investment directly. The digital divide, already wide, risks deepening (Opens in new window) as this infrastructure concentrates in wealthier countries. There is both a developmental case and a security case for Australia using whatever position it builds to negotiate Pacific access to AI capabilities, support regional capacity building, and ensure Pacific voices are present in governance forums.

Australia's track record of translating geostrategic interest in the Pacific into outcomes that actually serve Pacific islanders is mixed. And the leverage argument has Pacific implications Australia has not yet reckoned with honestly.

Data centres at this scale pressure electricity grids still transitioning to renewables, could push up energy prices, require substantial water for cooling, and raise legitimate concerns for affected communities. Lowy Institute polling this year (Opens in new window) found that 64% of Australians believe AI's risks outweigh its benefits, up 12 points since 2024. The Climate Council has sounded warnings (Opens in new window) about what unchecked data centre demand means for the energy transition.

There is a tension here that cannot be papered over. Pacific Island countries carry the weight of climate change more acutely than almost anyone. They have legitimate grounds for scrutiny (Opens in new window) of any proposal that increases energy demand in exchange for access to commercial technology built elsewhere. Australia will need to engage with that tension honestly rather than dress it up in development language.

The last decade of strategic competition with China in the Pacific was fought on infrastructure financing, with the Australian Infrastructure Financing Facility for the Pacific (AIFFP) one of its main instruments. AI is the next arena. The good news for bureaucratic continuity is that rebranding the AIFFP as the Artificial Intelligence Financing Facility for the Pacific requires only a minor typographic adjustment. The strategic logic, and its limitations, will be harder to update.

For middle powers, this is the central strategic problem of the AI era. Decision-making power over how AI develops sits with a handful of frontier labs and, to varying degrees, their respective governments. A way countries acquire meaningful leverage is by controlling a piece of the supply chain. Taiwan's semiconductor position is the clearest, if most extreme, example: it has made a small island democracy geopolitically indispensable in ways nothing else could.

Australia hosting a large share of global compute would not replicate that position, but it creates real options: a stronger basis for negotiating frontier model access, a more credible seat at AI governance tables, and a position in AI safety conversations that carries some weight.

This is where the Pacific can benefit.

Alliance membership is no substitute for leverage built on something tangible.

The government’s own expectations framework for data centre developers (Opens in new window), released in March, already requires large-scale compute providers to make capacity available to Australian researchers, start-ups and not-for-profits on favourable terms. That principle points toward something more significant: a government with a credible claim on Australia's compute capacity can direct some of it toward national priorities, including cybersecurity capabilities that would otherwise depend entirely on foreign systems and foreign goodwill.

That last point was made concrete earlier this year. As Tom Barber wrote (Opens in new window), when the Trump administration temporarily suspended global access to Anthropic's most advanced models, neither the alliance nor Five Eyes membership made any difference. Alliance membership is no substitute for leverage built on something tangible.

So Albanese’s speech needs to get three things right.

The first is public consent: a confident, specific case for why Australian leverage with the world's leading AI companies serves ordinary Australians, paired with credible environmental commitments rather than aspirational ones.

The second is the deal: any agreement with Anthropic must be structured to preserve real government power rather than trade it away at signing for the promise of investment and jobs.

The third is the Pacific: if this build-out gives Australia something real to offer, it should say so explicitly, and early.

About the author

Geordie Fung

Geordie Fung is the Director of Analysis & Advising at the Development Intelligence Lab, a Canberra-based think tank focused on international development policy and Australia’s role in the world.

Topics

Artificial intelligence Pacific Islands Australia Aid & development Cyber & technology