Australia is set to host the infrastructure of the AI boom. Microsoft has pledged A$25 billion by 2029, while Amazon Web Services has committed another A$20 billion. The National AI Plan frames AI as central to a more competitive, productive, and resilient economy, but the government has stepped back from tougher AI rules. This creates a tension: if Australian land, energy, water, workers, research institutions, and communities power the AI boom, then who benefits and who bears the burden?
Senator David Pocock sees a familiar pattern. Australia allowed multinational gas companies to extract immense value from finite resources while paying too little tax and leaving households exposed to high domestic prices. Norway took a different path, using its sovereign fund to ensure oil and gas wealth benefited present and future generations. Pocock warns that AI data centres may reproduce Australia’s mistake: foreign firms use local resources and labour, profits flow offshore, and communities carry the costs.
Pocock is right that Australians deserve a fair return. The Climate Council reports 162 operational data centres and more than 90 proposed projects. If new demand is met through gas rather than renewables and storage, energy prices could rise 26 per cent in NSW and 23 per cent in Victoria by 2035. Energy demand may triple by 2030, while water demand could more than triple over five years, with some connection requests reaching 40 million litres daily. This matters when CSIRO projects that Australia’s cities may need 73 per cent more water by 2050 as climate change reduces supply.
Yet the gas analogy only goes so far. AI infrastructure is not simply another extractive industry. Gas affects revenue, prices, emissions, and energy security. AI also shapes work, education, public services, democratic discourse, knowledge production, emotional life, and national autonomy. Data centres are not merely server warehouses. They are the mines, ports, and railways of the AI age: physical chokepoints through which computational power is produced, owned, and distributed.
Treating the issue only as resource extraction risks repeating the failures of digital platforms. Social media polarised societies not because users lacked manners, but because surveillance, engagement maximisation, and data extraction became profitable infrastructure. Private actors captured the gains, while psychological, social, and democratic costs were dispersed across users, families, schools, and public institutions. AI data centres could deepen this pattern unless Australia sets stronger terms.
Backlash should not be dismissed as technophobia. Gallup found that 71 per cent of Americans oppose local AI data centres, citing water and energy use, environmental harms, quality of life, jobs, and distrust of AI. In Mississippi, residents have sued Elon Musk’s xAI and SpaceX over alleged “omnipresent and inescapable” noise from a power plant serving nearby data centres. At the violent fringe, authorities have reported attempted arson linked to anti-AI extremism.
Violence must be condemned without qualification. But condemnation is not analysis. Backlash grows when infrastructure feels imposed, extractive, and unaccountable. This is the contemporary Luddite lesson. The Luddites were not irrational enemies of technology; they resisted machinery that displaced skilled labour, depressed wages, and concentrated power among factory owners. Their machine-breaking was famously described as “collective bargaining by riot”, before the movement was violently suppressed. Today’s digital Luddism similarly contests platform capitalism, data extraction, and shareholder-governed infrastructure. Resistance to AI data centres is often not anti-progress. It resists a political economy in which communities bear burdens without ownership, voice, benefit, or meaningful participation.
Australia can avoid that path, but not through non-binding expectations, executive photo opportunities, or narrow tax debates. The federal government’s expectations for AI infrastructure developers sit alongside existing laws; they do not create a democratic ownership model.
The challenge is digital sovereignty: retaining legitimate authority over the digital domains that shape collective life. This does not require autarky, protectionism, or technological nationalism. Australia cannot and should not build everything alone. But infrastructure shaping Australian work, education, public services, civic trust, and national security should be aligned with democratic accountability and the common good.
A useful way forward is collective sovereignty: a public-interest approach to AI infrastructure based on shared ownership, citizen control, and needs-based distributive justice.
First, shared ownership. Critical data centres should not be governed entirely by foreign corporations. Australia could pursue public equity stakes, community-benefit agreements, sovereign compute reserves, public-sector access rights, First Nations participation, and data trusts. The aim is meaningful Australian ownership, voice, and benefit. This is no longer a fringe idea. US Senator Bernie Sanders has proposed a 50 per cent public ownership stake in major AI firms, arguing that technologies built on collective knowledge should deliver collective returns. Yet taxation alone is insufficient; Australia also needs bargaining power, including through alliances with middle powers and value-aligned partners.
Second, citizen control. Communities affected by AI infrastructure need enforceable voice before projects are approved, not performative consultation once decisions are settled. This requires transparent reporting on energy, water, emissions, noise, tax, labour conditions, procurement, cybersecurity, and compute allocation, alongside independent audits and genuine veto points where harms outweigh benefits. Infrastructure that increasingly mediates work, learning, health, welfare, and civic discourse cannot be governed through opaque corporate discretion.
Third, needs-based distributive justice. AI infrastructure should serve those with the greatest needs, not merely those able to pay most. Major projects should reserve compute capacity for schools, universities, hospitals, climate adaptation, emergency services, local start-ups, non-profits, regional communities, and public-interest AI research. If AI displaces workers and reshapes professions, returns must be measured not only in tax revenue or GDP, but also in livelihoods and wellbeing.
These principles are not anti-business. They are pro-democracy, pro-responsibility, and pro-legitimacy. Brazil’s Pix payment system shows how a central bank can provide efficient, inclusive, and competitive public digital infrastructure. Switzerland’s Apertus shows that publicly developed, open AI infrastructure can be transparent, multilingual, and useful to society.
For a middle power, collective sovereignty requires international cooperation. Australia need not choose between US-style corporate domination and autocratic state control. It can help build a third path: democratic AI infrastructure governed through value-aligned partnerships with universities, civil society, and public institutions. A shared charter for AI infrastructure could establish enforceable standards for energy, water, labour, privacy, data portability, public access, Indigenous governance, and democratic oversight.
This would not retreat from global AI development. It would contest its terms. If data centres are treated only as investment pipelines, backlash will be predictable. If governed as shared infrastructure, they can expand public capacity. The test is not whether Australia can attract capital, but whether it can set terms over who controls and benefits from the digital future.
Dr Raffaele Ciriello is a scholar of compassionate digital innovation at the University of Sydney. His research examines ethical dilemmas in sociotechnical change, focusing on AI companions, decentralised platforms, and public digital infrastructure. His work appears in leading journals (e.g. Nature Machine Intelligence, The Lancet), alongside his book Compassionate Digital Innovation (Routledge). A Distinguished Member of the AIS, he serves in editorial roles at leading IS journals (EJIS, ISJ, JAIS) and has contributed over 300 reviews. He is also Debates Editor at CAIS and founding chair of the ACIS track on Digital Innovation for the Common Good, supporting socially engaged IS scholarship. He engages in public debate, contributing over 100 media commentaries (e.g. The Economist, Financial Times, 60 Minutes) and advising government bodies, including the NSW Department of Education and Australia’s eSafety Commissioner. He currently focuses on designing and governing emerging technologies for the common good.
This article is published under a Creative Commons License and may be republished with attribution.
