The numbers involved in the current artificial intelligence infrastructure boom have become so large they risk losing meaning. Hundreds of billions of dollars. Thousands of megawatts of power. Millions of square metres of server space. Yet behind the abstractions lies something concrete: a small group of American corporations is building the foundational infrastructure upon which the next generation of global technology will run, and they are doing so at a pace that has left governments, regulators, and rival economies scrambling to respond.
According to reporting by TechCrunch, the major players in this build-out include Meta, Microsoft, Google, Oracle, and OpenAI, each committing to infrastructure investments that, taken together, likely exceed one trillion US dollars over the coming decade. These are not speculative projections. They are announced capital expenditure figures, data centre construction contracts, and power purchase agreements that are already reshaping energy grids, land use patterns, and labour markets across the United States and beyond.
Meta has pledged to spend up to 65 billion US dollars on AI infrastructure in 2025 alone, a figure that dwarfs the annual budgets of most national science agencies worldwide. The company is also reported to be laying a subsea cable that would circle the entire globe, a project of a scale previously associated only with sovereign telecommunications authorities. Microsoft, through its partnership with OpenAI, has committed to the Stargate project, a joint venture with Oracle and SoftBank targeting 500 billion US dollars in US-based AI infrastructure investment. Oracle, for its part, is constructing what it describes as the largest AI data centre in the world in Texas, a facility that will require its own dedicated power generation.
The clinical significance of this investment, to borrow a phrase more common in medicine than economics, is not merely about competitive advantage between corporations. It is about who controls the compute layer of the global economy. Artificial intelligence systems, from the large language models now embedded in productivity software to the diagnostic tools being trialled in hospitals, run on physical hardware. That hardware is overwhelmingly concentrated in the United States, with secondary clusters in Europe and East Asia. Australia, despite its advanced economy and strong research institutions, remains a relatively minor player in this infrastructure race.
The Australian Department of Industry, Science and Resources has acknowledged the strategic importance of sovereign AI capability, and the Albanese government has made gestures toward supporting local AI development through its National Reconstruction Fund and various research grants. But the gap between Australian ambition and American scale is considerable. A few hundred million dollars in domestic investment looks modest when set against Meta's annual data centre budget.
There is a legitimate argument, advanced by those on the centre-left of the policy debate, that Australia should not be trying to compete directly with Silicon Valley on infrastructure. Instead, this view holds, Australia's comparative advantage lies in specialised applications, in applying AI tools to agriculture, mining, healthcare, and environmental monitoring in ways that suit our specific geography and industrial base. The CSIRO has pointed to exactly these sectors as areas where Australian AI research has genuine global relevance.
That argument has real merit. But it does not fully address the sovereignty question. When the underlying compute infrastructure is owned and operated by foreign corporations subject to foreign law, Australian businesses and government agencies using those systems are exposed to risks that extend beyond ordinary commercial considerations. Data residency, jurisdictional access by foreign intelligence agencies, and the terms under which AI services can be withdrawn or modified are all concerns that a serious national technology strategy must confront directly.
The environmental dimension adds another layer of complexity. The energy appetite of large-scale AI training is significant and growing. The International Energy Agency has estimated that data centres could account for up to 4 per cent of global electricity consumption by 2026, up from around 1.5 per cent in 2022. The major technology companies have made renewable energy commitments, and some, like Google and Microsoft, have genuinely ambitious decarbonisation programmes. But the speed of infrastructure expansion is outpacing the availability of clean power in many regions, and there are credible questions about whether net-zero pledges can be maintained at this rate of growth.
For Australia, the energy question cuts in an interesting direction. The country has some of the best renewable energy resources in the world, abundant land, and a relatively stable regulatory environment. There is a credible case that Australia could position itself as a preferred location for data centre investment by technology companies seeking to meet their sustainability targets, particularly given the growing demand for AI compute in the Asia-Pacific region. Singapore, which has imposed moratoriums on new data centre construction due to power constraints, is no longer able to absorb regional demand on its own.
What the data actually tells us is that the AI infrastructure race is not a passing phase. The capital already committed, the contracts already signed, and the hardware already on order represent a structural shift in how the global technology sector is organised. The question for Australian policymakers is not whether to engage with this shift, but how. Fiscal responsibility argues against blank-cheque subsidies for data centre construction. National interest argues against passive acceptance of a status quo in which Australian digital infrastructure depends entirely on the goodwill of foreign corporations.
The most defensible path forward probably involves targeted investment in research capability, clear regulatory frameworks for data sovereignty, and active efforts to attract responsible foreign investment in local infrastructure, while maintaining honest expectations about the limits of what a middle-power economy can build on its own. The scale of what is happening in the United States is genuinely extraordinary. Australia does not need to match it. But it does need a considered response to it, and that response is still taking shape.