The Global Contest in AI Infrastructure: Why governance decisions made now will determine Australia's place in the emerging digital order

John H Howard, 22 December 2025

This Innovation Insight examines how the global race in artificial intelligence is increasingly being fought through hard infrastructure, and what that means for Australian governance and public policy. It argues that decisions about power, planning, subsea connectivity and data stewardship will determine whether Australia becomes a trusted Indo-Pacific AI hub or remains a peripheral server-farm landlord.

For governments, this is no longer a narrow technology policy issue. It is a governance challenge that cuts across energy, planning, competition, foreign investment, cyber security and industrial strategy. Countries that recognise this will shape the norms of the emerging digital order. Those who do not will find the rules written by others.

A new geography of power

Industry analysts report that AI computing capacity is centralising into a few high-power corridors where fibre, electricity, capital and talent already concentrate. Northern Virginia is now widely regarded as the densest cluster of data centres in the world, accounting for around 13 per cent of global data centre capacity according to Virginia's legislature.

These 'AI spines' are said to enjoy powerful network effects: dense interconnection, fast connectivity, specialised contractors and rapid repeat permitting. If these accounts are correct, the advantages would be hard for latecomers to replicate and would lock in benefits for incumbent operators and host jurisdictions.

As the computing demands of AI systems grow, commentators suggest that the binding constraints are shifting from semiconductors and software to megawatts, grid resilience, land, cooling and regulatory predictability. Data centres, it is argued, have become de facto national-security and industrial-policy assets.

Policy risk appears to run in both directions: over-concentration in a few hubs may create geopolitical vulnerabilities, but jurisdictions that move too slowly on energy infrastructure and approvals risk being bypassed in the global build-out.

Other governments have responded accordingly. In the United States, federal energy agencies now treat AI-related data-centre demand as a distinct planning category, with scenario analysis and priority transmission projects for AI hubs. The Trump Administration's July 2025 executive order on accelerating federal permitting for data centre infrastructure, and the associated rollback of constraints on gas-infrastructure approvals, were explicitly justified in terms of powering AI data centres.

European institutions are incorporating AI-related demand into energy-security and climate discussions, prompting calls for efficiency standards and coordination between the digital industry and energy regulators. The International Energy Agency projects that data-centre electricity demand will roughly double globally by 2030, mainly due to AI, and suggests this will require hundreds of billions of dollars in new generation and grid infrastructure.

Where Australia sits

Australia will not become a Northern Virginia-scale node, but industry reports suggest it is emerging as one of the more attractive data-centre markets in the Asia-Pacific. Sydney and Melbourne are described as core regional nodes, attracting large commitments from global cloud providers. Political stability, rule of law and strong subsea connectivity are cited as factors making Australia a preferred location for large-scale data centre investment.

According to Knight Frank's Global Data Centres Report, 2025, Australia received US$6.7 billion in data centre investment in 2024, making it the second-largest recipient globally behind the United States.

Yet an 'energy gap' appears to be opening. According to industry and market operator sources, grid constraints, long connection lead times and limited firm generating capacity mean operators receive only a fraction of the megawatts they request. Some analysts have projected a shortfall between 0.7 and 1.7 GW even as billions in potential investment line up. If these projections hold, Australia may be structurally attractive and growing fast, but at risk of losing ground if it cannot secure scalable, ultimately low-carbon power at the pace investors require.

The realistic strategic objective is to become the trusted, resilient Indo-Pacific hub that global and regional firms rely on for sovereign, fast and secure AI capacity in a stable democracy. Achieving this means treating large-scale data centres as strategic infrastructure, akin to ports or LNG terminals, and aligning energy, planning, security and industrial policy around that choice.

The governance gap

Australian policymakers have recognised parts of this agenda, but the treatment has been fragmented. Subsea cables are classified as critical infrastructure and wrapped in protection zones and cybersecurity strategies. The National AI Plan positions Australia as an AI and data-centre hub and foreshadows national data-centre principles. Yet external commentary notes the Plan is ambitious but light on solutions to permitting and grid bottlenecks.

The Plan‘s reliance on market forces sits uneasily with the need for deliberate corridor and hub design. The Indo-Pacific 'digital belt' is held together by subsea cables that carry almost all international data traffic, making their landing points both strategic opportunities and vulnerabilities.

For high-value datasets, choices about which corridors to use determine which foreign legal regimes and surveillance capabilities sit in the path between Australian origin and overseas users.

What needs to happen

Competing in the global AI infrastructure race requires shifting from project-by-project approvals to an explicit AI-infrastructure industrial strategy. This means considering several issues:

• On power, a deliberate strategy for dedicated supply to data-centre precincts, blending large-scale renewables, back-up generation and new transmission, with clear rules on how these loads interact with the broader grid and emissions trajectories. If gas is to play a transitional role, it should be time-bound with strict emissions-intensity caps and sunset schedules, conditional on efficiency measures in the data centres themselves.

• On approvals and permitting, the proposed national data-centre principles must translate into faster, more predictable approval pathways for projects that meet the criteria. The 'map' of AI infrastructure should become a conscious instrument of industrial policy: government backs a set of AI-ready corridors and systematically de-risks power, grid and permitting for those zones in return for explicit commitments from investors.

• On subsea governance, consider which Indo-Pacific cable routes matter most for Australia’s datasets and negotiate corridor-specific arrangements combining technical routing preferences, joint investment in redundancy, and agreed privacy and surveillance standards with key partners.

• On institutions, a single federal lead with authority to integrate AI Plan implementation, grid planning, data-centre principles, foreign investment settings and subsea-corridor strategy is essential. The current fragmentation across energy, communications, home affairs, industry and treasury portfolios makes coherent action difficult.

Hub or landlord?

AI infrastructure is now a central instrument of statecraft. The global race is being won and lost through decisions about power, planning, subsea connectivity and data stewardship. Australia faces a choice between two futures.

In one scenario, Australia becomes the trusted Indo-Pacific AI hub: a large-scale, largely renewable-backed data-centre base across AI-ready corridors, plugged into carefully governed regional cable routes that carry Australian datasets through trusted partners under predictable legal regimes.

Multinational AI firms default to Australian locations because the rules are clear, approvals are predictable, and data corridor governance is credible. Physical build-out is harnessed to domestic capability through skills, research, procurement and local firms. Australia captures value, sets standards and builds high-value AI-enabled services.

In another scenario, Australia becomes a peripheral server-farm landlord: hosting important infrastructure, but as a price-taker on governance and value capture. Global platform companies and cable consortia dictate the geography, resilience and legal environment of the AI backbone. Australian-created and curated datasets flow through systems whose rules are set by others. The infrastructure race is decided elsewhere.

The difference between these futures lies in governance. Siloed risk management, slow permitting and reactive approaches to subsea corridors lead to the second outcome. Treating AI infrastructure as a governed strategic asset, with deliberate coordination of power, planning and subsea strategy around clear national objectives, leads to the first.

The window for shaping the networks on which Australia's digital future will rest is open now. It will not stay open indefinitely.