Australian data centres sit at the centre of the AI boom: massive economic opportunity colliding with hard physical limits on power, water, and the grid.

Main Issues

Energy demand and grid pressure. Data centres currently consume around 2% of electricity on the National Electricity Market (roughly 4 TWh, or the equivalent of 700,000+ homes). Demand has already nearly doubled in Victoria and risen ~18% in NSW in a single year. AEMO forecasts it tripling to ~12 TWh (around 6% of the NEM) by 2030, with longer-term estimates reaching 8–11% of national consumption by 2035.

The development pipeline is enormous: tens of gigawatts of proposed capacity (one figure puts maximum demand from all proposed projects above 21 GW, more than seven times Eraring, Australia's largest coal station). Much of this is "phantom demand" (connection requests that will never fully materialise), but the volume still overwhelms planning.

Risks include higher wholesale power prices (modelling shows potential rises of 20%+ nationally, and up to 23–26% in NSW and Victoria by 2035 if new load is not matched by additional renewables and storage), delayed coal retirements, greater reliance on gas, and grid stability problems. Large inverter-based loads can trip simultaneously during disturbances, creating blackout risks similar to incidents overseas. Governments are responding with proposed "BYO energy" rules requiring centres to add at least as much new renewable capacity as they consume, plus demand flexibility (curtailing at peak times) and network cost contributions. Timelines are mismatched: data centres can be built faster than the renewables meant to supply them.

Water use: Current national consumption is modest (~5.5 GL/year, or ~0.04% of total water use, far below mining or manufacturing). Locally it is more significant (around 0.7% of Sydney's supply and 0.2% of Melbourne's). Industry forecasts see demand more than tripling. High-growth scenarios for Sydney project data centres taking 15–25% of supply or hundreds of megalitres per day by the mid-2030s if water-intensive cooling dominates. Australia's dry climate and drought risk amplify the issue. Some projects are shifting to waterless cooling, which raises energy use instead.

Community opposition, planning, and delivery constraints: Noise, visual bulk, diesel backup generators, heat, traffic, and proximity to homes or schools generate strong local pushback (record objections at sites such as Lane Cove; some Perth councils considering outright bans). Skills shortages affect nearly 70% of critical construction occupations. Specialised global supply chains create long-lead risks. Approval processes, grid connections, and water access add delays.

Scale of investment and economic activity Australia ranks among the top global destinations for data-centre capital (second only to the US in recent years). The pipeline is estimated at $150 billion or more by 2030 (some figures run higher). This is already driving sharp rises in private capital expenditure in the information and telecommunications sector and is projected to add several percentage points to business investment growth in the near term.

Value-add is high relative to energy use: the tech sector (including data centres) generates roughly $12.6 billion in gross value added per terawatt-hour, ahead of mining and manufacturing. Construction creates temporary jobs and stimulates supply chains; once operating, the facilities enable far broader economic activity.

Countries that host the infrastructure capture a disproportionate share of the productivity and sovereignty benefits. Australia has structural advantages: available land, strong renewable energy potential, political and legal stability, and proximity to Asian markets. Hyperscalers and AI firms (Microsoft, OpenAI, Amazon, Google, Meta and others) are already committing capital. Analyses suggest becoming a regional compute hub could deliver tens of billions in extra annual GDP and tens of thousands of jobs over time, while supporting data sovereignty and cyber resilience.

Enabling effects: Properly regulated, the demand can accelerate storage build-out, fund grid upgrades (sometimes creating excess capacity usable by others), and drive more efficient centralised compute versus scattered on-premise systems. The alternative: falling behind in digital infrastructure, means importing AI capability and ceding the productivity gains to communist China.

Unstoppable Until the Crash

The IT/AI world currently behaves like a classic infrastructure supercycle. Global demand for compute is real and rising steeply because large language models, inference at scale, and enterprise AI adoption are compute-hungry. Capital is flooding in, announcements multiply, and governments compete to attract the investment. In Australia the surge already resembles (and in some metrics temporarily eclipses) earlier mining investment booms: large, lumpy, relatively interest-rate insensitive, and capable of sucking in labour, materials, and policy attention.

Like previous cycles, it contains the seeds of overshoot. "Phantom" capacity applications inflate the numbers. Energy and water constraints, community resistance, skills shortages, and the possibility that AI returns prove slower or narrower than the hype all create the conditions for a sharp correction. When the downswing arrives, whether from overbuild, power shortages, higher costs, or disappointment on productivity payoffs, the drag on investment and activity can be severe, just as the mining boom's hangover was.

Until that point, the momentum remains powerful. The combination of genuine technological demand, first-mover advantages in hosting capacity, and the fear of being left behind keeps the build-out rolling. Australia's challenge is not whether data centres will expand, but whether the expansion can be forced onto a sustainable energy and water footing fast enough to capture the upside without locking in the downside.