The Cost of AI Data Centres in Australia

What does an AI data centre cost?

There is no single Australian price. A small enterprise facility and a hyperscale AI campus have very different designs. Location, power availability, connection works, land, cooling, redundancy, chip supply and construction timing can all materially change the result.

The main cost layers include:

• Land and construction: secure buildings, reinforced floors, fire systems and physical protection.
• Computing hardware: graphics processors, memory, storage and high-speed networking.
• Power infrastructure: grid connections, substations, transformers, cabling and sometimes new generation.
• Cooling and water: chillers, cooling towers, pumps, treatment and recycled-water systems.
• Resilience: batteries, generators, duplicate network paths and spare equipment.
• Operations: electricity, maintenance, security, software, staff and replacement of ageing hardware.
  
  

The hardware cycle matters. AI processors can become commercially outdated well before the building does, so operators may face repeated capital upgrades during a facility’s life.

The electricity and grid cost

The Australian Energy Market Operator says 162 operational data centres account for around 2% of current grid-supplied electricity use. It has projected consumption reaching about 12 terawatt hours, or around 6% of grid-supplied electricity, by 2030 and about 34 terawatt hours, or 12%, by 2050.

That growth is not guaranteed, but it is large enough to affect system planning. At the end of the March 2026 quarter, 11 large projects representing 5.4 gigawatts of maximum demand were progressing through the transmission connection process, mostly in New South Wales and Victoria.

Connecting a large new load can require network upgrades and additional generation. The cost is not just the electrons a facility buys. It can include the infrastructure needed to deliver reliable power at the right place and time.

Water, land and local infrastructure

Data centres mainly use water for cooling and heat rejection. Actual use varies widely with design, climate and operating choices. Infrastructure Australia estimates a 250 MW hyperscale facility could use potable water equivalent to thousands or tens of thousands of households each year, depending on water-use effectiveness.

The local effect matters more than a national average. Water demand can be concentrated in urban growth areas that already need investment for homes, industry and drought resilience. Infrastructure Australia expects data centres to invest between $500 million and $1.1 billion in recycled-water infrastructure by 2030.

There are also opportunity costs. Large sites compete for appropriately zoned land, construction capacity, electrical equipment and skilled labour. Communities may experience road works, transmission construction, noise from backup generation and pressure on local services.

The benefits on the other side

Data centres can bring construction spending, long-term technical jobs, tax revenue, digital resilience and local access to computing capacity. They can also support renewable energy projects through long-term power purchase agreements.

Some computing workloads can be scheduled when renewable supply is abundant, although many customer and cloud services still require continuous availability. The public value depends on how facilities are connected, operated and integrated with the energy system.

Who should pay?

The Australian Government’s 2026 expectations for data centre and AI infrastructure developers say operators should underwrite new renewable supply, pay their full share of new grid connectivity so costs are not passed to households or other businesses, support demand flexibility and use water sustainably.

These expectations reflect the central policy issue: private investment can create national benefits, but the supporting costs should be transparent and allocated fairly. Planning, connection and water decisions need to consider both the project and the surrounding community.

What the cost means for AI buyers

Most Australian organisations will not build a data centre, but they still pay for infrastructure through cloud subscriptions, usage fees, contracts and future price changes. AI governance should therefore include:

• understanding where data is processed and which suppliers provide the computing capacity
• checking how prices may change as usage, energy costs or hardware demand grows
• planning for outages, capacity limits and supplier exit
• asking for meaningful energy, water and emissions information where it matters to procurement
• avoiding unnecessary AI use when a simpler system can perform the task
  
  

For related guidance, see Australian data centres and AI infrastructure demand, AI growth and chip prices, AI risks and the workplace AI governance framework.