Australia’s AI buildout will be decided project by project

PwC estimates that Australia could attract US$703 billion in cumulative AI infrastructure investment through 2050. Its September 9 outlook projects annual data centre spending rising from US$16.5 billion in 2026 to US$41.6 billion in 2050.

The long horizon deserves as much attention as the large number. This is a forecast of spending across decades, including equipment that will need replacement. It is not a list of projects with committed funding.

For Australia, the useful discussion starts with what would allow individual projects to proceed, and what the country receives in return.

The equipment changes faster than the building

PwC projects that ICT equipment will account for US$38.2 billion of annual Australian investment by 2050, compared with US$3.3 billion for construction. Its forecast describes an infrastructure base that continues to require substantial spending after the buildings exist.

A completed shell does not settle the cost of keeping a facility competitive. The owner needs to consider which equipment tenants will require, how the facility accommodates replacement and who bears the expense when specifications change.

A hypothetical project with a long-lived site and shorter-lived equipment combines different risks. The site may have durable value because of its connection to power and networks. The equipment earns its value through work performed while customers want to use it. Treating the entire investment as one uniform asset can obscure that distinction.

Local benefits need a mechanism

The Australian Government’s September consultation on AI infrastructure addresses energy, water, community costs, skills and local capability. These are proposed requirements under consultation. They show the questions government is asking as investment grows.

A project can generate benefits through construction, demand for local suppliers, access to computing capacity or research partnerships. Those benefits have different recipients and time horizons. A temporary construction workforce should be described separately from ongoing operational employment. Access for local researchers depends on the terms offered, not simply on the physical presence of servers.

The same care applies to costs. A developer’s plan for electricity needs to explain how supply becomes available and when. A commitment on water needs an operating assumption behind it. Communities can assess a specific proposal more usefully when those details are visible.

A forecast is a starting point for diligence

Our view is that the national headline is most useful as a reason to examine the project pipeline. A credible pipeline would distinguish proposed capacity, contracted capacity, sites with necessary approvals and facilities already operating. Moving from one category to the next is the evidence that a buildout is progressing.

PwC’s forecast makes the potential scale legible. The decisions that determine Australia’s share will involve particular sites, electricity connections, customers and contracts. Those are also the places to test whether the investment creates the local capability its supporters expect.