Australia could capture a growing share of Asia-Pacific AI demand while reducing emissions compared with more carbon-intensive markets, according to industry research commissioned by Data Centres Australia and prepared by economics consultancy Mandala Partners.
The report argues that Australia is unusually well placed to become an exporter of AI compute because of its renewable energy potential, available land, skilled workforce and political stability. It estimates that Australia could capture between 1.9GW and 2.9GW of additional latency-tolerant AI demand from APAC and global markets by 2030, on top of projected domestic capacity.
That additional capacity could generate between AUD 2.7 billion and AUD 4.1 billion in annual economic activity and support up to 17,120 ongoing jobs, while requiring as much as AUD 52 billion in additional construction investment, Mandala estimates.
The report, The Economic Contribution of the Data Centre Industry in Australia, uses an input-output model based on Australian Bureau of Statistics data and financial information supplied by a group of DCA member organisations. It estimates that Australia’s existing 1.6GW of data centre capacity generates AUD 2.3 billion in annual economic activity and supports around 9,450 ongoing jobs across the industry and supply chain.
By 2030, projected domestic capacity of 3.9GW would require around AUD 42 billion of investment and is forecast to generate AUD 5.6 billion in annual economic activity and support 23,040 ongoing jobs. Construction of the additional 2.3GW is estimated to generate AUD 19.8 billion in economic activity and support 18,930 construction jobs.
Batch opportunity
The report’s more ambitious case is based on the proposition that a significant proportion of AI training and batch inference can be located away from end users. Mandala estimates that 34% of global data centre demand in 2030 will be in APAC and assumes half of APAC workloads will be AI-related. It then applies a 55% share to split inference into batch and real-time, while treating AI training as fully latency-tolerant, to derive the addressable export market.
Australia’s potential share of that market is then determined through a comparative scoring model covering power availability, land, water, build time, political stability, workforce and connectivity. The resulting score is converted into a market share and applied to the estimated addressable demand.
The environmental case is a central part of the report. Mandala estimates that hosting 1.9GW of additional export compute in Australia would produce 3.2 million tonnes of CO2 emissions in 2030, compared with 11.9 million tonnes if the same capacity were hosted in Indonesia. It therefore estimates Australian compute would have 3.7 times lower emissions on this basis.
Mandala describes this as effectively exporting renewable energy through compute, arguing that Australia can embed the value of its renewable electricity in digital services rather than attempting to export electricity physically. Mandala partner Tom McMahon said Australia’s existing subsea cable infrastructure makes this approach particularly attractive.
Scope 2 treatment
The comparison, however, is specifically a comparison of location-based Scope 2 emissions from grid electricity consumption. The report notes that it does not include Scope 1 emissions such as backup generation and refrigerants, embodied emissions from construction, or the additional firming and backup generation required to manage renewable intermittency. Those factors are assumed to be constant between countries.
As a result, this reads as a measure of economic activity generated through direct expenditure and supply chains, as opposed to a conventional cost-benefit assessment of data centres against alternative uses of capital, energy or infrastructure. The jobs estimates also include full- and part-time employment and are explicitly not equivalent to FTE employment.
Those qualifications are significant given the scale of the report’s projections, particularly the potential AUD 52 billion of additional investment associated with the AI export scenario.
Data Centres Australia CEO Belinda Dennett said the findings demonstrated the importance of digital infrastructure to Australia’s economic and energy-transition ambitions. “By creating the right policy environment, we can capture more of the AI value chain locally, support thousands of high-skilled jobs, and play a pivotal role in the clean energy transition,” Dennett said.
Clean Energy Council CEO Jackie Trad said data centres could help unlock new renewable generation and create new markets for Australian compute. “When data centres help bring in renewable energy projects, they can both boost Australia’s energy supply and create new, untapped markets for Australian compute,” she said.
Compute hub
The report also makes a strategic case for Australia becoming an AI compute hub, arguing that hosting infrastructure could give Australia greater influence over AI governance, data sovereignty and security, while attracting technology investment, talent and research activity.
Mandala acknowledges near the end of the report that its projections are “an opportunity, not a guarantee”, and says the gap between Australia’s current trajectory and the full export opportunity will not be closed by market forces alone. It identifies energy, workforce, planning, investment and policy settings as determining factors, while explicitly recognising trade-offs involving energy systems, communities and the national interest.
That final caveat is important: the report provides a modelled case for what Australia could capture if it becomes a competitive destination for AI infrastructure, rather than a forecast that the additional capacity, investment or emissions reductions will necessarily occur.