Australia’s data centre pipeline growing to $260bn in a decade: CBA

Australia’s data centre construction boom is accelerating as artificial intelligence and cloud computing generate demand for more processing capacity.

CBA economists have upgraded their estimate of Australia’s feasible additional data centre pipeline from 6 gigawatts to 7GW.

The expansion could more than quadruple Australia’s total rated capacity to around 9.2GW by the mid-2030s, with construction activity expected to peak in the 2028 and 2029 financial years.

CBA Associate Economist Lucinda Jerogin said the build-out would be significant even after accounting for projects that may not proceed.

“Importantly, the scale of this additional 7GW pipeline remains a very large development for Australia, suggesting total rated data centre capacity could more than quadruple in under a decade with a construction boom underway,” Jerogin said.

Announced capacity exceeds the feasible pipeline

About 23GW of data centre capacity has been publicly announced across Australia, but most projects remain some distance from construction.

Around 3.6GW is under construction or being expanded, while a further 2GW has received development approval or moved into pre-construction. Together, these projects account for 24 per cent of the announced pipeline.

The remaining 76 per cent is in planning or at an earlier stage, making its timing and eventual scale less certain.

“Projects can be redesigned, delayed or delivered in smaller phases as customer demand, grid access and financing become clearer,” Jerogin said.

Data from the Australian Energy Market Operator points to an even larger potential pipeline, with 225 projects seeking almost 67GW of connection capacity as of June 2026.

However, CBA said this was likely to overstate the capacity eventually delivered because applications can overlap and developers may be competing for the same customers. More than one-third of connection requests made over the previous year were withdrawn, while the median probability of a project progressing from its current stage to completion was estimated at 28 per cent.

“The high degree of attrition and duplication in AEMO’s connection data demonstrates even meaningful progress through the connection process does not guarantee delivery,” Jerogin said.

“This reinforces our scepticism of the gross 23GW pipeline.”

AI demand supports further development

CBA’s central estimate is broadly consistent with AEMO’s electricity demand forecasts, which imply an additional 5.5GW to 6.5GW of rated data centre capacity by the 2036 financial year.

The analysis also considers potential Australian demand associated with OpenAI and Anthropic, alongside demand from cloud providers, businesses, governments and research organisations.

Under CBA’s central scenario, Australia captures 10 per cent of the two companies’ estimated global compute demand. On the assumptions used in the report, total additional Australian capacity would reach approximately 6.5GW to 7.9GW.

“A meaningful Australian allocation is plausible,” Jerogin said.

An upside scenario in which Australia becomes a major location for AI model training and large-scale inference could lift additional capacity to between 9.2GW and 11GW. These are scenario assumptions rather than disclosed commitments.

Investment could reach $260bn billion

CBA expects the data centre build-out to attract around $150 billion in investment by 2030, supporting approximately 4GW of additional rated capacity.

A further $110 billion after 2030 would take total investment to around $260 billion as the central 7GW pipeline is delivered by the mid-2030s. About 70 per cent of construction costs are expected to relate to specialised IT equipment rather than buildings and structures.

Constraints could slow delivery

Energy infrastructure is the most immediate constraint. According to AEMO, Australian data centres consumed an estimated 5 terawatt hours of electricity in 2025–26 , equivalent to 2.8 per cent of electricity supplied through the National Electricity Market.

AEMO forecasts consumption could rise to approximately 15TWh in 2029–30 and 34TWh by 2035–36.

Community opposition to large developments has also increased, while uncertainty around copyright law could affect whether large-scale AI training is based in Australia.

“Risks are two sided,” Jerogin said.

“If Australia becomes a major location for frontier model training and large scale inference, rated capacity could increase by a much larger 9-11GWs. Conversely, weaker customer demand or more binding energy, social licence and policy constraints could delay or limit delivery.”

You can read Lucinda's full economic research report here.