Australia’s data centre boom is accelerating but delivery risks are rising

Key insights

• Data centre commencements have risen almost 800% over the past three years, with the pipeline of work predominantly in Sydney and Melbourne
• Of the 78 occupations that RLB identified as critical to data centre construction, 54 (69%) are currently in shortage
• Key constraints to deliverability of the pipeline are power and water supply, specialist skills shortages, risks from global supply chains, and rising community opposition
• For builders and developers, deliverability and speed to market will depend on early, integrated planning, proactive procurement and DfMA to manage cost, supply-chain and energisation risks

Australia’s data centre construction boom is already under way

Data centres house the digital infrastructure needed to support cloud computing, artificial intelligence (AI), enterprise IT, streaming and other data-driven services.

Data centres have been built in Australia since the 2000s, but construction activity really took off in 2023.

Construction work the Australian Bureau of Statistics (ABS) calls ‘commercial buildings not elsewhere classified’ was valued at A$5.1 billion in the year to March 2026, up 226% from three years ago (see Chart 1). This category is used as a proxy for data centre construction, as data centres now make up almost the entirety of the category.

ABS data also shows data centre activity will ramp up considerably in coming years. Data centre commencements have risen almost 800% over the past three years, with commencements worth over $15 billion in the year to March 2026. In total there is currently $21.6 billion of work under construction (Chart 1).

The surge in data centre construction is being driven by rising AI workloads, cloud adoption, digital sovereignty and cybersecurity concerns. Australia is seen as an attractive location for new data centres due to political stability, space, connection to Asia and abundant renewable energy resources.

The immediate pipeline of work is predominantly in Sydney and Melbourne. There is over $11 billion of work under construction in NSW and $8 billion in Victoria (Chart 2). Data centre construction has been concentrated in Sydney and Melbourne due to the benefits of clustering facilities near existing expertise and skilled labour, major clients such as tech firms, banks and government agencies, and critical infrastructure.

Building approvals data also points to further strong growth in data centre construction activity in the near term, with over $17 billion of ‘commercial buildings nec’ approved over the past year, half of which are in Victoria and 41% are in NSW (see Chart 3).

Data centre approvals now make up around 19% of all non-residential building approvals, up from 6% just two years ago. In Victoria, approvals for new data centres account for 30% of all non-residential building approvals. Notable projects recently approved include the 504MW CDC data centre campus in Marsden Park and the 162MW M4 NEXTDC data centre campus in Fishermans Bend.

Structural works, finishes and labour represent major cost drivers in traditional building sectors. In contrast, data centres are dominated by electrical and mechanical systems including high-voltage infrastructure, Uninterruptible Power Supply (UPS) systems, generators, cooling plants, controls and commissioning. On top of this sits the IT fitout, including racks, servers, storage and networking hardware.

The long-term pipeline is huge

Data centre investment is forecast to surge in the coming decades, with the pipeline of activity potentially rivalling the mining construction boom of the 2010s.

Announced or early-stage large-scale projects include AirTrunk’s 1.2 GW hyperscale campus in Kemps Creek, the 550 MW NEXTDC and OpenAI S7 hyperscale campus in Sydney, Goodman’s 500 MW Project Atlas campus in Sydney, and AirTrunk’s 354 MW MEL2 hyperscale centre in Melbourne.

A range of estimates based on committed, planned or possible projects forecast that ‘operational capacity’ will more than double between 2025 and 2030, and then the potential pipeline rises even further in the 2030s and 2040s (Chart 4). ‘Operational capacity’ is the amount of installed data centre IT load that is available for customer or operator use, measured in gigawatts (GW) of IT load.

Westpac estimates this investment phase will be worth A$155 billion including associated energy infrastructure over the next decade.

The investment surge is expected to be concentrated in Sydney and Melbourne, with an estimated 95% of the potential pipeline of projects in Australia’s two biggest cities (see Table 1).

TABLE 1: Projected future data centre pipeline

Notes: Total pipeline includes under construction, committed and early stage projects. Source: Data Centres Australia, March 2026.

Hyperscale growth and environmental performance are accelerating the shift to Design for Manufacture and Assembly (DfMA)

The data centre industry is undergoing a significant design transformation, with hyperscale developments driving a shift towards Design for Manufacture and Assembly (DfMA).

This approach shifts construction from site-based delivery to controlled factory environments, supporting modular, repeatable and scalable designs. Across electrical and mechanical systems, DfMA can reduce onsite labour, trade interfaces and commissioning risk while accelerating deployment. As data centre campuses expand, modular architecture is likely to become a foundational design principle rather than a specialist solution.

• Electrical systems: Electrical infrastructure is increasingly delivered as factory-built modules integrating switchgear, transformers, UPS systems, controls and batteries. These modules can be fully tested prior to shipment and provide a scalable framework for campus expansion.
• Mechanical systems: Mechanical DfMA, including chilled water plant skids, pump skids and prefabricated pipework assemblies, is now mainstream on many developments. These systems package pumps, heat exchangers, valves, controls and associated pipework into factory-built assemblies that arrive onsite fully tested and ready for installation.
• Integrated infrastructure modules: The next phase of DfMA is moving beyond individual components towards fully integrated infrastructure modules. As AI workloads drive higher rack densities and more complex cooling requirements, the industry is increasingly adopting AI-ready cooling modules that combine Cooling Distribution Units (CDUs), heat exchangers, pumps, controls and monitoring systems into a single factory-manufactured solution. 
• Power and cooling pods: A further evolution is the emergence of integrated power and cooling pods, where electrical and mechanical infrastructure are combined into a single modular utility block. These pods incorporate UPS systems, switchgear, cooling equipment, controls and monitoring platforms within a factory-built enclosure. By reducing trade interfaces and consolidating commissioning, integrated pods simplify deployment and support a more standardised campus design.

Scrutiny of data centre energy consumption, carbon emissions and water usage is also reinforcing the adoption of DfMA. Key design metrics now include Power Usage Effectiveness (PUE), Water Usage Effectiveness (WUE), embodied carbon, renewable energy integration and lifecycle performance. Design teams are being asked to optimise efficiency while accommodating future growth requirements. Factory-based manufacturing generally reduces material waste, improves quality, minimises rework, and enables more efficient use of resources.

The key challenge is deliverability in the face of growing constraints  

Power and grid connection 

Access to power and grid connection will be a major constraint on deliverability. Historically, site selection was driven by land cost, connectivity and proximity to customers. For AI data centres, access to power capacity, transmission infrastructure and renewable energy supply is increasingly the key factor influencing investment decisions. In many cases, securing grid capacity has become more critical than securing the construction site itself.

Recent forecasts from AEMO predict data centres will consume between 8% and 12% of all electricity supplied through the National Electricity Market by the 2040s, up from around 2% currently (and an even higher share in Sydney and Melbourne). This ramp-up in demand could push up electricity prices if data centre developers don’t also invest in new renewables and storage. The government has announced plans to make new large-scale data centres underwrite new power supply, fund new renewable energy projects and pay for grid connections.

Data centres will add pressure to the grid and emissions, but they may also provide some flexible demand that helps absorb surplus generation, for example by running data-intensive AI training at times of peak renewables output. And while data centres are intensive users of electricity, they are more efficient than on-premises servers.

Specialist labour shortages will intensify

Availability of skilled and specialised labour will be another key constraint on the data centre construction pipeline. Data centres require both general construction trades and more specialised data centre trades. General trades include electricians, mechanical services plumbers and fire-protection trades. Many of these skilled workers are also needed for hospitals, life sciences facilities, defence, transport, renewables projects and some of the infrastructure for the Brisbane 2032 Olympics.

There is a particularly strong demand for electrical trades as hyperscale AI data centres require substantially more and complex electrical infrastructure due to higher rack densities and the shift towards 100MW-plus campuses. This shift is intensifying competition for specialists across high-voltage electrical systems, mechanical services, controls and commissioning. Hyperscale data centres also require chilled water systems, liquid cooling technologies and sophisticated controls platforms. This is increasing demand for specialist mechanical engineers and technicians with experience in thermal management, plant optimisation and energy efficiency.

Major hyperscale operators are expanding capacity across multiple regions, contributing to higher labour costs, greater reliance on international expertise and growing pressure on contractors to secure key personnel well before construction begins.

Jobs and Skills Australia analysis shows many of the skilled workers needed to build data centres are currently in shortage, particularly trades such as electricians, plumbers, air-conditioning and refrigeration mechanics, and mechanical services trades (Chart 5). Of the 78 occupations that RLB identified as critical to data centre construction, 54 (69%) are currently in shortage.

Global supply chains create long-lead procurement risk

Unlike many Australian construction projects, which rely heavily on local materials and established domestic supply chains, data centres depend on a highly specialised global manufacturing ecosystem. Critical equipment – including generators, UPS systems, switchgear, cooling infrastructure, controls systems and IT hardware – is predominantly sourced from international suppliers, making project delivery dependent on global production capacity. 

This reliance introduces a complex set of commercial risks. Project costs can be affected by currency fluctuations, international freight costs, geopolitical instability, trade restrictions and manufacturing backlog across key vendors. The rapid growth of hyperscale cloud computing and AI infrastructure globally has further increased competition for critical components, placing additional pressure on pricing and availability.

Lead times for major electrical and mechanical packages can extend beyond 12 months, requiring procurement decisions to be made before the design is fully finalised.

Water availability and cooling

Water and cooling are critical planning and operating constraints for data centre development. Water availability can shape site selection, community acceptance and approval risk, particularly where evaporative cooling is proposed.

Operators are increasingly adopting more efficient and higher-density cooling designs, including liquid cooling, as AI workloads lift rack densities beyond what traditional air-cooling systems can comfortably support. These technologies can reduce water intensity, but often increase design complexity, upfront cost and power requirements for heat rejection. Sustainable and responsible water use is therefore becoming a core deliverability test (and could potentially be mandatory under the government’s AI regulation agenda).

Planning restrictions and community acceptance could reshape site selection

Data centres are facing increasing community opposition over power and water use, visual bulk, noise, heat and construction traffic, in addition to broader community concerns about the social and employment impacts of AI.

These issues are more acute near residential or mixed-use areas. It is possible that community pushback may encourage larger campuses to be built in industrial, peri-urban or regional locations (although proximity to fibre networks and customers will still matter).

Overcoming constraints will shape what and how much gets built

Australia’s data centre pipeline is substantial. For developers and builders to overcome growing constraints, deliverability and speed to market will depend on resolving these issues early and together:

• Early project integration: Early, integrated cost, program and procurement advice will be central to project bankability and speed to market.
• Cost planning: Cost plans need to reflect the sector’s services-heavy cost base, long-lead equipment, escalation and the potential cost of delayed energisation.
• DfMA delivery: DfMA will be an important part of the response. Modular delivery requires earlier design decisions, supplier commitments and logistics planning.
• Supply chain management: proactive procurement strategies, including early supplier engagement, framework agreements and increased contingency allowances to mitigate program and cost risk.