As AI drives unprecedented demand for power, cooling and land, regional Australia has an opportunity to turn its energy advantage into a new generation of digital infrastructure, investment and skilled jobs.
Australia’s demand for data centre capacity is accelerating. Artificial intelligence, cloud computing and increasingly data-intensive businesses are driving an extraordinary requirement for new digital infrastructure. AEMO estimates data centres currently consume around 2% of grid-supplied electricity in Australia, predominantly in Sydney and Melbourne, but under its Step Change scenario that could rise to around 6% by 2030 and continue growing substantially from there. (AEMO)
That creates a problem. The places where we have traditionally built large data centres are increasingly the places where building the next generation of them is becoming most difficult. Sydney, Melbourne and other metropolitan centres face growing constraints around electricity generation and transmission, water availability, land, network capacity and competing infrastructure demands. Data centres are increasingly competing for many of the same resources required by rapidly growing cities and existing industry.
Perhaps we are asking the wrong question. Instead of continually asking how we can squeeze more data centres into metropolitan areas, we should be asking whether some of the next generation of Australian data centres should be built in regional Australia, where land and energy resources can potentially be developed around the computing requirement.
Follow the energy, not just the population
Historically, proximity to major population centres and telecommunications infrastructure was a major factor in determining where computing infrastructure was located. That equation is changing as data centres become larger and electricity becomes an increasingly important factor in where they can practically be built.
A modern hyperscale or AI data centre is not simply another commercial building. It is an industrial facility that consumes enormous amounts of electricity and converts almost all of it into heat. For many workloads, particularly AI training, batch processing, cloud storage and other applications that are less sensitive to small increases in latency, there is a strong argument for locating at least some computing infrastructure closer to abundant energy rather than expensive metropolitan land.
Western Queensland provides an interesting example. Towns such as Roma and Charleville sit within regions possessing several characteristics that deserve serious consideration for future data centre development: abundant land, high solar irradiance, existing energy infrastructure and improving telecommunications options.
The solar resource alone is substantial. Bureau of Meteorology long-term climate statistics for Charleville record average daily solar exposure of approximately 22.7 MJ/m², illustrating the strength of the solar resource available across western Queensland. (Bureau of Meteorology)
Solar alone, however, does not solve the data centre problem. Servers do not stop operating when the sun goes down, which means regional data centres would need to be designed around a combination of renewable generation, storage, grid connections and reliable firming capacity.
Gas infrastructure could provide firming capacity
This is where locations such as Roma become particularly interesting. Nearby Wallumbilla is one of Australia's significant gas hubs, connecting major pipelines across Queensland and into southern markets. APA's Roma Brisbane Pipeline runs approximately 438 kilometres from the Wallumbilla gas hub near Roma to Brisbane, while the South West Queensland Pipeline provides another major connection west toward Ballera and the broader east-coast gas network. (APA – Roma Brisbane Pipeline) (APA – Wallumbilla Gas Hub)
That existing infrastructure creates an opportunity worth investigating. Local gas turbine or reciprocating generation could potentially provide firm capacity when renewable generation is insufficient, while large-scale solar and battery storage provide an increasing proportion of total electricity consumption.
Rather than viewing the energy question as solar or gas, a regional data centre precinct could be designed around an integrated energy system combining solar, battery storage, gas firming and a grid connection. Gas would not necessarily need to remain the dominant source of electricity over the lifetime of the facility. It could provide reliability while additional storage, renewable generation, enhanced geothermal or other firm low-emissions technologies become commercially available.
This is not purely theoretical infrastructure planning. APA is already developing a pipeline connection from its Roma-to-Brisbane system to support CS Energy's proposed Brigalow Peaking Power Plant near Kogan, demonstrating how existing regional gas transmission infrastructure can support new electricity generation projects. (APA)
The geothermal lesson
Recent developments in the United States demonstrate why we should think more creatively about the relationship between energy and data centres. Google partnered with Fervo Energy in Nevada to develop enhanced geothermal generation supporting the electricity system serving Google's data centres.
Traditional geothermal power depends heavily on finding naturally favourable underground conditions. Enhanced geothermal systems approach the problem differently. Fervo has adapted horizontal drilling and reservoir engineering techniques developed by the oil and gas industry to access hot underground rock, circulate fluid through engineered fractures and return that heat to the surface, where it can be converted into electricity.
The initial Google-supported project began delivering electricity in 2023, and the relationship has since expanded. In 2025, Nevada regulators approved an agreement adding 115 MW of new enhanced geothermal capacity from Fervo to support Google's Nevada operations. (Google)
The lesson for Australia is not necessarily that we should immediately begin drilling geothermal wells beneath western Queensland. The important lesson is the change in thinking. Google is demonstrating that the future of data centre development does not have to mean finding an existing location with enough spare electricity and building there. It can mean identifying a suitable location and building an energy system around the computing requirement.
Australia should be having the same conversation.
Cooling needs to be designed around the location
Power is not the only infrastructure constraint facing data centres. Cooling enormous quantities of computing equipment requires significant engineering and, depending on the technology used, potentially significant quantities of water. That can create understandable concerns when large data centres are proposed in metropolitan areas where water infrastructure is already supporting millions of residents and businesses.
Western Queensland presents an interesting engineering trade-off because the climate is hot but relatively dry. Evaporative cooling has been successfully used in dry climates for decades because low humidity allows water evaporation to remove substantial quantities of heat. The disadvantage is obvious: the process consumes water, which makes water availability and security an important consideration for any regional facility.
That does not mean evaporative cooling is the only option. Closed-loop cooling architectures using water-glycol mixtures, dry coolers, heat exchangers and increasingly sophisticated liquid-cooling systems could significantly reduce dependence on continuous evaporative water consumption. The trade-off is that dry cooling can require more electricity, particularly during extreme summer temperatures, so the appropriate solution would need to be engineered specifically for each location.
This is precisely why regional data centres should not simply replicate metropolitan designs on cheaper land. The energy system, cooling system and computing infrastructure should be designed together around the characteristics of the site.
What if the waste heat became useful?
There is another opportunity that deserves considerably more attention. Data centres consume electricity to perform computation, but almost all of that electrical energy eventually becomes heat. Enormous amounts of engineering effort and electricity are then devoted to getting that heat out of the building.
A regional technology precinct creates the opportunity to investigate whether some of that heat could instead become an input into neighbouring industries. Depending on temperatures and economics, potential applications could include controlled-environment agriculture, greenhouses, water treatment, industrial processes or other businesses requiring low-grade thermal energy.
Not every application will be commercially viable, particularly in Queensland's warmer climate, but the broader principle is important. Instead of treating a data centre as an isolated building surrounded by cooling equipment, we should consider whether it can become part of a broader energy and industrial ecosystem.
Connectivity is becoming a solvable engineering problem
Connectivity would once have been one of the strongest arguments against regional data centres, and terrestrial fibre would still be essential for any serious hyperscale development. Data centres require enormous bandwidth, low latency for certain workloads and multiple redundant network routes. Starlink or other satellite services are not substitutes for that capacity.
However, the communications landscape has changed substantially. Inland fibre routes already cross regional Australia, and additional fibre infrastructure can be constructed where sufficient commercial demand exists. At the same time, low-Earth-orbit satellite networks have created another layer of communications resilience that simply did not exist at scale a decade ago.
Starlink has demonstrated the capability of LEO satellite communications across regional Australia, while Amazon's Project Kuiper, now Amazon Leo, and other constellations are expected to add further competition and capacity. These systems should be viewed as supplementary resilience rather than replacements for high-capacity terrestrial fibre.
The important point is that connectivity is no longer an automatic reason to dismiss regional locations. It becomes an infrastructure question: what fibre already exists, what additional routes are required, what redundancy needs to be constructed, and does the overall project justify that investment?
For sufficiently large data centre developments, those are solvable engineering and commercial questions.
Regional Australia should not just supply the energy
There is an even more important reason to investigate this model: regional economic development.
Australia is likely to invest billions of dollars in the infrastructure required to support artificial intelligence and cloud computing over the coming decades. Where that investment occurs will determine who benefits from it.
A major regional data centre precinct creates considerably more economic activity than simply employing technicians inside a server building. There is civil construction, electrical contracting, mechanical services, fibre installation, solar construction and maintenance, power generation, security, facilities management, engineering, transport and equipment maintenance.
That creates opportunities for local contractors and suppliers, but it also creates something regional communities particularly need: demand for skilled workers. Electrical apprentices, refrigeration mechanics, network technicians, instrumentation specialists, plant operators and engineers all become part of the ecosystem required to keep these facilities operating.
Partnerships with regional TAFEs and training providers could develop pathways directly into those jobs. Once energy and telecommunications infrastructure has been established, it may also attract additional industries that benefit from access to reliable power and high-capacity digital connectivity.
The objective should therefore be bigger than attracting a single data centre. It should be about creating regional technology and energy precincts capable of attracting entire clusters of investment.
Build the ecosystem, not just the data centre
Imagine a regional Queensland technology precinct built around hyperscale and AI computing infrastructure, large-scale solar generation, battery storage, gas generation providing firming and emergency capacity, redundant grid connections, multiple inland fibre routes, LEO satellite connectivity and closed-loop cooling systems designed specifically for the local climate.
Over time, additional technologies could be added. Enhanced geothermal may eventually become commercially attractive in suitable Australian locations. Long-duration batteries could reduce dependence on gas generation. Waste heat could support complementary industrial activity. Additional fibre could be installed as more customers arrive.
Most importantly, the infrastructure begins to reinforce itself. The first major data centre justifies investment in electricity and fibre. That infrastructure makes the location more attractive to the second operator. Increased demand supports more renewable generation and storage. Skilled workers move into the region because there is long-term employment. Suppliers establish local operations because there is sufficient demand.
That is how industrial clusters develop.
There are legitimate challenges
None of this means that Roma, Charleville or any other regional town automatically makes sense for a hyperscale data centre. Transmission capacity, gas availability and pricing, fibre capacity, environmental approvals, water security, cooling performance during extreme heat, bushfire and flood resilience, workforce availability and the logistics of transporting equipment into regional locations all require serious investigation.
Latency also matters. Some workloads need to remain physically close to major users and network interconnection points. The argument is not that every Australian data centre should move inland. It is that not every workload needs to sit in Sydney or Melbourne either.
AEMO's forecasts make clear that data centre electricity demand is becoming material to Australia's entire electricity system. If the industry grows as expected, decisions about where that demand is located will increasingly become decisions about energy infrastructure, not simply property development. (AEMO)
Australia has a choice
The AI boom is going to require enormous amounts of electricity, land and computing infrastructure. We can continue concentrating that demand around cities already struggling to expand electricity, water, housing and transport infrastructure, or we can treat the coming investment cycle as an opportunity to build something strategically different.
Regional Australia has land. Western Queensland has exceptional solar resources. Parts of it sit alongside major existing energy infrastructure. Telecommunications technology and inland fibre are making regional connectivity increasingly achievable. Most importantly, regional communities need long-term investment, skilled employment and industries capable of supporting the next generation.
Google and Fervo's geothermal project in Nevada demonstrates something important about the future of digital infrastructure. Instead of simply searching for somewhere that already has enough electricity for another data centre, companies are beginning to ask how they can build the energy infrastructure required to support the computing.
Australia should apply the same thinking to our own geography. Perhaps the next generation of Australian data centres should not be competing with our capital cities for increasingly constrained resources.
Perhaps we should start building more of the compute where the energy, land and opportunity already are.
