Data centres will use enormous amounts of electricity. Some can use enormous amounts of water. They can compete for trades, occupy large sites and, once built, employ fewer people than their extraordinary capital cost might suggest.
Those are not reasons to stop talking about them.
They are reasons for Australia to demand something much better.
Australia is having a very important argument about data centres.
Unfortunately, parts of that argument are already becoming too simple.
On one side are billions of dollars of investment, artificial intelligence, cloud computing, jobs and the promise that Australia could become a major digital economy.
On the other are very legitimate questions.
Who provides all the electricity?
Will households pay more for it?
Where does the water come from?
What happens during drought?
Why should an enormous foreign-owned facility be given access to scarce Australian infrastructure if relatively few people eventually work inside it?
And what, exactly, does the Australian public get in return?
They are excellent questions.
In fact, Australia should be asking them before approving the next generation of projects.
But there is another question that deserves just as much attention:
What does Australia lose if we decide this infrastructure is simply too difficult and allow the next generation of computing to be built somewhere else?
That question takes us well beyond the idea of a conventional data centre.
At JtA, we think Australia should start talking instead about digital infrastructure campuses.
Because the opportunity is much bigger than buildings full of computers.
First, the uncomfortable part: the critics are right about quite a lot
There is no useful case for Australia’s data-centre future that begins by pretending these facilities do not consume enormous quantities of electricity.
They do.
AEMO’s new 2026 Electricity Statement of Opportunities forecasts electricity consumed by data centres in the National Electricity Market rising from about 5 terawatt-hours in 2025–26 to 34 TWh in 2035–36.
That means moving from roughly 3 per cent to 13 per cent of electricity supplied through the grid within a decade.
The development pipeline is also extraordinary.
AEMO has now considered the demands of 225 known data-centre developments seeking to connect to the NEM, more than double the 97 projects being considered a year earlier. Not all will be built, and AEMO has explicitly improved its forecasting to account for project attrition, gradual ramp-up and the difference between nominal connection capacity and actual consumption.
So when somebody says the grid implications are serious, they are correct.
Water is serious too.
Some traditional cooling systems can consume very large volumes, particularly when concentrated in already water-constrained cities. The Climate Council points to Australian utilities receiving individual data-centre enquiries requiring as much as 40 million litres per day. It argues that unchecked growth could place real pressure on water supplies during a hotter and more drought-prone future.
And then there are jobs.
This deserves more honesty than it sometimes receives.
A multi-billion-dollar data centre does not become a multi-billion-dollar permanent jobs factory.
The Reserve Bank noted this month that data-centre investment is already materially boosting Australian business investment, but also observed that much of the equipment fit-out is imported, limiting the immediate addition to domestic GDP.
The ABS makes an equally important distinction. Data centres are extremely capital intensive, so direct productivity per worker can be very high, but their relatively small employment share means their largest economy-wide productivity impact will depend on what the rest of the economy does with the cloud computing, AI and data services that infrastructure enables.
That is exactly the point.
We should stop selling data centres primarily as job factories. They are infrastructure.
We don’t judge a fibre-optic cable by how many people are permanently employed standing beside it.
We judge it by what people can do because it exists.
A data centre is a building. A digital infrastructure campus could be much more.
Imagine the proposition differently.
Instead of approving:
a large industrial building + servers + grid connection
Australia could demand:
compute + new energy + storage + grid services + advanced cooling + recycled water + fibre + skills + research capability + community infrastructure.
That is a fundamentally different project.
It is a digital infrastructure campus.
And once we look at it that way, the conversation changes.
The electricity requirement stops being only a problem and can become part of the investment proposition.
The enormous demand for fibre can extend connectivity.
The need for water can finance recycled-water infrastructure rather than consume potable supplies.
The concentration of sophisticated electrical, mechanical, cooling and computing systems can become a training ground for Australian technicians and engineers.
And the GPUs themselves can become infrastructure available not only to global technology companies, but to Australian researchers, start-ups, universities, not-for-profits and businesses.
That last point matters enormously.
The Australian Government’s own Data Centre Expectations now explicitly say providers of large-scale compute should make computing power available to Australian start-ups, innovative small businesses, researchers and not-for-profits on favourable terms, while deploying engineers and researchers in Australia and developing local capability.
That should not be a footnote.
It should be central to the deal.
What could Australians actually get?
Consider electricity first.
The obvious fear is that an enormous new electricity consumer arrives, connects to a constrained network, bids aggressively for existing generation and leaves everybody else paying more.
That is a perfectly rational thing to oppose.
But there is another model.
Require major new campuses to bring enough additional generation and firming capacity to substantially support their own demand.
Require them to pay the appropriate cost of their network connections and necessary augmentation.
Require batteries.
Require the ability to reduce or shift suitable workloads during constrained periods.
And reward campuses that can actually support the grid.
The Commonwealth’s current expectations already move strongly in this direction: additional clean generation and/or storage, appropriate cost recovery for transmission and distribution, energy efficiency, demand flexibility and peak-load management. The government has said it intends to turn key elements into mandatory standards.
There is currently a political disagreement over precisely what forms of energy should qualify.
The Commonwealth wants large new facilities tied to additional renewable electricity, with firming, while Queensland and the Northern Territory are arguing for a more technology-neutral approach that can include their existing and developing energy resources.
That debate will continue.
But underneath the political disagreement is a principle on which Australia should be able to find considerable common ground:
A new industrial load measured in hundreds of megawatts should not simply send the bill for its supporting infrastructure to everybody else.
Bring power.
Bring storage.
Pay for the infrastructure you require.
Be flexible when the system needs it.
And demonstrate that the Australian energy system is stronger with you connected than it would have been without you.
That is a much more useful test than whether somebody is simply “pro” or “anti” data centre.
Queensland already has a glimpse of what this can look like
At Brendale, north of Brisbane, Quinbrook’s Supernode project provides perhaps Australia’s clearest early illustration of the idea.
Supernode is being developed beside the South Pine high-voltage switchyard as an integrated energy and digital infrastructure precinct.
Its masterplan includes up to 500 MW of potential IT data-centre capacity, three high-voltage connections and a battery complex planned to reach approximately 780 MW / 3,074 MWh.
Stages 1 and 2 of the battery are now operating, providing a combined 520 MW / 1,858 MWh, with further development underway.
That architecture is important.
The battery isn’t merely there to keep servers alive for a few minutes during an outage.
It is major energy infrastructure capable of interacting with the broader electricity market.
That starts to look less like:
computers consuming Queensland’s electricity
and more like:
a new energy-and-compute precinct participating in Queensland’s electricity system.
The Sunshine Coast presents another version of the idea.
NEXTDC’s SC1 at Maroochydore hosts the Sunshine Coast International Broadband Network cable landing station and connects into the 7,000-kilometre Japan–Guam–Australia South submarine cable. NEXTDC is now developing the larger, AI-ready SC2, planned for about 6 MW of IT capacity.
Then in 2025 the new Tabua submarine cable landed at Maroochydore, creating another international connection through the Pacific towards the United States.
Suddenly a regional Queensland city is not merely consuming internet from Sydney.
It is part of the physical international infrastructure of the internet.
That can change what businesses can locate there.
It can change latency.
It can improve route diversity and resilience.
It can make regional Queensland a place where digital services are produced rather than simply consumed.
That is what infrastructure does.
And then there is the Western Downs
The newly lodged Western Downs Digital Park proposal near Kogan and Dalby takes the concept to an entirely different scale.
It is still a proposal — not an approved project — and some of the enormous headline capacity and investment numbers being reported vary depending on whether the immediate development or ultimate campus potential is being discussed.
That alone is a reason for careful assessment rather than breathless acceptance.
But the location demonstrates something important.
Why force every major AI computing facility into Sydney or Melbourne if some computational workloads do not need to be beside the end user?
Training an AI model can be very different from serving a real-time request.
Large training runs and some batch workloads can potentially be performed where land, power and supporting infrastructure make most sense.
The Western Downs has enormous energy infrastructure, existing generation, transmission corridors and space.
Its newest major data-centre proposal is consequently being designed around the energy landscape rather than trying to shoehorn a huge new load into an already congested metropolitan precinct.
Whether this particular proposal ultimately proves to be the right one is for the planning, energy, environmental and community processes to determine.
But the larger idea deserves attention:
Could regional Australia convert one of its great advantages — energy — into a much higher-value digital export?
Australia has spent generations exporting energy. What if we also exported compute?
This may be one of the most interesting parts of the opportunity.
Australia exports coal.
We export LNG.
We export minerals.
Increasingly, we will export embodied renewable energy through processed minerals, metals and new industrial products.
Compute potentially introduces another category.
Electricity enters a campus.
What leaves is not a tonne of material.
It is computation.
Model training.
Scientific processing.
Cloud services.
Simulation.
Inference.
Digital products.
In economic terms, Australia has potentially transformed electrons into a high-value service that can be sold around the region.
It is effectively an energy-intensive export industry without a ship leaving the port.
That possibility is not uncontested and the projected gains vary significantly.
Deloitte Access Economics — in research commissioned by Google — has modelled a scenario in which Australia becomes an Asia-Pacific digital infrastructure hub and estimates up to $134 billion of additional GDP over 25 years. McKinsey has modelled still larger potential outcomes under more aggressive infrastructure scenarios. These are scenarios, not guarantees, and should be read as such.
A newly released study commissioned by Data Centres Australia similarly estimates substantial construction and ongoing economic contributions and argues that latency-tolerant AI workloads could create a compute-export industry. Again, it is industry-backed modelling rather than an independent promise of future wealth.
But the global competition itself is unquestionably real.
Asia-Pacific currently has a 26.5 GW data-centre development pipeline. Johor, Bangkok, Mumbai and Jakarta are expanding quickly, while Australia, India, Japan and Malaysia are expected to push beyond 2 GW of operational capacity by 2028.
If Australia decides not to host that infrastructure, it does not mean the infrastructure disappears.
Malaysia will build it.
Thailand will build it.
India will build it.
The United States will build it.
Someone will build it.
And Australia can purchase the resulting computing services from them.
That is an entirely legitimate strategic choice.
But it is a choice.
“We can just use overseas cloud” is true — up to a point
Australia will not suddenly lose access to artificial intelligence if it doesn’t build thousands of megawatts of local compute.
Undersea fibre works.
Cloud computing already crosses borders.
Many Australian workloads can perfectly reasonably run elsewhere.
But not all computing infrastructure is interchangeable.
Defence provides the clearest example.
The Commonwealth is investing at least $2 billion in a sovereign Top Secret Cloud built in Australia for Defence and the National Intelligence Community, specifically so highly classified information can be stored, analysed and processed securely at scale while supporting AI and machine learning.
Australia’s Defence supercomputer is similarly described as a secure sovereign capability enabling Australian researchers to tackle sensitive problems locally.
So local compute is not an abstract technology-industry wish.
At a certain level it becomes part of national capability.
The same principle, at different security levels, matters for government, health information, financial services, critical infrastructure, research, commercially sensitive intellectual property and Australian businesses that simply want more control over where their information is processed.
Sovereignty does not mean every byte of Australian data must remain in Australia.
It means Australia retains meaningful choices.
And choices require infrastructure.
The biggest benefit may not happen inside the campus at all
This is where the jobs criticism needs reframing.
Industry-commissioned research released this week estimates Australia’s existing 1.6 GW of operational data-centre capacity supports around 9,450 ongoing jobs across the economy, with expansion potentially supporting more than 23,000 ongoing jobs by 2030.
Those are worthwhile jobs.
Construction is labour intensive too.
But if Australia builds these facilities only because it wants people maintaining servers and switchboards, we have misunderstood their economic purpose.
The real prize is what happens outside the fence.
The Productivity Commission recently concluded that AI will likely lift Australian productivity, while stressing that estimates of the size of the dividend vary widely. Its assessment is that economy-wide multifactor productivity gains above 2.3 per cent over the coming decade are likely.
That does not mean a new data centre automatically creates productivity.
It doesn’t.
Which means Australia needs both halves of the equation:
compute infrastructure + Australians able to use it.
- A regional manufacturer using AI to find defects earlier.
- A farmer using models to combine weather, soil, satellite and crop information.
- A radiologist getting better decision support.
- A university researcher running simulations that were previously impractical.
- A small Australian technology business training a specialised model without needing enormous upfront computing infrastructure.
- A council modelling floods.
- A logistics company reducing fuel and wasted kilometres.
- A teacher producing genuinely adaptive learning support.
- A scientist analysing huge biological datasets.
- A Queensland start-up developing a product for the world.
Those things are the economic case.
The campus merely enables them.
This is also why JtAi exists
There is a slightly unusual element to this story.
JtAi helped build it.
This article began as a human question: Australia is seeing a wave of data-centre applications — what could an intelligent version of that future actually look like?
That question became an iterative collaboration.
JtAi helped range across AEMO forecasting, ABS economic data, Productivity Commission work, International Energy Agency analysis, Queensland planning, operator announcements, critical commentary, government policy, regional infrastructure and the competing economic claims being made by industry.
Then came the more important part:
challenging the claims, separating proposals from approvals, distinguishing industry modelling from official statistics, finding the uncomfortable evidence, and applying human judgement to the argument.
That is the model behind JtAintelligence — JtAi.
JtAi isn’t intended to be an oracle and it isn’t simply a chatbot generating paragraphs. It combines artificial intelligence, human experience, structured knowledge, workflows and real-world judgement: AI where it helps, humans where judgement, responsibility and context matter.
This article is a small example of why local access to increasingly capable computing matters.
Today, systems like JtAi can help a relatively small organisation investigate an issue at a depth and speed that would once have required vastly more resources.
Extend that capability across Australian small businesses, research organisations, farms, hospitals, engineering firms, manufacturers, councils and community organisations and the infrastructure conversation becomes very different.
The servers are not the final product.
Human capability amplified by them is.
But we should not give the industry a blank cheque
Being optimistic about digital infrastructure does not require Australia to accept every proposal.
Quite the opposite.
Australia now has bargaining power precisely because developers want to build here.
- We have land.
- We have political and legal stability.
- We sit in the Indo-Pacific.
- We have direct submarine connectivity.
- We have enormous energy resources and exceptional renewable potential.
- We have advanced research institutions.
- And Australia remains an unusually attractive place for long-life infrastructure investment.
The National AI Plan notes that Australia attracted about $10 billion of data-centre investment in 2024, ranking second globally for the asset class that year, and that more than $100 billion of potential Australian data-centre investment was announced between 2023 and 2025.
That means we do not need to beg.
We can negotiate.
An Australian Digital Infrastructure Compact
Here is where the national debate could go.
Instead of asking whether Australia is “for” or “against” data centres, establish a clear bargain for very large digital infrastructure campuses.
1. Bring infrastructure with you
Large campuses should demonstrate the additional electricity generation, storage, transmission and firming required to accommodate them without unfairly transferring costs to existing customers.
2. Become part of the grid, not merely a load on it
Where technically possible, campuses should provide batteries, demand flexibility and grid-support services.
The International Energy Agency says growing large loads such as data centres increase the need for flexibility, while battery storage can help integrate them securely and cost-effectively.
Some AI work does not have to run at the exact second it is requested.
There should eventually be an economic incentive for flexible computational workloads to chase abundant electricity rather than requiring electricity to chase them.
3. Don’t drink the town dry
Best-practice closed-loop cooling, rainwater harvesting and recycled or treated water should become the starting point.
Potable water should be the exception where practical alternatives exist, not the lazy default.
And major proponents should pay for the water infrastructure they require.
4. Give Australia access to the compute
A portion of strategically significant computing capacity should be accessible on favourable terms to Australian universities, researchers, start-ups, not-for-profits and innovative businesses.
The Commonwealth has already put this principle into its expectations.
It should become measurable.
5. Build Australians alongside the buildings
Major campuses should carry apprenticeship, training and Australian workforce commitments proportional to their scale.
This matters because there is a legitimate competing-demand problem: data centres, housing, renewable energy, defence and infrastructure are all chasing electricians and other skilled workers.
The answer cannot simply be to outbid the housing sector for the people we already have.
Grow the workforce.
6. Leave useful infrastructure outside the fence
Where a campus requires new fibre, roads, substations, water recycling or telecommunications infrastructure, governments should look for opportunities to design those investments so they create wider community or industrial benefit.
A digital infrastructure campus should improve its precinct.
7. Earn community consent
The larger the project, the less acceptable it is to treat community consultation as a planning technicality.
Noise, generators, cooling equipment, visual impact, traffic, biodiversity, water, fire risk and local infrastructure all matter.
Queensland is proposing social-impact assessments and community-benefit agreements for major data-centre developments — an idea worth serious consideration regardless of where one sits in the broader energy policy argument.
8. Report what you use and what you return
- Electricity.
- Water.
- Emissions.
- Storage.
- Demand-response performance.
- Australian jobs.
- Apprentices.
- Local procurement.
- Compute made available to the Australian innovation ecosystem.
Major infrastructure receiving privileged access to scarce national resources should be able to demonstrate its side of the bargain.
What happens if Australia gets this wrong?
There are two ways to get it badly wrong.
The first is to approve everything.
Build wherever somebody owns land.
Allow enormous new loads to compete for existing electricity.
Let shared infrastructure costs wash through to everyone else.
Accept potable-water consumption because it is easiest.
Announce thousands of “jobs” without explaining how many survive construction.
Call every project sovereign simply because the building has an Australian postcode.
That would rapidly destroy the industry’s social licence.
The backlash now occurring overseas shows how quickly that can happen. The IEA notes that grid constraints and connection backlogs are already threatening to delay around 20 per cent of planned global data-centre capacity to 2030.
But Australia can make the opposite mistake too.
We can become so frightened of the demand created by the next industrial platform that we effectively tell global and Australian investors:
build it somewhere else.
If that happens, Australians will probably still use AI.
Our businesses will still buy cloud services.
Our researchers will still need compute.
Government will still digitise.
We will simply import more of that capability.
The capital investment occurs elsewhere.
The engineers cluster elsewhere.
The research partnerships happen elsewhere.
The power infrastructure is financed elsewhere.
The surrounding ecosystem develops elsewhere.
The regional fibre is built elsewhere.
And the next Australian technology company may discover that its most important infrastructure — and increasingly its most important industrial input — sits in another jurisdiction.
We will have avoided the difficult infrastructure problem.
And potentially surrendered much of the opportunity with it.
This shouldn’t be a choice between technology and Australians
The debate now reaching National Cabinet is often framed as a contest.
- AI versus electricity prices.
- Data centres versus water.
- Investment versus communities.
- Renewables versus other energy technologies.
- Canberra versus Queensland.
Those conflicts are real, but they do not have to define the end result.
The more ambitious question is:
How do we make this wave of investment build things Australians needed anyway?
- More generation.
- More storage.
- Stronger grids.
- Better water recycling.
- More resilient fibre.
- More apprentices.
- More engineers.
- More research capability.
- More sovereign computing capacity.
- Better regional connectivity.
- And access to computational power that allows Australian people and businesses to do things that previously belonged only to the world’s largest organisations.
If a proposed campus cannot contribute meaningfully to that bargain, Australia should be perfectly comfortable saying no.
But when one can?
We should be capable of saying yes — quickly, confidently and with ambition.
Because the countries competing with Australia are not deciding whether the digital economy exists.
They are deciding where its physical infrastructure will live.
The cloud was always on the ground
Perhaps the strangest thing about the digital age is the language we chose for it.
We talk about the cloud as though our information floats invisibly somewhere above us.
It never did.
The cloud is concrete.
Copper.
Glass fibre.
Transformers.
Substations.
Cooling systems.
Water pipes.
Batteries.
Generators.
Chips.
People.
Land.
And enormous amounts of electricity.
Artificial intelligence makes that physical reality impossible to ignore.
That is not necessarily a weakness.
For a country like Australia, it could be an extraordinary opportunity.
We have spent more than a century building industries around the resources beneath our feet and the energy available across this continent.
The next great infrastructure question may be what happens when we connect those advantages to computation.
Not simply building more data centres.
Building digital infrastructure campuses that generate value well beyond their fences.
Campuses that strengthen the systems they connect to.
Campuses that help Australians create things.
Campuses that carry infrastructure, knowledge and opportunity with them.
Campuses built on Australian terms.
The future will require compute somewhere.
The real decision confronting Australia is whether we want simply to consume that future —
or help build it.