Australia's appetite for compute is accelerating. The bigger question is whether our energy, infrastructure and skills can scale with it.
Australia's High-Performance Computing landscape is entering a very different phase.
Demand for advanced compute is growing across scientific research, artificial intelligence, engineering, defence, climate modelling, healthcare and increasingly data-intensive enterprise workloads.
At the same time, AI is reshaping the physical infrastructure required to provide that compute.
More GPUs mean greater rack density. Greater density means more power. More power creates greater cooling requirements.
And the scale of the infrastructure now being proposed means HPC and AI can no longer be discussed purely as a technology issue.
It is becoming an energy, infrastructure and workforce issue.
How do we actually power the next generation of HPC and AI infrastructure?
One of the clearest indicators comes from Australia's existing national supercomputing infrastructure.
The National Computational Infrastructure (NCI) and Pawsey Supercomputing Research Centre operate Australia's two Tier-1 HPC facilities, Gadi and Setonix.
Demand for access has surged.
Compute hours requested by Australian researchers through the 2026 national allocation scheme.
Increase in requested HPC resources between the 2023 and 2026 allocation rounds.
Increase in requests for Setonix GPU capacity while available GPU allocation remained flat.
Those figures represent publicly funded research computing rather than the entire Australian commercial HPC market.
But they demonstrate the direction clearly.
Australia wants more compute than its existing national infrastructure can currently provide.
And AI is adding an entirely new source of demand.
Sources: NCI and NCMAS 2026 Outcomes.
The HPC conversation can no longer be separated from Australia's rapidly expanding data-centre and AI infrastructure market.
The scale of investment now underway is substantial.
Announced in 2026 to accelerate AI-ready infrastructure following significant contracted growth.
CDC described the agreement as the largest data-centre contract in Australian history.
Hyperscale campuses across Sydney and Melbourne form part of Australia's expanding digital infrastructure layer.
IC3 Super West is designed for dense AI and cloud workloads using air, liquid and hybrid cooling.
These aren't isolated infrastructure projects.
Together they point towards something much larger:
Australia is building the physical layer required for an increasingly compute-intensive economy.
This is where the HPC story becomes an energy story.
At the end of March 2026, the Australian Energy Market Operator had 11 large-scale data-centre projects representing 5.4GW of maximum demand progressing through the transmission connection process.
Approximately 60% of that capacity was located in New South Wales and 40% in Victoria.
For context, average data-centre demand across the National Electricity Market during the first quarter of 2026 was around 600MW.
The pipeline therefore represents a potentially enormous increase in demand.
Forecast data-centre electricity consumption in the NEM during 2025–26.
Forecast annual consumption by 2035–36.
Potential share of electricity supplied through the grid consumed by data centres by 2035–36.
For HPC and AI infrastructure operators, access to compute may therefore increasingly depend on something considerably less glamorous:
Access to electrons.
Source: AEMO — Digital Demand Surge.
Historically, organisations could think about computing infrastructure largely through the lens of technology.
Servers. Storage. Networks. Software.
AI and modern HPC environments make the equation more interconnected.
Each layer potentially constrains the next.
There is little value in securing thousands of GPUs if sufficient power isn't available.
Power alone doesn't solve the thermal requirements of increasingly dense racks.
A world-class facility doesn't guarantee adequate high-performance networking.
And none of the above creates business value without people capable of architecting, operating and optimising the environment.
That is why the next generation of HPC infrastructure can't simply be treated as an IT procurement exercise.
The scale of the infrastructure being proposed has caught the attention of government.
In 2026, the Australian Government established national expectations for data-centre and AI infrastructure development.
The Government subsequently announced plans for Australian Standards for AI that would go further.
Under the proposed framework, large data centres would be expected to underwrite new power supply, contribute to connection costs and support grid stability.
The future of AI infrastructure isn't solely about attracting investment. It is about ensuring Australia can physically support that investment.
We're already beginning to see what that future might look like.
Google, AirTrunk and European Energy Australia partnered on the 25MW Mulwala Solar Farm in regional NSW, with the project designed to add new renewable capacity to the National Electricity Market while supporting digital infrastructure demand.
AirTrunk is also exploring battery energy storage in Australia as part of its longer-term strategy around grid flexibility and carbon-free energy.
CDC reports that electricity consumed across its facilities during FY26 was matched with renewable electricity, while its infrastructure is increasingly being designed around high-density and liquid-cooled AI and supercomputing workloads.
And the relationship between compute and energy is likely to become even tighter.
Instead of asking:
“Where should we build the data centre?”
The question increasingly becomes:
“Where can we access power, connectivity, water or appropriate cooling, land and technical capability at the scale the workload requires?”
That could materially influence where Australia's future HPC clusters develop.
The market is becoming broader than the traditional definition of an HPC provider.
At the national research layer, NCI and Pawsey remain critical sovereign infrastructure, supporting thousands of users across universities, research institutions and government agencies.
Around that national infrastructure sits a rapidly expanding commercial ecosystem involving data-centre operators, hyperscalers, semiconductor companies, energy providers, networking vendors and AI infrastructure specialists.
One example is Project Southgate, bringing together CDC, Firmus Technologies and NVIDIA around sovereign, renewable-powered AI infrastructure in Australia.
The HPC ecosystem is therefore no longer a neat category.
It is becoming a convergence of multiple industries and infrastructure disciplines.
Australia can approve another data centre.
An organisation can procure more GPUs.
Capital can fund another infrastructure project.
But specialist expertise cannot necessarily be manufactured at the same speed.
The capability required around modern HPC and AI infrastructure crosses numerous disciplines:
And increasingly, organisations need people capable of operating across those domains.
This is an important distinction.
Australia's future HPC skills problem may not simply be a shortage of people carrying an “HPC Engineer” title.
It could be a shortage of people who understand how increasingly complex compute, infrastructure and AI environments fit together.
The debate around Australia's HPC and AI future will understandably focus on infrastructure.
How many GPUs? How many megawatts? How much renewable generation? How many data centres? How much sovereign compute?
They're all important questions.
But there is another capacity metric that deserves the same attention:
The processing infrastructure required to support increasingly complex workloads.
The generation, transmission and resilience required to power that infrastructure.
The specialist skills required to architect, operate, secure and optimise it.
If one significantly outpaces the others, the entire system becomes constrained.
Australia has many of the ingredients required to participate meaningfully in the next generation of global computing.
Strong research institutions. Existing sovereign HPC capability. Growing renewable generation. A rapidly expanding data-centre sector. Increasing AI investment. And demand that is clearly already present.
The opportunity is significant.
But the winners won't simply be the organisations that secure the most compute.
They will be the organisations capable of bringing together compute, energy, infrastructure and specialist talent — and turning all four into productive capability.
Buying infrastructure creates capacity. Knowing what to do with it creates value.
At Logical, we're tracking the specialist talent markets developing around HPC and AI infrastructure — from HPC engineering and GPU infrastructure through to cloud, networking, data, cybersecurity and platform engineering.
As Australia's compute infrastructure scales, our focus is on understanding where specialist capability exists, where skills can transfer from adjacent markets, and where workforce constraints could become barriers to infrastructure investment.
Because Australia's next HPC challenge won't simply be building more compute.
It will be building the capability to power, operate and optimise it.
Talk to Logical →