Australia risks capturing only a fraction of the economic value created by artificial intelligence unless its infrastructure investment is matched by a concerted effort to build local AI capability, according to ResetData co-founder and co-CEO Marcel Zalloua.
Billions of dollars are flowing into the energy and data centre infrastructure needed to support AI, but Mr Zalloua says the bigger opportunity is ensuring Australian businesses, researchers and institutions can use that infrastructure to build valuable intellectual property of their own.
“We are laying the tracks of the new industrial revolution. This is the artificial intelligence revolution,” Mr Zalloua told InnovationAus.com editor Justin Hendry on the latest episode of Resilience through Capability podcast.
But he believes a critical part of that transition is currently being overlooked.
ResetData calls it the ‘missing middle’ – the small and medium-sized enterprises, startups, research organisations and universities that need access to advanced AI infrastructure but often lack the capital and scale to build it themselves.
“How are we actually supporting our local industry enterprises to transition and adopt artificial intelligence?” Mr Zalloua said. “We need to secure what we need for our country first.”
The challenge is particularly acute in research, where institutions are often developing their own infrastructure independently.
AI compute is now so expensive and technically complex that this fragmentation risks becoming a competitive disadvantage, Mr Zalloua argues. A shared platform giving researchers access to the latest technology while keeping their data and intellectual property (IP) within Australia could instead help local institutions compete globally.
The same principle applies across what ResetData describes as the five layers of the AI stack: energy, infrastructure, compute chips, models and applications.
Australia is already attracting enormous investment at the energy and infrastructure layers, but Mr Zalloua believes much of the enduring economic value will be created further up the stack.
“The bottom two layers are probably the least significant in terms of long-term economic returns for the country,” he said.
“We need to find a balance. How do we support the energy and the infrastructure layer, while also supporting businesses that can take us further up the stack to grab that value?”
This distinction also underpins Mr Zalloua’s view of AI sovereignty.
Traditional cloud services largely involve storing and processing existing business data. AI, by comparison, can take an organisation’s IP and use it to generate new intelligence – and potentially new revenue.
For Mr Zalloua, genuine sovereignty therefore goes beyond the physical location of a server. It extends to who operates the infrastructure, what data is stored on it, how that data is accessed and where the resulting information is sent.
This makes access to sovereign compute an increasingly important piece of national infrastructure.
ResetData is already in discussions with governments about ways to reserve capacity that Australian organisations can grow into over time.
Mr Zalloua believes ringfencing a proportion of Australia’s future AI compute for local startups, researchers and government agencies could prove critical as global demand expands.
ResetData’s own AIF1 facility in Melbourne was launched as a sovereign AI factory, with the Australian Institute for Machine Learning and Adelaide University among its first users.
The company has also sought to widen access through a competition offering $1 million worth of compute tokens. The winning research organisation, ProCan, is using AI compute to analyse protein data as part of work aimed at improving the identification of cancer treatments.
For Mr Zalloua, use cases like this demonstrate why the national debate needs to stretch beyond simply how many data centres Australia can build.
“How do we provide a platform for the other companies like ProCan that are solving these really important problems for families?” he said.
Government has an important role to play in creating the conditions for that capability to develop.
Mr Zalloua wants Australia’s emerging Office of AI to take a long-term view of what the country wants AI adoption to achieve, rather than focusing solely on attracting international investment.
Australia’s advantages – available land, energy resources, technical capability and stable institutions – mean global AI investment will continue to seek a foothold here, he argues.
“The market will drive itself and drive investment,” he said. “But we need to be really focused on Australia first.”
He also cautions against assuming Australia has only a two or three-year window in which to secure AI investment.
Demand is already immense, but the technology remains at the beginning of what he expects will be a much longer transformation.
“Yes, it’s a race, so to speak, but the finish line isn’t in three years’ time,” Mr Zalloua said.
A more measured approach would give government an opportunity to balance regulation and risk management with the incentives businesses need to experiment and invest.
This is particularly important in the public sector, where AI projects can struggle to move beyond proofs of concept.
“Solve small problems first,” is Mr Zalloua’s advice.
Rather than attempting organisation-wide transformation immediately, agencies can test AI against defined problems, identify where it delivers genuine value and progressively expand successful applications.
The pace of technological improvement makes experimentation particularly important.
“The AI of today is not going to be as good as the AI of tomorrow,” Mr Zalloua said. “Every month something drops that’s better than the best one in market.”
Looking five years ahead, his preferred outcome is not simply that Australia hosts more AI infrastructure, but that it has developed businesses and research organisations capable of using it to compete internationally.
“What I would like to see is that Australia has taken the opportunity to enhance its own capability locally, be thriving as a country and a net exporter of AI intelligence,” he said.
“What I fear might happen is that in five years we’re still chasing it.”
Avoiding that outcome will require decisions made with a much longer horizon than the next investment announcement, or political cycle.
“This is an opportunity to set the path for Australia for the next 10 or 15 years,” Mr Zalloua said.
Resilience through Capability is an InnovationAus.com podcast series produced in partnership with ResetData.
Do you know more? Contact James Riley via Email.