Gavin Dudley launched his IT career by selling internet services to business customers — a role that offered a front-row seat to the transformative power of emerging technology. Reflecting on those early days, he recalls a conversation that serves as a cautionary tale for businesses today that underestimate the inevitable impact of AI on their future.
“I remember talking to a guy that owned a couple of panels beating stores,” Dudley recalled in a conversation with TechRepublic. “He said to me: ‘I will never need the internet, don’t talk to me about the internet, it will never impact my business.’ Well, I think it’s a bit like that with AI at the moment. And AI is going to impact every business.”
Now VP of sales at Macquarie Data Centres, an Australian data centre company, Dudley said the AI boom is only “just beginning.” He believes that significant amounts of AI will be deployed in the next 10 years. The result? The Australian market, like the rest of the world, is going to have huge data centres that need to be met.
What are the key trends in data centre construction?
As the AI boom gains momentum, the demand for massive computational power is reshaping the infrastructure behind it. Cutting-edge data centres are serving as the backbone of this technological revolution.
In Australia, the need for AI processing is leading to deploying powerful — and hotter — GPUs in data centres. Dudley said the shift in hardware had caused a move toward “fungible” data centre construction, which allows operators to pivot between air and liquid cooling depending on the computing being used.
“We’ve spent some time in the U.S. working with people that are already running genuine AI clusters,” Dudley said. “We’re seeing a [data centre] structure look more like three racks or four racks of high-performance computing, and then up to 12 racks of storage and CPUs.”
SEE: AI Causing Data Centre Power and Cooling Conundrum in Australia
This computing mix creates a blended need for air and liquid cooling. Dudley said data centre halls also need to be designed beyond the 100% cooling capacity of the past, toward 140-150% cooling capacity. He explained that this provides the flexibility of pivoting between air and liquid cooling techniques.
Macquarie Data Centres, which is currently building a new AUD $350 million IC3 Super West data centre in Sydney, is also designing Australian government security protocols, as the company anticipates AI platforms will need to be “severely locked down” to safeguard sovereign AI security and value.
How are customers approaching data centre needs?
Macquarie Data Centres predicts that increasing customer demands for AI will require more capacity from hyperscale cloud providers. However, the high costs of extensive computing in cloud environments may drive more enterprises to invest in their own platforms.
“That’s going to drive both the hyperscale cloud and a new breed, a new round of internal infrastructure that companies will go and buy to implement,” Dudley said. “At present, Macquarie Data Centres is seeing different types of customers with different needs for data centres in Australia”
Hyperscale cloud providers
Hyperscale cloud providers now have their own immense hunger for power and space, driven by the rapid growth of AI applications and cloud services. Their needs are “through the roof,” Dudley said, with some providers in the market now willing to purchase any capacity available in strategic locations.
They are pursuing a blended approach: investing in their own purpose-built facilities, partnering with co-location providers, and commissioning build-to-suit data centers tailored to specifications. Dudley said hyperscalers are now planning five to 10 years in advance to secure the resources they need.
Government agencies
Government clients are currently only “dabbling” in AI, with most federal agencies relying on SaaS and hyperscale cloud services rather than investing heavily in their own infrastructure. However, defense and other large government agencies are now beginning to procure AI capacity to support critical operations.
SEE: AI Surge Could Trigger Global Chip Shortage by 2026, Research Finds
Private sector enterprises
Dudley said many large “stealth projects” are underway in large enterprises in Australia. These projects involve securing infrastructure by engaging with high-performance compute OEMs backed by chip manufacturers.
He said the rest of the local Australian enterprise market is still experimenting with AI and leveraging the hyperscale cloud to do so. This computing is happening at a “super premium rate” as organisations build out strategies for utilising AI in the future.
What role will sovereign AI have in Australia’s future?
Macquarie Data Centres expects the sheer value held in AI models in the future to prompt regulation around protecting this intellectual property in sovereign data centres. Dudley explained that business AI models, as they learn and grow, will become increasingly valuable assets and central to business value.
Australia will also have compelling reasons to develop sovereign AI models locally:
Culture: Reliance on offshore AI models could dilute Australian values, delivering decisions influenced by foreign norms rather than reflecting local societal and cultural principles. A sovereign AI approach will ensure technologies align with Australian community values and support a strong national identity.
AI ecosystem: Sovereign AI could drive productivity, foster local AI development, and position Australia to reap the benefits of innovation in AI, Dudley said. By cultivating in-country AI expertise and infrastructure, Australia could stimulate growth and maintain its competitiveness in the global economy.
SEE: Is it time for Australia to build its own LLM?
Security and governance: Locally hosted AI ensures Australian laws govern sensitive data such as medical records and personal information. This protects intellectual property and establishes clear legal accountability for AI-driven decisions — an essential safeguard for ethical and legal integrity.
Government leadership may be needed to foster a strong sovereign AI ecosystem, Dudley said, including clarification of AI regulatory frameworks and investment to encourage local AI market innovation. To be ready for sovereign AI, data centre infrastructure will also need to be expanded and optimised for a mix of client needs, and Australia’s energy grid will need to be modernised to support sustainable growth.