Compute Infrastructure Is a Design Problem, Not Just a Capital Problem

Sumit Mukhija from DCI Data Centers opened the morning with a framing that set the tone for everything that followed. The goal is not simply to build compute capacity. It is to stay relevant over a long period of time. Utilisation is the metric that matters, not megawatts alone. In a market where silicon generations are turning over annually and rack densities are pushing well beyond 25kW, the infrastructure built today must absorb what comes next without full replacement.

Tony Gaunt from Vertiv extended that thinking in practical terms. The core challenge is not liquid cooling versus air cooling. It is designing a facility that is not electrically obsolete before it opens. High-voltage DC is gaining traction as rack densities increase. BIM modelling complexity is growing. The annual pace of change from the silicon layer is compressing the design window in ways the industry has not previously had to manage. Australia's two decades of delivery experience is a genuine advantage here, but not an automatic one. Gaunt also raised a point that resonated through multiple sessions on the day: the industry needs more young people willing to challenge established approaches and bring fresh thinking to problems that have been solved the same way for too long.

Investment Conditions Are Shifting in Australia's Favour

The mid-morning investment session identified cloud loads, AI workloads, and the emerging neo-cloud segment as the three demand vectors pulling capital into Australia. Speculative development has struggled to keep pace, and near-term capacity constraints remain real. Power has extended development timelines significantly and raised the barrier for new entrants, but that same barrier is what makes Australian assets attractive to investors who seek markets where supply cannot be quickly replicated.

The structural shift attracting the most attention was the move away from single-owner development models. The separation of property ownership from operations is opening the market to new capital structures, allowing different investor profiles to access different parts of the stack. The project finance market was described as highly liquid. Friction points including land tax in Victoria, potential foreign exchange impact, and development economics that do not always maximise onshore returns were raised. The consensus, nonetheless, was that Australia remains a net positive destination for global data centre capital.

The Industry's Narrative Gap Is a Strategic Liability

The sharpest take of the morning came from Belinda Dennett, CEO of Data Centres Australia. Her argument was direct: the industry has a narrative problem, and it is costing Australia real ground in a global competition where capital is highly mobile.

The most damaging claim in public circulation is that data centres drive up household energy bills. The evidence does not support it. Australian data centres source 70% of their energy from renewables, account for just 0.04% of national water use, and pay their full share of connection costs, network tariffs, and upstream transmission augmentation. The sector contributes $12.6 billion in gross value added per terawatt hour, ahead of mining and manufacturing. With a $142 billion annual opportunity identified for Australia against $680 billion in forecast global hyperscale spend for 2026, and 8,300 new roles projected by 2030, the stakes are significant. Australia currently operates 252 data centres at 1.4GW of capacity. The United States operates approximately 5,400 facilities at 53.7GW. That gap is the growth opportunity.

Dennett's call to the room was to stop importing American headlines into Australian policy discussions. The community awareness gap is widening. Public coverage frames data centres as threats to energy affordability and water supply while the industry sees world-class infrastructure credentials and a generational investment opportunity. Closing that gap requires the sector to communicate its contribution accurately and consistently.

Planning Reform Remains the Sector's Most Pressing Constraint

The panel on unlocking land and navigating planning brought together Belinda Dennett, Nauman Akhtar from InSite DC, Fredrik Johansson from Balder Investment, Michael Gold from Macquarie Technology Group, and Jennifer Cooper from Urbis as moderator. Preserving industrial land in Sydney was identified as an active and worsening constraint. Securing power and planning approvals in parallel rather than in sequence was described as the only delivery model that meets hyperscale customer timelines. AEMO is tightening grid connection standards in response to AI load requirements, adding further complexity to projects already under timeline pressure.

Akhtar argued for a national body to standardise planning processes across jurisdictions. Dennett disagreed, arguing that an additional institutional layer risks slowing the system further. Her position was that the planning framework needs to become adaptive rather than linear, capable of adjusting as project conditions evolve. The panel noted that Malaysia has moved decisively to build an integrated ecosystem around data centre development, and that regulatory certainty is what global operators consistently cite when making location decisions.

Day Two Takeaway

The Summit closed with a clear statement of what Australia's data centre sector needs to be over the next five years: a credible, sovereign, AI-capable infrastructure platform that the rest of the world chooses to connect to. The technical foundations are largely in place. Capital is available. Demand is structural. What remains is execution on planning reform, community narrative, workforce development, and infrastructure design that holds its value as compute requirements continue to evolve.

See you next year!