Kevin Miller, HDR

Kevin Miller is TMT Director (APAC), HDR

As agentic AI looks set to dictate the next wave of the digital transformation

April 10, 2026

The Australian data center market is experiencing unprecedented growth. Demand for artificial intelligence is the key driver for this, with the country’s AI data center sector forecast to reach revenues of US$978 million by 2030, representing a compound annual growth rate of 34 percent.

Agentic AI will inform the next wave of transformation, fundamentally reshaping what data centers need to deliver. Unlike traditional AI workloads, which focus on single-task inference or training, agentic AI involves autonomous systems making decisions, coordinating with other agents, and performing multi-step reasoning. This demands greater compute as well as new interaction patterns, resulting in even higher density, variability, and resilience requirements within data center environments.

– Thinkstock

Navigating the hardware curve

Today’s data center racks are reaching 80-100kW, five to ten times what was considered the norm just a few years ago, and liquid cooling, once an experimental option, is now becoming the default for AI-ready data centers. Agentic AI will accelerate this momentum further, placing new pressures on demand, availability, resilience, and adaptability.

Consequently, for data center owners and operators, the key question has shifted from how to power and cool this generation of chips to how to build adaptive, data-driven platforms and supporting infrastructure that are resilient and capable of evolving together with fast-advancing technology.

Hyperscalers are already preparing for this future. Nvidia’s Vera Rubin GPU will deliver around three times the performance of today’s Blackwell Ultra, while Rubin Ultra, expected by 2027, is predicted to be fourteen times faster.

Harnessing asset data for resilience

Far from a simple, incremental upgrade, each new generation of chip rewrites the thermal and electrical profile of the data hall. Therefore, rather than approaching each refresh as an isolated challenge, operators need to put in place strong, forward-looking strategies that anticipate continuous change and escalation in data center performance requirements. This is crucial in the age of agentic AI, which calls for smarter and more agile centers, and it is in this area that digital transformation plays a major role.

Consultants work in rich model-based environments during design; the data and intelligence embedded in those models are often lost at handover because digital delivery requirements are not specified by most clients throughout the design, construction, and commissioning phases. This means that valuable information is stripped away at the point when it would deliver the biggest benefits.

With agentic AI pushing facilities to operate at the Edge of their design envelope, data center operators need access to accurate and structured data to gain full visibility of their assets. A model-based handover, accompanied by a robust digital strategy outlining how to derive the maximum value from it, is no longer a ‘nice to have’ – it is an essential foundation for resilience.

We are seeing increasing interest in leveraging these data-rich models to add value to their business decision-making.

Digital twin technology as the bridge

We are seeing increasing interest in leveraging these data-rich models to add value to their business decision-making and in digital twins. This trend will continue as clients recognize the benefits of replicating their physical assets virtually using real-time data to keep ahead of the curve and maximize value by optimizing operations, improving efficiencies, and achieving cost savings. Digital twins are emerging as the necessary bridge between agentic AI demand and operators’ ability to adapt.

More than a visual model, digital twins link design intent, as-built detail, commissioning records, and live operational feeds. This enables operators to run scenario analysis ahead of new GPU deployments, stress-test cooling systems before chips land, and optimize energy consumption in real time. The potential advantages are significant, bringing more predictive maintenance, smoother upgrades, and fewer surprises when the next generation of chips arrives. The largest gain is agility, enabling data center infrastructure to adapt in tandem with hyperscalers.

Digital twin models can improve sustainability, too, a vital consideration in today’s landscape where climate change is top of the agenda for the built environment. With the help of machine learning, it is possible to boost operational efficiencies by mapping out facilities in a digital environment without making any changes to equipment.

In fact, research shows that implementing digital twin technologies can result in energy savings of up to 30 percent in a range of building types. In a data center scenario, this can be achieved by tracking information such as natural temperature changes based on locations or assessing specific water temperature input for servers, bearing in mind that next-generation chips are able to take higher temperatures. Adjusting cooling and power settings accordingly can substantially reduce energy consumption and waste, while delivering operational and cost efficiencies.

Australia is in the early stages of this shift, yet the trajectory is clear. Agentic AI will redefine the demands placed on data centers, and digital twins provide the means to meet them. The winners in this market will be those who recognize that the next generation of chips is not just faster – it is rewriting the rules for how facilities must be designed, delivered, and operated.