Sustainable by Design: Building AI Infrastructure

The instinct in a boom is to build first and clean up later. For data centres, that order does not work. A facility’s environmental performance is set at the design stage. Its cooling architecture, its power topology, its location relative to renewable generation and its water strategy shape its impact for the whole operating life. Retrofitting efficiency into a building that was never planned for it is expensive, partial and slow. Engineering it in from the start is cheaper, cleaner and increasingly what customers and investors demand.

The technology to do this well already exists. Direct-to-chip and immersion liquid cooling are now becoming standard for high-density AI halls. These cooling systems can cut direct water use by 70 to 90 percent and lift energy efficiency for the dense GPU clusters that high performance computing demands.¹ Power usage effectiveness, or PUE, is the industry’s core efficiency metric. It improves sharply when you design cooling for the workload rather than adapt it afterwards. Pairing facilities with renewable supply, placing them on grids with real clean-energy capacity, and planning for water stewardship are no longer rare. They are fast becoming the baseline expectation for modern data centre infrastructure.

AI is not one workload but many. Training a large generative AI model, running real time inference, and serving enterprise data centre applications each place different demands on power, cooling and space. Data storage is growing just as fast as compute. Every model needs vast, fast storage systems close to the GPUs, and that storage carries its own energy and cooling cost. A sustainable facility plans for all of it together: compute, storage, networking and power as one system rather than separate data centre components bolted on over time.

Demand also rises and falls. Workloads from hyperscalers and public clouds surge and ease, and high-performance computing tenants scale quickly. A design that holds its efficiency across that whole range, not only at full load, is the one that stays sustainable in the real world. None of this is theoretical. Generative AI has pulled demand forward faster than most forecasts expected, and that pressure now lands on real grids, real water systems and real communities. Designing for it honestly, rather than hoping the problem stays small, is the difference between infrastructure that lasts and infrastructure that has to be retrofitted under duress.

This is the environment INSITE DC, an Australian data centre developer, has built its platform for. Our Melbourne data centre campus is engineered for high-density, liquid- and immersion-cooled AI workloads spanning both training and inference, with a scalable electrical architecture and a target PUE of circa 1.3. It is a GPU data centre aligned to next-generation accelerator platforms and designed with modularity at its core so that capacity and cooling can evolve as the workloads do. Our co-design philosophy (aligning with partners on density, cooling and network topology before the scope hardens) is particularly suited to inference customers, whose deployments grow and change continuously rather than landing all at once.

The training era proved what AI could do. The inference era is where that capability meets the real world, billions of queries at a time. For data centre operators, it is the more demanding test (continuous, latency-bound and relentlessly cost-sensitive), and it is the one that will define the next decade of AI infrastructure. The growth engine has changed gears. The infrastructure has to change with it.

Sustainability is not only an engineering question. It is a governance one. Credible environmental performance has to be measured, disclosed and held to account, not just claimed in a brochure. This is the logic behind INSITE DC’s ESG Policy. The policy is built around three priority topics: good governance and ethics; diversity, equity and inclusion; and climate change mitigation, resilience and adaptation. We support the United Nations Principles for Responsible Investment, we screen out activities that clash with our values, and we report on our ESG progress every year, shortcomings included. An ESG Committee oversees the policy and updates it as standards and regulation move.

Our approach favours credible transition over empty gestures. We believe a practical path, engaging openly, measuring honestly and improving step by step, delivers better real-world outcomes than slogans. Environmental responsibility is not a side department at INSITE DC. It is one of the six values encoded in our name. The “E” in INSITE stands for exactly this: running efficiently and sustainably and helping our customers meet their own performance and ESG goals.

The commercial logic is straightforward. The customers driving AI demand are hyperscalers, cloud providers and large enterprises. Many of them run cloud services and cloud computing platforms such as Microsoft Azure and Amazon Web Services on this infrastructure. They carry their own net-zero commitments, and they increasingly choose partners who help them meet those commitments rather than undermine them.⁴ A facility designed for a low PUE, for water efficiency and for renewable supply is better for the grid and the community. It is also more attractive to the very tenants whose workloads make it viable. Sustainability and competitiveness have converged.

Our Melbourne campus reflects this thinking. We are purpose-building it for next-generation AI with liquid and immersion cooling, a target PUE of around 1.3, and a design that treats efficiency and adaptability as first-order requirements rather than later compromises. Every major system (power, cooling, networking and storage) is specified against the same test: does it cut energy, water and carbon over the life of the campus, not just on day one? The aim is simple to state and hard to deliver: meet the explosive demand for AI compute without handing an unacceptable bill to the environment or the communities that host it. In the AI era, that is what responsible data centre infrastructure looks like, and it is the only kind worth building.

INSITE DC is an Australian developer of next-generation AI and hyperscale data centre infrastructure. Our flagship Melbourne campus is engineered from the ground up for high-density GPU compute, liquid and immersion cooling, and a target PUE of circa 1.3, with a development pathway scaling beyond 400MW and a 1GW+ pipeline across Australasia. Built on our values of Insight, Never-Fail Reliability, Service Excellence, Integrity, Trusted Partnership and Environmental Responsibility, we co-design infrastructure with our partners rather than for them.