The digital economy is hungry. Artificial intelligence, the Internet of Things (IoT), and our growing data needs are driving exponential increases in the scale, complexity, and power requirements of data centres (DCs). These facilities are no longer just utility buildings; they are the backbone of the digital world we live in.
With demands projected to double by 2030 and power use already rivalling that of small cities,¹ the challenge for DC developers is no longer just one of technical capacity, but of resilience, sustainability, and speed.
We understand that staying ahead of the curve in this shifting landscape requires more than industry experience, it demands active research into the interplay of demand, location, and consumption – research that enables evidence-based decisions, agile design responses, and alignment with evolving client and regulatory expectations.
Anticipating the future: Demand, design, and decarbonisation
Data centres currently consume about 1% of Australia’s total electricity, around 3 TWh per year.² Globally, the IT sector is expected to use 13% of all electricity in the near future.³ Much of this is driven by generative AI, which requires immense computing power training; ChatGPT-3 alone consumed 1,287 MWh4. DC rack densities are increasing at a compound annual growth rate (CAGR) of 7.8%,⁴ even as facility footprints contract and demand for power infrastructure intensifies.
These shifts are not abstract – they demand tangible design responses. Vertical data centres in land-scarce cities. Campus-style facilities in outer metro zones. Co-located on-site power production. Sophisticated liquid cooling systems are becoming essential to manage escalating thermal loads – NVIDIA’s Blackwell architecture, for instance, mandates direct-to-chip liquid cooling, signalling a major industry shift away from traditional air-cooled systems and redefining the spatial, structural, and mechanical requirements of AI data centre design.⁵
As a multidisciplinary design practice at the forefront of DC development, Hames Sharley engages with these trajectories directly. Our work anticipates the evolution from rack-ready warehouse shells to high-performance, contextually responsive assets whether in dense urban locations or emerging ex-urban tech corridors.
Navigating the energy dilemma: Towards smarter power solutions
The DC sector is increasingly under scrutiny for its energy consumption and carbon footprint. On average, a data centre consumes more than 1000 W/m² over ten times that of a typical office.⁶ In the U.S. alone, DCs consume 70 billion kWh annually, costing $7 billion.¹
Cooling alone accounts for 40–50% of total DC energy use (Bertoldi, 2014), and traditional systems are rapidly becoming obsolete. As a result, many operators are turning to onsite renewables, power purchase agreements, and hybrid cooling methods. Yet despite this, most DCs still rely heavily on brown energy like coal and gas, particularly at peak times, which undermines net-zero claims.¹
This has fuelled global interest in nuclear energy, especially among hyperscalers like Microsoft and Amazon. Amazon’s acquisition of a nuclear-powered DC campus⁷ and Microsoft’s new division dedicated to nuclear energy integration8 underscore this trend. While nuclear remains contentious and currently banned in Australia, the international direction is clear: future-proofing DCs will require a robust and diversified energy mix.
Hames Sharley acknowledges that it is being seriously explored by industry leaders and that our clients must understand its implications. From siting and ownership models to regulatory and safety considerations, nuclear energy represents a R&D frontier we must be prepared to navigate alongside our clients.
A role for architecture: From passive delivery to strategic alignment
Architects must engage proactively with the energy discourse. This doesn’t mean evangelising for nuclear or renewables but understanding the operational and infrastructure implications of each. It means acknowledging that data centres, like healthcare facilities, are mission-critical assets.
At Hames Sharley, our approach is to frame architectural design within a broader ecosystem of performance, regulation, and future viability. We understand that enterprise, hyperscale, and colocation clients each have distinct operational priorities that shape the architectural logic of data centre design. Enterprise clients typically favour accessible, centrally located facilities, while hyperscale operators prioritise remote, large-footprint sites optimised for scalability. Colocation providers, meanwhile, balance high-density capacity with multi-tenant flexibility and stringent uptime requirements. Each typology calls for a tailored design response, from vertically integrated solutions for constrained urban sites to expansive low-rise campuses for regional deployments.
Effective design also requires understanding client operational needs: fibre-optic availability, latency thresholds, modular expansion strategies, and even visitor frequency. We design not only to meet the brief, but we also collaborate to anticipate and shape where the brief is going, ensuring design outcomes align with both present requirements and consider potential futures.
A research-driven practice for an uncertain horizon
At Hames Sharley, research underpins our approach to future-focused design. By investigating areas such as off-grid power segmentation, innovative infrastructure solutions, and next-generation battery technologies, ⁹, ¹⁰ we position ourselves as strategic partners that go beyond traditional service provision to offer foresight and guidance in an increasingly complex landscape.
Our clients seek certainty, not just in construction delivery, but in long-term operational resilience.
Conclusion: Designing for adaptation, not just construction
Australia’s DC sector is forecast to attract over $26 billion in infrastructure investment by 2030,² with projected capacity more than doubling. However, realisation timelines often stretch to three years and clash with the exponential pace of advancement in AI and cloud technologies. This makes foresight not optional, but essential.
Our strategy is simple: remain informed and lead through clarity. While no solution will solve the challenges ahead, understanding how different energy sources like renewables, backup systems, and emerging technologies combine to meet diverse needs is critical. We don’t advocate for what isn’t viable, but we do prepare for what may be inevitable.
As architects of critical infrastructure, our role is not to dictate the future, but to anticipate it. We design for what is needed today, and for what is likely tomorrow to enable our clients and communities to flourish in a landscape shaped by resilience, flexibility, and informed transition.
¹ Bhattacharya, T., Rahgouy, M., Peng, X., Takreeti, T., Cao, T., Mao, J., Das, A., Qin, X. and Sinha, A. (2022). Capping Carbon Emission from Green Data Centers. International Journal of Energy and Environmental Engineering, 14(4), pp.627–41. doi:https://doi.org/10.1007/s40095-022-00539-9.
² Mandala (2024). Empowering Australia’s Digital Future Data Centres: Essential Digital Infrastructure Underpinning Everyday Life. [online] Mandala. Available at: Empowering Australia’s Digital Future Data Centres: Essential digital infrastructure underpinning everyday life [Accessed 29 May 2025].
³ Koomey, J.G. (2008). Worldwide Electricity Used in Data Centers. Environmental Research Letters, 3(3), pp.1–8. doi:https://doi.org/10.1088/1748-9326/3/3/034008.
⁴ JLL (2024a). Growth of AI Creates Unprecedented Demand for Global Data Centers. [online] www.jll.com.au. Available at: https://www.jll.com.au/en/newsroom/growth-of-ai-creates-unprecedented-demand-for-global-data-centers [Accessed 21 May 2024].
⁵ NVIDIA Corporation. (2024, March 18). NVIDIA unveils Blackwell: The world’s most powerful chip for AI. https://blogs.nvidia.com/blog/blackwell/
⁶ Stein, J. (2002). More Efficient Technology Will Ease the Way for Future Data Centers. In: Information and Electronic Technologies: Promises and Pitfalls. ACEEE Summer Study on Energy Efficiency in Buildings.
⁷ Swinhoe, D. (2024). AWS acquires Talen’s nuclear data center campus in Pennsylvania. [online] Data Center Dynamics (DCD). Available at: https://www.datacenterdynamics.com/en/news/aws-acquires-talens-nuclear-data-center-campus-in-pennsylvania/ [Accessed 18 May 2024].
⁸ Moss, S. (2024). Microsoft hires Erin Henderson to head ‘nuclear development acceleration’ for data centers. [online] Data Center Dynamics (DCD). Available at: https://www.datacenterdynamics.com/en/news/microsoft-hires-erin-henderson-to-head-nuclear-development-acceleration-for-data-centers/ [Accessed 18 May 2024].
⁹ Argonne (n.d.). A Novel Radioisotope Battery Made from Nuclear Waste - Chain Reaction Innovations. [online] Argonne. Available at: https://chainreaction.anl.gov/a-novel-radioisotope-battery-made-from-nuclear-waste/ [Accessed 10 Jun. 2025].
¹⁰ Hughes, K. (2024). Nuclear Power in Your pocket? 50-year Battery Innovation | CAS. [online] www.cas.org. Available at: https://www.cas.org/resources/cas-insights/sustainability/nuclear-power-your-pocket-50-year-battery-innovation [Accessed 10 Jun. 2025].
