As demand for AI continues to grow globally, the race to build more data centers is gathering pace. But when a facility needs enormous amounts of electricity, water, and supporting infrastructure, an important question is emerging: Does every location make sense?
This article looks at how the growth of AI is changing the way engineering professionals think about where data centers should be built. While the expansion of data centers signals technological progress, growing concerns around their impact raise another question: if this is the future of technology, why is their development facing growing opposition?
From Bytes to Big Builds
Data centers trace their roots way back in the 1940s, when the United States military’s Electrical Numerator Integrator and Computer (ENIAC) became operational at the University of Pennsylvania. It was far from today’s data center facilities, but the room-sized machine still required a dedicated space, substantial power, and cooling.
As computing evolved from mainframes to client-server systems and cloud computing, the facilities supporting data centers also grew. Server virtualization allowed multiple workloads to run on a single physical server, while high-performance computing (HPC) enabled complex and data-intensive tasks. At the same time, cloud platforms such as Amazon Web Services (AWS), Microsoft Azure, and Google Cloud made computing and storage resources available over the internet. Together, these developments increased the need for larger and more specialized data centers.
Behind this seemingly simple process is a growing need for computing power, with data centers increasingly being used to train Large Language Models (LLMs), support Generative AI, and power AI-driven scientific computing. These workloads require large numbers of high-performance processors, particularly Graphics Processing Units (GPUs) and Tensor Processing Units (TPUs), which generate significant computing and cooling demands. As AI workloads grow, data centers are being designed to accommodate higher power densities and much greater computing capacity.
A Surge in Data Centers
The shift to the physical expansion of data centers is already becoming visible across global markets. Governments and industries are planning new facilities to increase existing capacity and invest in the infrastructure needed to support this growing digital economy. The scale of this development varies by country, but the direction is increasingly clear.
According to the Clean Energy Finance Corporation (CEFC) in its article report “Getting the Balance Right: Data Center Growth and the Energy Transition”, Australia’s operational data center capacity could increase from around 0.3 GW in 2024–25 to between 2.2 GW and 3.2 GW by 2035. As this capacity expands, data centers could account for 8–11% of Australia’s projected electricity consumption in 2035, compared with around 1% in 2024–25.
This growth could also drive up to $135 billion in data center investment. The picture is similar in the United Kingdom, wherein according to the UK Government’s UK Compute Roadmap, the country will need at least 6 GW of AI-capable data center capacity by 2030, representing a threefold increase from the capacity available as of July 2025.
South Africa is also seeing rapid growth in its data center market. According to Mordor Intelligence, the country’s data center market is expected to grow significantly over the next five years with a 16.35% compound annual growth rate (CAGR) between 2026 to 2031, with Johannesburg and Cape Town being the core hubs.
Why Data Center Location Matters
As more facilities are concentrated in these established hubs, however, the question of where to build them becomes increasingly important. A suitable location must offer more than available land. It must also provide reliable electricity, water, connectivity, and supporting infrastructure needed to support cooling and operations. These requirements can make site selection increasingly complex, especially when new facilities place additional pressure on resources already serving local communities. For data center developers, location eventually becomes an engineering decision shaped by the availability of energy, water, connectivity, and other critical resources.
As EIT Lecturer Dr. Hadi Harb notes, “The bottleneck is increasingly becoming the electricity behind all this computing power. Data center locations can be optimized to minimize cooling requirements, but there are also constraints on where such data centers can be based, as distance to users is a consideration that should not be neglected.”
Where Should Data Centers Be Built?
According to the International Energy Agency (IEA), the strongest locations for large-scale data centers are areas with sufficient generation capacity, available grid connection capacity, and robust transmission infrastructure to accommodate the high, continuous loads of AI computing. The United States, for example, benefits from extensive grid infrastructure; China from rapidly expanding renewable generation, while Nordic countries such as Sweden and Norway combine abundant low-carbon electricity with cooler climates that can reduce cooling demand.
However, Dr. Harb points to another potential solution: shifting some AI workloads away from centralized facilities and onto edge devices. As he explains, “Capable machines exist today that can run LLMs, Computer Vision Models, and Automatic Speech Recognition locally.” This also underscores the need for engineering professionals with the skills to design and manage the infrastructure supporting both centralized and edge-based AI systems.
The challenge, then, is not simply building enough capacity to meet AI’s growth. It is by developing engineering judgement and skills to determine what infrastructure is needed, where it is needed, and how it can operate sustainably – and this is where industry-focused engineering education plays a critical role.
Reference:
What is a Data Center
Data Center Growth and the Energy Transition
UK Compute Roadmap
South Africa Data Center Market Size & Share Analysis
Are We Building Data Centers in the Wrong Places
This article was published September 15th, 2026 and the content is current as at the date of publication.
