Concerns about resource depletion are not new to economists. From fears of “peak oil” to anxieties over food shortages, history has shown that as demand surges, supply mechanisms adapt accordingly. Today, a similar narrative is emerging around data centres. A closer examination, however, reveals that, despite burgeoning demands, the industry is not approaching a resource-constrained peak. Instead, data centre infrastructure is evolving dynamically to meet increasing needs, driven by advancements in artificial intelligence (AI), cloud computing, and the expansion strategies of hyperscale companies.
Daniele Viappiani, Portfolio Manager at GC1 Ventures, explores this further:
Data centres are intricate ecosystems requiring stable electricity, advanced cooling systems, robust connectivity, and stringent security measures. While, theoretically, they can be constructed anywhere, optimal locations are chosen based on factors such as minimal natural disaster risk, proximity to essential infrastructure, and favourable environmental conditions. The industry is witnessing the emergence of massive, highly sophisticated data centres alongside smaller, modular facilities. These two complementary approaches combine to suitably address growing demand.
Modular data centres are prefabricated units that offer rapid deployment and scalability, addressing the need for swift expansion. According to reports, the global market for modular data centres is projected to grow significantly, reaching $93.3 billion (£68.6 billion) by 2030, driven by the demand for plug-and-play solutions that can be deployed quickly to meet immediate needs.
The complexity of data centres requires highly specialised labour, including engineers, electricians, and network experts, who are currently in high demand. Constructing large-scale data centres is a capital-intensive endeavour with lengthy lead times. While smaller facilities may take months, large centres can require up to three years to become operational.
Despite this, existing data centres often operate below full capacity, allowing for short-term scaling. Many facilities are intentionally over-provisioned to manage utilisation peaks and accommodate growth, often running at under 50% utilisation. This design allows for the addition of servers or workloads within hours or days, provided the physical infrastructure supports it. Upgrading older servers to more efficient models can further enhance capacity, though limitations are primarily dictated by electrical and cooling infrastructures.
Rapid expansion faces obstacles such as power grid limitations and zoning regulations, particularly in urban areas. The surge in electricity demand from data centres has led utilities to grapple with unprecedented power requests. For instance, Oncor Electric in Texas, USA, received requests totalling 119 gigawatts, far exceeding its current capacity. Utilities are responding by increasing capital spending and exploring infrastructure expansions, though challenges like overbuilding and rising construction costs persist.
Zoning regulations also pose challenges, as finding suitable locations near physical infrastructure without overwhelming existing systems requires careful planning. In response, the industry is adopting innovative strategies, including the repurposing existing real estate such as old malls and factories, and expanding into emerging markets in Southeast Asia, Africa, and Latin America.
With high energy consumption, sustainability becomes a priority, so investments in renewable energy, passive cooling, and nuclear power to overcome grid limitations are key. In 2024, renewable sources like wind, hydro, and solar provided a record 32% of global electricity, surpassing the 30% share in 2023. This growth aligns with the data centre industry’s shift towards greener operations.
Some companies are exploring nuclear power as a solution to provide massive, always-on power, free of carbon emissions. The US Department of Energy has identified federal sites, including major national laboratories, as potential locations for data centres aimed at accelerating AI development, leveraging existing energy infrastructure and the potential for expedited permitting, especially for nuclear energy projects.
The proliferation of AI has exacerbated concerns over increases in data centre demand. Training large models requires substantial computational power, contributing to a significant rise in electricity consumption. However, it’s also possible that we may see diminishing marginal returns from using more data centres for AI and, eventually, demand growth could slow down. Emerging AI models requiring fewer processing chips may reduce future power needs, indicating that while current demand is high, future growth may stabilise.
While concerns about data centre shortages are understandable given the rapid advancements in technology and increasing digital demands, the industry is demonstrating resilience and adaptability. Through the development of both massive and modular data centres, strategic location planning, investment in sustainable energy sources, and continuous innovation, the data centre sector is well-equipped to scale efficiently. The challenges remain significant, but with proactive strategies and technological advancements, the industry is poised to meet the demands of the digital era without approaching a resource-constrained peak.
