Over the next decade, the most significant force shaping data centre design will undoubtedly be the growing influence of artificial intelligence (AI). While hyperscale companies and the dominant technology giants—Google, Apple, Meta, Amazon, and Microsoft (GAMAM)—are leading the charge, broader adoption will inevitably follow as AI applications become more accessible. However, this will not replace traditional IT infrastructure but rather add another layer of complexity, requiring data centres to support an evolving mix of computing needs. Quantum computing may also enter the equation, potentially offsetting rising energy demands. Yet, until the real benefits and requirements of AI and other emerging technologies are fully grasped, planning with precision remains a challenge.
The readiness of today’s data centre stock to accommodate AI-driven workloads varies significantly. Only those operators who have closely monitored emerging trends and collaborated with organisations such as the Open Compute Project (OCP) will be well-positioned. The adoption of high-density racks, advanced cooling techniques—including direct-to-chip heat sinks and immersion cooling—and flexible infrastructure design are critical factors in ensuring AI-readiness. Data centres with the foresight to provision for structural flexibility will be best placed to adapt, enabling them to shrink ‘white space’ (occupied by IT equipment) and expand ‘grey space’ (housing critical infrastructure such as power and cooling systems). However, even the best-prepared facilities may need to enhance their cooling capacity and secure additional power supply to meet growing demands.
Sustainability is a pressing concern for the industry, yet many facilities remain underprepared to meet emerging regulatory and environmental targets. While reporting frameworks are developing, there is still a lack of standardised mechanisms to track Scope 3 emissions, energy consumption, and water use—especially in colocation centres where tenants may be reluctant to disclose operational data. Although tools like Data Centre Infrastructure Management (DCIM) software provide a means to monitor energy efficiency, confusion persists over what should be reported. The European Energy Directive (EED), for instance, targets a Power Usage Effectiveness (PUE) of 1.3, yet global averages remain closer to 1.56. More than simply improving PUE, the industry must focus on holistic measures of data centre efficiency that reflect real-world IT processing effectiveness. Enhanced collaboration between governments and operators is essential to defining these metrics and identifying truly sustainable facilities.
The single most critical factor in future-proofing data centres will be securing flexible power provision. While headline-grabbing figures suggest that 250kW racks are becoming the norm, the reality is more nuanced. IT platforms will continue to require a diverse range of equipment, and legacy systems will remain in operation for years to come. Although technological advancements are progressing rapidly, adoption tends to lag behind. The industry must strike a balance between preparing for the future and ensuring that current infrastructure remains viable.
As AI workloads grow, the spatial dynamics of data centres are shifting. Traditional ‘white space’ footprints are shrinking as individual racks become significantly more power-dense. A facility originally designed to accommodate 500 racks at 5kW per rack may soon need to support only 20–50 racks at far higher densities. However, these changes bring significant infrastructural challenges. While liquid cooling solutions are becoming more widespread, some degree of air cooling will still be required, necessitating careful spatial planning. Additionally, the need for larger power supply systems and external heat rejection infrastructure means that plant space—often referred to as ‘grey space’—must expand accordingly. Crucially, retrofitting older facilities to support AI workloads will depend on the availability of external space for heat rejection systems and backup power.
Among all the hurdles in retrofitting older data centres, the most significant is the availability of external space. Many existing facilities were designed under the assumption that power densities would remain relatively stable. However, as AI accelerates demand, the need for increased backup power—traditionally provided by hydrocarbon-fuelled generators—and advanced cooling infrastructure is becoming a pressing concern. In a scenario where a 9MW facility is transformed into a 90MW operation, the sheer footprint required for additional cooling towers, dry coolers, and air-cooled chillers cannot be underestimated. Where roof space is insufficient, land must be secured—an increasingly complex challenge given urban constraints, airflow requirements, and noise mitigation measures.
As the data centre industry navigates this era of rapid technological evolution, flexibility, power availability, and sustainability will define its future success. The ability to adapt existing infrastructure while planning for new demands will separate forward-thinking operators from those struggling to keep pace. AI, quantum computing, and sustainability targets will push the industry to reimagine its approach to design, making this next decade one of the most transformative yet.
