Unlocking compute within the UK’s power limits
Unlocking compute within the UK’s power limits
Data centres now underpin almost every aspect of life in the UK. NHS patient records, banking systems, government services, retail platforms, and personal devices all rely on them. Recognising this, the government designated data centres as critical national infrastructure in 2024, placing them alongside essential systems such as energy and water. Today, nearly every job in the UK economy is connected, directly or indirectly, to data centres through the digital services businesses depend on.
Energy constraints
In 2024, the UK had approximately 1.6 gigawatts of data centre capacity. With the right conditions, the sector could grow tenfold by 2035. However, this expansion depends heavily on overcoming constraints related to energy supply, regulation, and planning. Energy remains the most significant challenge. According to the National Energy System Operator (NESO), data centres consumed around 5.0 TWh of electricity in 2023 (about 2% of total UK demand), with projections suggesting demand could exceed 26 TWh by 2030, a more than fivefold increase.
A major concern for operators is the speed and uncertainty surrounding access to new grid connections. This directly influences investment decisions in UK data centre and AI infrastructure. Even when investment is secured, development timelines remain long due to these bottlenecks, limiting the sector’s ability to support broader economic growth.
Accelerating deployment
One way to accelerate deployment is by prioritising brownfield sites with access to existing or stranded power. At the same time, improving the efficiency of existing facilities is critical. A key metric in this effort is Power Compute Effectiveness (PCE), which measures how efficiently a data centre converts its available power into active IT load. PCE is not intended to replace Power Usage Effectiveness (PUE), but to complement it by focusing on how much of the provisioned power is actually used for computing.
Cooling systems play a central role here, as they account for a significant portion of energy consumption. Many legacy, compressor-based and air-cooling architectures limit the ability to convert delivered power into productive compute. By modernising these systems, operators can unlock additional capacity within existing power constraints, effectively increasing compute output without requiring new grid connections.
One promising solution is liquid spray cooling at the server level. If immersion cooling is comparable to submerging servers in a bath, liquid spray cooling is more like a targeted shower. By directly spraying heat-generating components such as CPUs and GPUs, it significantly improves heat transfer efficiency. This allows systems to operate at higher fluid temperatures without compromising performance, reducing the energy required for server cooling.
In many cases, high-performance servers using this approach can be cooled with only a dry cooler and circulating pump, even in ambient temperatures up to 50°C. This enables global deployment with year-round free cooling, eliminating the need for compressor-based systems and refrigerants. As a result, cooling energy demand is dramatically reduced, freeing up power that can be redirected to compute workloads and improving overall PCE.
Additionally, heat recovery and reuse are becoming increasingly important as data centres pursue net zero and carbon-negative goals. Liquid spray cooling enables higher-grade waste heat that is suitable for reuse in district heating and hot water systems, further enhancing sustainability while contributing to local energy ecosystems.
