Artificial intelligence is transforming the way data centres are designed and operated, and with it, the expectations placed on energy systems. The expansion of AI workloads is changing the volume and nature of energy use in the UK and the larger EMEA region, not just creating new demand. This makes it necessary to match the rate of digital growth with infrastructure that can deliver both sustainability and resilience.
A step change in energy demand
The increase in data centre power use over the next decade will be dramatic. In Europe, consumption currently stands at around 62 TWh annually, yet this is forecast to rise to 150 TWh by 2030. IT load is expected to more than triple from 10 GW in 2023 to 35 GW by the end of the decade. AI is a major driver of this growth, with dense and continuous computing requirements that far exceed those of traditional workloads.
Such rapid escalation has immediate consequences for the energy systems that support these facilities. In established hubs including London, Frankfurt, Amsterdam, Paris and Dublin, grid capacity is already stretched close to its limits. In Dublin, for instance, data centres are reported to account for up to 80 per cent of local grid demand. Connection queues are lengthening, with waits of seven to ten years now common and in some cases more than a decade. The International Energy Agency estimates that almost one in five planned global data centre projects faces delay because of grid bottlenecks, with moratoriums already in place in some markets.
The pressures are being felt not just in Europe’s traditional hubs but across the wider region, where governments are seeking to expand sovereign AI capacity and digital infrastructure at speed. Without a reliable and sustainable energy strategy, ambition will quickly outstrip what can be delivered.
The role of on-site generation
One outcome of these constraints has been the rapid adoption of on-site power generation. Developers are increasingly investing in “behind the meter” capacity, most often gas-fired turbines, which can be deployed in months rather than years and give operators direct control over their own availability and resilience.
This approach brings clear advantages in terms of speed to market and operational certainty, but it also raises difficult questions about sustainability. Diesel generators are ill-suited to continuous operation, with high emissions of nitrogen oxides, particulates and carbon dioxide that are subject to tight restrictions. Gas turbines offer a cleaner alternative, with lower emissions and fewer particulates, and can be permitted for continuous use more easily. They are also more reliable at scale and less exposed to the logistical risks that come with storing and transporting diesel. At the same time, their reliance on fossil fuels makes them a challenge to align with longer-term decarbonisation goals.
For many, gas is a bridge technology that provides breathing room while renewable capacity and advanced storage catch up. In the longer term, however, solutions such as battery systems, demand-side flexibility, small modular reactors and green hydrogen will be required if the industry is to achieve a sustainable balance between growth and carbon reduction.
From reporting to operational action
The conversation around sustainability has also shifted. For many years, data centre operators treated metrics such as power usage effectiveness, water consumption and energy mix as measures for annual reporting and compliance. Although this is still significant, the industry is increasingly employing these numbers as real-time tools to guide everyday decisions.
Performance data is now tracked in real time and connected to operational playbooks that direct immediate action rather than being tucked away in reports. If cooling efficiency drops, for example, teams follow a defined process to investigate equipment, make adjustments and escalate if needed. If water consumption rises unexpectedly, operators are equipped with clear steps to trace potential leaks, optimise system performance or switch to alternative sources.
By incorporating measurements into ongoing operations, the emphasis is shifted from compliance to tangible outcomes. Sustainability data becomes meaningful only when it shapes how facilities are run and how they respond to challenges in real time.
Managing energy and water together
Concerns about water use in cooling systems are growing alongside scrutiny of energy. Yet the reality is often misunderstood. Terms such as “water-cooled” or “direct liquid-cooled” do not imply that water is wasted. These are closed-loop systems that recycle the same liquid again and again to transfer heat, with no loss in the process. Where water use does occur is at the heat rejection stage, sometimes through evaporative cooling.
In the UK, this practice has never been widespread. Thanks to the country’s cooler climate, most operators use air-cooled chillers, which do not rely on evaporation at all. As a result, water consumption in UK data centres is already significantly lower than many assume.
Practical examples of how cooling can be part of the solution are beginning to appear. In Urmston, Manchester, Deep Green and Keysource are working together on an AI data centre that will capture heat and recycle it into the community swimming pool. The initiative is expected to save the leisure centre tens of thousands of pounds each year and reduce emissions by hundreds of tonnes over its lifetime. Although modest in scale compared with the sector’s overall footprint, it shows how rejecting heat can be turned into recovering energy and delivering benefits directly back to local communities.
Putting the scale in perspective
The debate about water use also needs to be placed in a broader context. According to the Environment Agency, the UK’s top ten water companies collectively lose around 2,700 million litres every day through leaking pipes. That amount of water would be enough to cool between 20 GW and 30 GW of IT capacity using legacy evaporative systems. To put this in perspective, the total global operational data centre capacity today is estimated at 42.4 GW.
In other words, the water lost by utilities in the UK every single day could almost cover cooling demand for the entire global data centre footprint. The sector is already adopting technologies that minimise water use, while also advancing methods to recover and repurpose energy that would otherwise be wasted. The greater priority should be placed on addressing systemic losses elsewhere in the water system.
Conclusion
With energy use increasing more quickly than grid capacity and environmental obligations becoming more pressing, artificial intelligence is driving the data centre industry into uncharted territory. Operators, legislators, and energy suppliers must work closely together to address these issues, and they must be prepared to view sustainability as an active operational discipline rather than a compliance exercise.
The design, management, and integration of facilities with their surroundings are now guided by metrics that formerly assessed progress after the fact. Small but significant energy reuse examples also demonstrate how the sector may innovate to benefit operators and communities. The way forward is clear: to construct durable, effective, and sustainable infrastructure while simultaneously meeting AI’s expectations.
