In simple terms, AI is not good for the environment. While it may be streamlining processes, making us work smarter, leaner and on the face of it, greener, the reality is that the power required to keep this tool working is huge; 10 times more electricity than a Google search, which according to Goldman Sachs, could lead to a 160% increase in power demand in the next five years.
It’s clear that data centres have a voracious appetite for electricity – in 2023 they are reckoned to have used a whopping 7.4 gigawatts (GW) of power across the globe, a 55% leap from the 4.9GW in 2022. Moreover, the average modern data centre uses around 40% of its power for cooling; an inescapable essential of data centre operations.
The very nature of this sector, however, means that no expense is spared in developing ever-more efficient solutions to bring down cost and improve environmental credentials. State of the art cooling systems that benefit from low global warming potential (GWP) HFO refrigerants, for example, can reduce energy consumption by up to 50%. Natural refrigerants, such as propane, are also gaining in popularity.
In this landscape of cost vs sustainability, liquid cooling is becoming the preferred option thanks to specific benefits that suit the increasing power density and heat output of modern hardware – especially those focused on AI. Traditional air-cooling systems struggle to effectively dissipate the heat generated by these dense workloads, leading to hotspots, reduced operational efficiency, and potentially compromised reliability.
By reducing energy consumption and improving the lifespan of components, liquid cooling will lower operational costs. Higher server density can also lead to better utilisation of space, while better facilitating the capture of heat. Making use of this heat, an inescapable by-product of data centre operations, is key to a more sustainable and holistic approach to data centre location planning.
And when it comes to planning for the future, the scalability of liquid cooling systems makes them well-suited for accommodating growth and expansion, allowing data centres to easily scale their cooling capacity in response to changing demands.
In the UK, which is largely densely populated, the benefits of liquid cooling fit with the ‘edge’ model, siting data centres next to net users of heat. Using district heating systems, the data centre provides a source of ‘free’ heat for nearby buildings, while also potentially improving data processing efficiency. Rejecting heat into the air uses more energy than reusing it using an ambient loop.
This is happening now, with examples including Queen Mary University of London, which exports heat from its onsite data centre to provide hot water for campus facilities, and also an innovative scheme in Exmouth, where a public swimming pool uses excess data centre heat, slashing its bills by 60%. In these examples, one energy input results in two useful energy outputs, positively impacting the carbon footprint of both the data centre and the buildings in the loop.
Beyond saving money and improving sustainability credentials, this approach may also serve to safeguard under threat services. Since 2010, England has lost nearly 400 swimming pools, often forced to close because of rising costs, negatively impacting the health and wellbeing of affected communities. As the Exmouth model shows, data centres could hold the key to helping community assets stay operational and affordable.
To smooth the road to a symbiotic future where our unwavering reliance on the internet is, in-part, offset by data centres in combination with district heating schemes, we need to overcome some of the barriers. Planning laws are one such hurdle, which are hopefully being addressed by Labour’s planning reforms, set to loosen the rules for nationally significant infrastructure projects.
The other major challenge comes from shared responsibility – of heat capturing and distributing technology and the heat itself. Incentives for both parties must be clearly defined.
We can’t escape AI’s increasing impact on the modern world. If managed the right way, however, the march of the data centre could actually lead to an acceleration in the ‘smart city’ model, where the wastage of one asset is used by another.
