IS WATER AVAILABILITY THE NEXT BARRIER FOR DATA CENTRE DEVELOPMENT?

Drinking us dry

Following on from the UK Government’s announcement of an ambitious AI action plan that includes the development of "AI Growth Zones" to encourage data centre construction and boost the UK's AI capabilities, there have been some concerns raised over the potential impact on the water supply and whether demand will outstrip supply. 

This has resulted in some alarmist headlines about there not being enough fresh, drinkable water to supply homes and businesses and the BBC has reported that UK water industry company Thames Water is already speaking with the government about plans to confront possible water shortages. 

It is a similar situation in the US as highlighted by a recent report by Black & Veatch, which indicates that many water utilities are not fully aware of how data centre growth might affect water resources. More than half of the participants (54%) acknowledged they have not accounted for the rise of data centres in their resource planning strategies.

UK water companies are under pressure

Certainly in the UK, the water companies are under pressure. The general perception is that the chronic lack of investment in infrastructure has meant that despite there being enough rain to more than support datacentres and the population – this may not be possible. Recent analysis by the University of Greenwich suggests that shareholders in some of the UK's largest water companies have taken out tens of billions of pounds but failed to invest. It found that investors have withdrawn £85.2bn from 10 water and sewage firms in England and Wales since the industry was privatised more than 30 years ago – although Ofwat, the industry regulator, said it "strongly refuted" the figures. However, it did confirm that between 2020 and 2022, 14 water companies underspent their budget on improving their water network and it is worth noting that the last reservoir was built in 1991 and in that time the population of London has increased from 6.8 million to 9 million people! Recently, Thames Water has been hit by a huge £122.7million fine for breaking rules on its wastewaster operations and dividend payments. The fine is the largest Ofwat has ever issued.

If we accept that data centres are indispensable for the continued advancement of artificial intelligence then we will need a solution and in the same way there are expectations on the national grid to improve grid capacity to support expansion,  surely the same should be true for the water companies?

Different regions have different requirements

While the cooler climate in the UK means that most data centres operate with relatively low water usage, global trends are shifting particularly to support AI processing, which creates greater heat and therefore requires more water to cool it. Leading international data centre operators are reporting increasing water consumption, driven largely by the growing demands of AI and cloud infrastructure. Notably, both Google and Microsoft have disclosed year-on-year rises in water usage at their data centres since 2020, reflecting the resource-intensive nature of these technological advancements.

Sector innovation

However, the sector is also responding to the challenge with innovative solutions. For example, Microsoft is planning to implement a cooling approach that eliminates water evaporation, potentially saving more than 125 million litres annually per data centre.

In the UK and other nations where the water infrastructure is potable, grey water harvesting (rain etc) and utilisation, so that water required isn't a diversion of drinking water, is a feasible option. There are other alternatives available too, such as dock water cooling, where no water is utilised, but it is pumped through a heat exchanger and returned at a slightly higher temperature. This has been utilised on data centres in London and is similar in principle to how Sizewell C and other power stations operate.

Does AI hold the answer?

One of the benefits of AI is that whilst the front-end interface part of AI requires low latency, the machine learning doesn't and therefore can be located in regions where they can be designed to use minimal water - cooler regions and those where there are large bodies of cool water. Scotland and Scandinavia are ideal for this as we can use loch/ fjord water cooling in these locations and eliminate the need for evaporative cooling. As a result, smart site selection and integration of water usage requirements in planning will reduce reliance on evaporative cooling.

Moreover, AI itself can be leveraged to enhance the efficiency of data centres. AI algorithms can optimise energy use by predicting cooling needs, managing workloads more efficiently, and reducing idle times for servers. Predictive maintenance powered by AI can also prevent equipment failures, thereby reducing the need for excessive cooling. 

A multi-faceted approach

There is no doubt that the interplay between AI data centres and resource usage necessitates innovative approaches to mitigate environmental impacts, but it is safe to say that the data centre sector is on the case and has been for several years now as we have been dealing with these issues as part of the more general ongoing demand – AI is just the next challenge.

Addressing these issues requires a multifaceted approach that includes transitioning to renewable energy, adopting innovative cooling technologies, and leveraging AI for operational efficiency.  As the demand for AI capabilities continues to grow, so too must our efforts to ensure that this growth is sustainable and responsible, balancing technological progress with environmental stewardship. At BCS, we help clients navigate this path to deliver optimum locations and positive outcomes for our global community.