09 April 2024
Billy Durie, global sector head for data centres, Aggreko
AI has become the word on seemingly everyone’s lips at the minute, with tools such as ChatGPT receiving widespread attention for their applications across several industries. However, amid this increase in demand, the strain on data centres, and in turn the natural environment, is increasing at a similar rate.
The impact on water usage is already being felt by the industry. A recent article from The Telegraph investigated the scope of this issue, revealing that each command given to ChatGPT ‘drinks’ the equivalent of a sip of water, with 20 tasks equating to half a litre. The positive applications of AI are innumerable, but it is important to recognise that the use of AI also harbours a very real physical impact.
Microsoft recently revealed in its latest environmental report that its global consumption of water rose significantly between 2021 and 2022, growing by over 34%. Some researchers believe this to be because of research into AI. Google also reported similar findings of water increases across its American sites.
This increased use and research of groundbreaking artificial intelligence systems comes with a drawback of higher water consumption due to the high amounts of cooling required to maintain operations. However, this also coincides with increasing instances of drought across Europe and the wider world, with countries such as Italy and Spain particularly affected in 2023.
Uncertain conditions
The scale of the challenge of water scarcity is only continuing to grow. Water scarcity is set to affect more people than ever in 2024, with 1.8 billion people will be living in regions or countries with absolute water scarcity, which implies having less than 500 cubic meters of water per person per annum. This issue is further complicated through the fact that data centres are sometimes in areas with water scarcity issues. In 2022, 23% of Microsoft’s and 18% of Google’s freshwater withdrawals were from areas facing scarcity problems. This means that it is especially important that companies must enhance their efficiency and look to reduce their water consumption, even if these operations continue to become more intensive.
One way that firms could tackle this issue head on would be through the use of new cooling technologies that may help data centres cope with the demand of such operations while curbing their water consumption.
One cooling technology that could be vital to helping reduce water-use could be closed-loop cooling systems. This is where a chiller forms the primary cooling solution through distributing water-chilled air through a cooling circuit consisting of a compressor, evaporator, expansion valve, condenser, and multiple Computer Room Air Handers (CRAHs). Each section of the system then supplies cooled air to the racks.
From here, the ejected warm air is directed to a heat exchanger and cooled again, before being recirculated. Crucially, this method results in little to no water losses, as the system uses a fixed volume of water that recirculates.
Traditionally, the only thing really being considered by data centre decision makers was Power Usage Effectiveness – or PUE. However, as data centres become increasingly thirsty and water becomes less accessible, water usage effectiveness – WUE – must also be considered.
This is where closed-loop cooling systems give users an advantage. They achieve an excellent WUE value as they can function on a fixed volume of water. Compared to a standard open-loop system, which only consists of a water-cooled chiller with a cooling tower, which will require a constant supply of water due to evaporation, drifts, and blowdown.
Closed-loop future
This is why the clear way forward involves closed-loop cooling. Climate-conscious data centre operators can make a huge push towards sustainability by choosing this system, which can go a long way to helping to reduce the amount of water being used by the industry during testing environmental conditions.
