My Ph.D. supervisors and I recently published a paper on the water footprint of data centres. Considering how important data centres are in storing and transmitting much of the data that we engage with online, it is disappointing how little attention has been paid to their water use so far.

Our headline finding was that worldwide one gigabyte of data sent from the average data centre can have a water footprint as high as 200 litres. To put this in perspective, if I were to download the full Radiohead discography (around 1 GB) from a data centre, I could generate a water footprint roughly equivalent to that generated by the production of one kilogramme of tomatoes.

To qualify this, there is an extremely large uncertainty in this finding as it could be as low as only 1 litre of water per gigabyte. The real value will most likely be somewhere in between but it is extremely difficult to say exactly where. This is because there is little to no reporting on water consumption and pollution in the key drivers of data centre water footprint: the amount of water evaporated and polluted by different cooling systems, the ubiquity of each system in the data centre population, and the indirect water footprint from the electricity powering the IT equipment and surrounding data centre infrastructure.

Since the lion’s share of data centre water footprint comes from this indirect footprint, it is the water intensive energy sources like biofuels, hydropower, and some hydrocarbons in electricity generation which are the main culprits. Prominent firms operating data centres are already working on reducing the CO2 and water footprints of their facilities. Google has built a new data centre in Finland which uses recycled wastewater and cool outside air. Apple has built solar panels to reduce carbon emissions, thereby seizing the co-benefit of reduced water footprint. However, most data centres are not as large as these nor nearly as prominent. They are therefore not under pressure to reduce environmental impacts as much.

The internet and its technologies have been spreading at an astonishing rate, becoming one of the fastest growing consumers of electricity in the EU and the US. While this could contribute to rising water stress, it’s not all bad news. Data centres can offer us new ways of reducing water footprint by decoupling economic activity from water use and by improving resilience to water stress.

An EU study found that the rising use of information and communication technologies is likely to reduce overall electricity consumption through improved efficiencies and virtualization of economic activity. Sending an e-book is less resource intensive than sending the hardback. The information and control provided by smart infrastructure and other applications of digitalization can help to make better use of resources. In these ways, data centres can help to improve efficiencies and decouple economic activity from environmental impact. The caveat being that this will only work if the rate at which efficiencies improve is faster than the growth in activity enabled by this greater efficiency.

Another benefit of spreading data centres is that they can optimize workload distribution across different sites to operate where the climate and electricity supply is most amenable to IT equipment, minimizing electricity costs, CO2 emissions, and water use. If an area is hit by a drought, data centre operators can quickly redirect data traffic away from that area to reduce the risk of downtime as well as reducing pressure on the environment from the electricity consumption and cooling. That being said, data centres are only a small part of the total of economic activity so the impact of redistributing workloads will be only very small. Data centres must, like others, first strive to reduce their own impact before helping others.

The question does not end there however. Given the urgency of water problems in many parts of the world, not least in high-tech California, demand side actions must also be considered. In the UK, the average household used approximately 17GB of data per month in 2011. With the move towards more high quality video content and mobile internet that value is sure to have risen substantially in the meantime. Assuming the high end of the range for the water footprint of data, downloads add roughly 3% to the average household’s water footprint.

What can a consumer do though, without control over the way their data is delivered to them? One option is that users could choose an environmentally friendly internet service provider. Inadequate reporting and unstandardized metrics makes such a decision difficult but more and more, providers are beginning to be aware of and report on their environmental impact. Hopefully in the future we will be able to compare different providers’ offers not just by cost but also by environmental performance thereby giving consumers greater choice over their impact.

Another option for consumers would be to reduce data use. Reducing data use requires behaviour change, which has time and again been shown to be a very difficult task and is effective only if everyone does it. Despite these criticisms, there are two important reasons why reduction of use is still important. Firstly, saying behaviour can be changed is the first step on the road to changing it. It is not unimaginable that we switch off Wifi when we are not using it and then go further. Secondly, policy and supply side decisions will not change as quickly without public pressure. While we do not have control over the decisions data centre operators make, we have control over our consumption and over our political engagement. By lobbying authorities to introduce reporting requirements and other environmental legislation we can change the conditions of our choices, thereby expanding our freedom and protecting the environment.

The full paper is open access and can be found here:

Posted September 2015

Water, Energy, and Data – Bora Ristic, World