The UK government’s environment watchdog has endorsed a dodgy and misleading report from Big Tech PR men, saying data centres don’t use much water – that’s a big problem. Here’s why
By Tim Squirrell
Big Tech lobbyists have produced a report telling us we can all stop worrying about our drinking water being guzzled by data centres before it gets to our taps – despite government, regulators and the water industry all warning that data centres are a major, growing source of strain on the system.
I work for Foxglove, a tech justice non-profit that has just launched the first ever legal challenge against construction of a hyperscale data centre in the UK. ‘Woodlands 2’ is being rammed through planning permission by the government, against the local council’s wishes, despite serious concerns over its impact on both the environment and increasingly scarce energy and water resources.
This report, produced in collaboration with the Environment Agency (!), is likely to be cited by those who want to carpet the UK with data centres. But it has three major problems.
First, the value of data the report is based on comes with a serious health warning. There are about 500 data centres in the UK. Tech UK’s report looks at self-reported water consumption data from just 73 of them, or about a sixth of the total. Not a great sample size, but perhaps they’ve done the work to make the best use of what they got?
Unfortunately not. The data collection is voluntary and anonymous. The sample wasn’t validated independently, nor did Tech UK collect any data on the size or the power demands or even location of the respondents.
Data centres come in all shapes and sizes, from glorified server cupboards to giga-behemoths using upwards of 1GW of energy to power GenAI tools, and they have wildly varying water consumptions. We don’t know which of these were represented in the survey.
We can’t take a report seriously as a tool for decision-making if it falls prey to methodological flaws we wouldn’t accept in an undergraduate essay. You can’t really learn anything about the UK’s data centres from a non-random sample, permeated with self-selection bias, which is what this is.
Tech UK’s report confidently states its findings: “challenge the assumptions that data centres are inherently water-intensive”. That might be helpful if we knew that all the hyperscale centres currently being built to facilitate chatbots were air-cooled, and so don’t use much water. But we don’t.
All this report tells us is that a self-selected sample of data centres of unknown size self-report that they don’t use much water for cooling. It could just as well be a series of slightly enlarged cupboards who responded to Tech UK’s survey to say “don’t worry, we’re only using air.”
It’s a bit like if we tried to tell you we knew how many cheeseburgers each of the 600+ players in the Premier League ate every year after having to chatted to about 100 of them. And they didn’t have to tell us their name, height, weight or what team they played for, or provide any proof they were telling the truth. Plus we work for the Premier League’s PR department and have published a report on why cheeseburgers are not a problem for the performance of professional football players.
On to the next problem. Despite the title – ‘Understanding data centre water use in England’, the report doesn’t do much to help us understand its subject better at all.
For example, it doesn’t say or even attempt to give an estimate on how much water data centres in the UK use.
Foxglove has actually tried to do this, working with The Times newspaper, where we found that existing data centres guzzle down at least 10 billion litres of water per year, though that is likely to be a significant underestimate.
But opting not to try and put a figure on total data centre water use itself hasn’t stopped TechUK from comparing its partial sample and analysis to verified figures from other industries, like agriculture, to make data centres look less demanding.
This is a silly comparison. Agriculture is intrinsically dependent on water, whereas this report itself argues data centres are not. There are indeed systems for cooling that do not use water but are either more expensive to install or less capable of dealing with the latest generation of chips, and inherently use more power because there is always a trade-off between water use and energy use.
The report says correctly that hundreds of millions of litres of water are abstracted annually for agriculture – the key word being abstracted. That means using surface or groundwater. Data centres broadly use clean drinking water direct from your pipes. It’s not a remotely credible comparison.
This isn’t research. It’s a lobbying tool. The final problem is that it’s both a shame and very worrying that the Environment Agency, a government regulator, has put its name on it.
At best, it’s not a great look. It tarnishes their reputation for scrutiny and high standards of rigorous analysis. At worst, it’s yet another sign our government has become far too close to a coterie of unelected billionaires squatting at the top of the tech giants – and increasingly calling the shots in Britain.