I was listening to a BBC interview between Rachel Millward, deputy leader of the Green Party, and Mark Acton of the Data Centre Alliance when I realised how quickly this particular debate can become polarised.
Millward was arguing for a temporary freeze on new data centres in England. Her case was environmental: water use, energy demand, local disruption, pressure on housing and infrastructure, and the wider implications for climate policy.
Acton pushed back hard. Some of the claims, he said, were simply wrong. The alarming figures being quoted about water use came largely from the United States and were not applicable to Britain. UK data centres were being unfairly portrayed as vast, thirsty industrial monsters when many used little or no water for cooling. Data centres were not optional, he argued. They were critical infrastructure. Banking, communications, websites, streaming, public services and much of modern life depended on them.
- The debate left me unconvinced by both sides
- Do UK data centres really use huge amounts of water? Size and location matters.
- Electricity usage may be the bigger issue.
- Housing, infrastructure and the question of priority.
- Britain still needs data centres.
- What I would expect from the planning system.
I came away neither persuaded that Britain should freeze data-centre development nor reassured that the environmental concerns can simply be dismissed. The evidence points instead towards a planning failure: we are promoting a new generation of extremely resource-intensive infrastructure without yet having sufficiently transparent, site-specific information about its demands on water, electricity and the places in which it is built.
Listening to the two of them, I found myself sympathising with both at different moments.
That usually makes me suspicious.
I am involved in local politics and spend a fair amount of time thinking about environmental policy, planning and the rather less glamorous business of how infrastructure actually works. I am therefore wary of arguments that sound compelling but depend on a headline statistic, particularly when that statistic has travelled halfway around the world before arriving in a British planning debate.
So I wanted to know what we actually know.
The first thing I discovered is that the argument about water is more complicated than either side made it sound.
There certainly are data centres in the United States that consume astonishing quantities of water. Some enormous facilities operate in hot, dry environments using evaporative cooling systems. Those figures are real.
But Britain is not Arizona.
Our climate is different. Cooling systems are different. A significant proportion of existing UK data centres use air cooling or closed-loop systems that consume relatively little potable water.
That means we should be extremely cautious about taking a dramatic American statistic and presenting it as though it describes a proposed data centre in Berkshire, Sussex or Yorkshire.
On that point, Mark Acton was right.
But he then went too far in the other direction.
To say that British data centres do not use water is not really sustainable either. Some do, and some large facilities can use significant amounts. More importantly, the national figure is not necessarily the important one.
Water is intensely local.
A development that represents only a tiny fraction of national water consumption can still matter considerably if it is located in a water-stressed catchment in southern England.
That distinction is particularly important as the climate warms.
England already faces serious questions about its future water supply. Population growth, housing, agriculture, industry and climate change are all increasing pressure on the system. Into that comes a new generation of potentially enormous digital infrastructure.
And here I think the debate changes.
The real issue is not whether existing British data centres collectively consume an alarming proportion of all our water. At present, they appear not to.
The issue is what happens next.
The data centres being contemplated for artificial intelligence can be on a different scale altogether. Their electricity requirements may be measured in hundreds of megawatts. Their water use will depend enormously on how they are designed, but we do not yet seem to have the comprehensive, mandatory and transparent reporting system that I would expect for an industry being promoted as strategically essential.
That troubles me.
If I were looking at a major planning application for a data centre, I would want some fairly basic questions answered.
How much water will it use?
Not a theoretical industry average. This facility.
Where will that water come from?
How much will it use during a heatwave?
What cooling system is proposed?
What happens if the local water company has to impose restrictions?
How does the demand look not just under today’s climate but under the conditions we may face in twenty or thirty years?
Those are not anti-technology questions. They are planning questions.
The same applies even more strongly to electricity.
This is where I think the public debate risks looking in the wrong direction. Water makes an immediate and understandable headline, but electricity may be the more consequential issue.
A large AI data centre can require an extraordinary amount of power.
Hundreds of megawatts is not a trivial load. It is an industrial-scale demand appearing on an electricity system that is simultaneously being asked to decarbonise transport, heating and manufacturing.
That creates choices.
If a huge data centre receives a grid connection, what else might have used that capacity?
Could it have supported housing?
New industry?
Electric vehicle infrastructure?
Heat pumps?
The answer may sometimes be that there is no direct conflict at all. Grid geography is complicated, and a data centre in one part of the country may even make useful use of generation that would otherwise be constrained.
But that is exactly why I want the question examined properly.
Millward said in the BBC interview that data centres were being given priority over homes and public services because they had been designated Critical National Infrastructure.
That is not quite right.
The designation itself does not simply entitle a data centre to jump the queue for water and electricity.
But the broader concern is not imaginary.
Government has decided that data centres and artificial intelligence infrastructure are strategically important. It is creating AI Growth Zones, looking at planning reform and trying to resolve grid-connection obstacles.
So there is government preference here.
That does not automatically make it wrong. Governments are supposed to make strategic choices.
But once the government starts giving a sector accelerated treatment, I think the reciprocal obligation should be greater transparency and stronger public-interest tests.
If a development is sufficiently important to receive preferential treatment, it should be sufficiently important to explain exactly what resources it will consume and what benefits it will deliver.
There was another aspect of the interview that bothered me.
Millward talked about hundreds of vast data centres being imposed on communities. Acton rejected that outright.
Again, the language gets in the way.
There are already hundreds of facilities of different kinds in Britain, and many more are proposed. But a small conventional data centre and a gigantic hyperscale AI campus are not remotely equivalent.
Counting them as though they were identical is almost meaningless.
What matters is scale.
A 10 MW facility and a 500 MW facility may both be called data centres, but from the point of view of energy planning they belong in different worlds.
The same problem applies to the phrase “data centre” itself.
It has become associated almost entirely with artificial intelligence, but data centres already underpin much of ordinary life.
Every time I use online banking, watch something on iPlayer, send a message, access a website, store something in the cloud, or use a digital public service, physical computing infrastructure is somewhere behind it.
Acton was absolutely right about that.
We cannot simultaneously demand digital services and pretend that the buildings housing the servers are unnecessary.
Nor is there necessarily an environmental gain in pushing all of this infrastructure abroad. Britain has a relatively temperate climate. Our electricity system is decarbonising. In some circumstances it may make more environmental sense to locate a facility here than in a hotter country with greater water scarcity and a more carbon-intensive grid.
But that does not mean every British data centre is environmentally benign.
This is where the carbon argument becomes complicated.
A developer may say that a facility runs on renewable electricity.
That sounds reassuring, but it is not necessarily the end of the calculation.
A data centre operating around the clock creates additional demand. The important question is what generation meets that demand at the margin.
If renewable generation, storage, transmission and firm low-carbon power expand rapidly enough, the additional load may be accommodated with relatively little carbon impact.
If they do not, gas generation may run for longer, or clean electricity that might otherwise have displaced fossil fuels elsewhere may be absorbed.
That is a systems question, not a branding exercise.
There are also opportunities.
Data centres produce huge quantities of waste heat. In the right place, that heat can potentially be captured and used in district heating systems.
They can be located strategically where renewable generation is abundant and network capacity exists.
They can be designed around low-water cooling.
They can be built on brownfield land.
They can contribute economically to areas in need of investment.
All of which brings me back to planning.
I do not think the answer is a simplistic “no more data centres”.
I am equally unconvinced by “leave it to the industry; there isn’t really a problem”.
What I want is strategic planning.
I want major developers to publish projected and actual energy use.
I want site-level water data.
I want cooling technology disclosed.
I want environmental impact assessments that consider future climate conditions rather than simply historic averages.
I want the grid consequences examined.
I want waste heat to be considered a resource rather than a nuisance.
And where a development is receiving government assistance, accelerated planning or preferential infrastructure access, I want to know what public benefit is being obtained in return.
Above all, I want the figures to be public.
That was perhaps the most striking thing I took from the BBC interview.
Two intelligent people could make almost completely contradictory claims about a rapidly growing piece of national infrastructure because the underlying evidence remains insufficiently transparent.
One says data centres consume enormous quantities of water.
The other says they effectively consume none.
One says they are being pushed through the planning system.
The other says getting permission is extraordinarily difficult.
One says they threaten climate progress.
The other says they are indispensable to the modern economy.
Parts of all of those statements can be true at the same time.
What is missing is specificity.
Which data centre?
How big?
Where?
Using which cooling technology?
Taking water from which catchment?
Drawing electricity from which part of the grid?
Generating what local economic benefit?
Producing what environmental cost?
Those are the questions that matter.
And they are precisely the questions local councillors and planning authorities are going to have to ask more frequently.
I suspect data centres will become one of those planning issues that initially appears highly technical and then suddenly becomes very local indeed: a proposed site, a new substation, a water connection, a landscape impact, noise complaints, grid capacity, promises of jobs and competing claims about national necessity.
At that point, slogans will not help very much.
The conclusion I reached after listening to Rachel Millward and Mark Acton is therefore rather less dramatic than either side of the argument.
Britain needs data centres.
It also needs to be much more discriminating about where they are built, how they are powered and cooled, and what claims developers are required to substantiate.
The environmental case should not depend on exaggerating the evidence.
The industry case should not depend on minimising genuine resource constraints.
If this really is Critical National Infrastructure, then we should plan it as critical national infrastructure.
That means deciding what we need, identifying where it should go, setting clear environmental standards, publishing the evidence and making the trade-offs explicit.
Rather than discovering, one planning application at a time, that we have accidentally built a national strategy.
My takeaway: councillors and planners need the information that will distinguish a well-sited, low-water, efficiently powered facility from one that places unacceptable pressure on a constrained locality. It made me think of the ‘pin the donkey’s tail’ party game! The donkey needs a tail, but you’ll never stick it in the right place wearing a blindfold.
Leave a Reply