Three readers have written in with proposals for containing the climate cost of Britain's data centre buildout, and between them they produce the most useful number in the whole debate: 16,000 tonnes of CO₂ a year, the national saving if every person in the country sent one fewer thank-you email a day. That figure arrives in the same correspondence as another one — the roughly 1 million tonnes attributed on 21 August to a single proposed hyperscale data centre in Havering, east London, an estimate that coverage described as incompatible with net zero. Put the two side by side and the argument mostly settles itself.
Richard Lamming of Shaldon in Devon takes the hardest line on who owns the problem. The prospect of future data centres drawing more electricity than the whole of the United Kingdom shows, in his reading, corporations transferring their difficulties onto the public and the electorate, and he does not think it should be permitted. His remedy is that every data centre be made responsible for generating all of its own electricity and for eliminating its CO₂ emissions outright. The buildings have large roofs and large car parks available for solar; wind turbines, battery storage and emissions-scrubbing plants can go on the site. If the companies reply that this is impossible, Lamming's answer is that the impossibility stays on their side of the fence rather than moving to society's, and that new technical solutions may simply have to be found.
Michael McClelland of Devizes in Wiltshire starts from the Havering project and the emissions figure published on 21 August, and notes what the coverage left out: the site's very large water consumption, conspicuous during a drought. He considers local opposition to Havering and to the many other data centres already built or in development entirely justified. He then turns the argument back on the reader. Data centres exist because people send, receive and store digital messages, and most of that traffic is junk. Hence the 16,000-tonne research finding, and hence his proposal that government impose small charges on digital messages to shrink the volume of unnecessary data. We pay to send a physical letter, he points out, so why not a digital one.
Malcolm Fraser of Edinburgh offers the shortest intervention: require AI to design data centres that do not consume enormous quantities of water and energy before the owners of AI companies are allowed to build them.
Now the arithmetic the letters do not perform on each other. One building in east London: around a million tonnes. The entire population of the United Kingdom changing its email habits every day for a year: 16,000 tonnes. The Havering estimate is not presented as an annual figure, which makes the comparison loose in the letters' favour — and it still leaves the behavioural fix worth a low single-digit percentage of one project. A levy on citizens' messages is not a small policy. It would need metering infrastructure, an enforcement regime and a public willing to be charged per email. That is a great deal of machinery to move a number this size.
I think McClelland's instinct is sound and his target is wrong. The postage analogy breaks at the point that matters: we pay to post a letter because each letter consumes labour and fuel individually, whereas the cost of a data centre is fixed by the decision to build it, not assembled from the messages that pass through afterwards. Charging for emails disciplines the demand curve. Havering is a supply-side commitment already on a planning desk.
Which makes Lamming's letter the one with teeth, and for a reason that has nothing to do with rooftop solar. His proposal moves the burden of proof. Under current arrangements a developer asserts a power requirement and the grid, the water supply and the public accommodate it. Under his, the developer has to bring its own electricity and its own emissions accounting to the table before anything is approved, and he closes the escape route in advance: a claim of impossibility becomes the applicant's problem to solve, not the country's problem to absorb. Fraser's suggestion is rhetoric, but it points the same way — make the industry demonstrate the design before it pours the concrete.
The gap running through all three letters is water. McClelland raises it as the thing the Havering reporting ignored, and then his own proposal does not address it either; Lamming's self-generation scheme covers electricity and carbon and stops there; Fraser names water without quantifying it. Nobody has put a figure on how much a hyperscale site in east London would draw, or from where, in a drought year. Carbon at least has a number to argue about.
What the letters share, without saying so, is that every fix in them is a planning condition rather than a technology. Britain can decide what a data centre must bring with it before it is permitted to draw power — that decision is available and it is political. What it cannot do is let 16,000 tonnes of unsent emails stand in for making it.