Who should pay for the power hungry AI data centres? Readers say: the companies building them
Two planned UK data centres could emit more carbon than ExxonMobil. Letter writers want the tech industry to generate its own electricity and clean up its own mess.
Key points
- Two planned UK data centres could produce higher carbon emissions than ExxonMobil, according to analysis published 25 August.
- The combined electricity demand of future data centres may exceed the total current power consumption of the United Kingdom.
- Letter writers are calling for each data centre to generate 100% of its own electricity and reach net-zero carbon output.
- Proposals include fitting rooftops and car parks with solar panels, adding wind turbines on site, and installing battery storage and emissions scrubbers.
Two data centres, not yet built, could between them produce more carbon than ExxonMobil, one of the world's largest oil companies. That claim comes from analysis first reported by The Guardian AI, published on 25 August, and it has landed in the letters pages like a brick.
For ordinary readers, the numbers are worth sitting with. The electricity demand from planned data centres, the vast warehouses of computer servers that store and process digital information, including AI, could exceed everything the United Kingdom currently uses. Every home, every hospital, every factory.
Who should foot the bill?
Right now, that extra electricity demand falls on the national grid, which means it falls on everyone. Richard Lamming, one of the letter writers, calls this a straightforward transfer of corporate costs onto society. He wants the rules changed.
His proposal: every data centre should be legally required to generate all of its own electricity. No drawing on the grid. No passing the emissions problem to the public.
The physical case is not obviously absurd. Data centres are enormous. Their rooftops, their car parks, the land around them could host solar panels, small wind turbines, and battery storage systems that hold surplus power for when the sun is not shining. Emissions scrubbers, devices that capture carbon dioxide before it enters the atmosphere, could handle what remains.
A second letter writer, Michael McClelland, approaches the problem from the demand side rather than the supply side. His argument: we should simply send fewer digital messages. Every email, every video stream, every AI query consumes electricity. Reducing unnecessary digital communication is the cheapest clean-up of all.
Malcolm Fraser also contributed to the letters, though his specific proposal was not detailed in the source material.
What does this mean for patients and the public?
For people who rely on AI-powered health tools, diagnostic software, or electronic health records, none of this changes anything today. But the energy cost of running those tools is real, and growing fast. A single AI query uses roughly ten times more electricity than a standard web search.
If regulators do act, data centre operators could face new building codes or energy licences. That could slow some AI expansion, raise costs for cloud services, or, depending on how the rules are written, push investment toward greener infrastructure.
None of these letters carry the force of law. They are public pressure, directed at politicians who will eventually have to decide whether AI's electricity appetite gets treated as a private matter or a public one.
That decision is coming. The grid cannot wait indefinitely.
Common questions
Can a data centre really run entirely on its own solar and wind power?
In principle, yes, though it requires significant on-site investment and battery storage to cover nights and calm days. Some smaller facilities already run close to this model, but the largest AI data centres consume power at a scale that makes full self-sufficiency very difficult with current technology.
Does reducing my personal digital use actually make a difference?
A small one, individually. Collectively, yes. Streaming in lower resolution, avoiding unnecessary large file transfers, and limiting AI tool use for trivial tasks all reduce demand at the margins. The bigger levers sit with the companies designing and building the infrastructure.
