UK Officials Underestimated AI Datacentre Emissions by 100x

The UK government vastly underestimated the climate impact of artificial intelligence, it has emerged, after officials raised their estimate of carbon emissions from AI by a factor of more than 100. According to new data quietly published this week, energy use by AI datacentres in the UK could cause the emission of up to 123m tonnes of carbon dioxide (CO₂) – about as much as generated by 2.7 million people – over the next 10 years.

That latest figure replaces a previous estimate – since deleted – that claimed emissions would reach a maximum of 0.142m tonnes of CO₂ in a single year. The discrepancy is not a rounding error. It is a fundamental miscalculation that reveals how little the Department for Science, Innovation and Technology (DSIT) understood about the physical reality of AI infrastructure before committing to it.

The Guardian report details how the revision appeared in the UK "compute roadmap," which sets out the government's plan "to build a world-class compute ecosystem" for delivering artificial intelligence in the UK. This is a goal on which the government has staked its hopes for economic growth. The problem is that AI datacentres require huge amounts of electricity to operate – much more than the datacentres used to store online data – and most of that continues to be generated by fossil fuels.

According to the DSIT's latest estimates, the carbon impact of the planned AI buildout could range from 34m to 123m tonnes of CO₂ – about 0.9% to 3.4% of the UK's projected total emissions between 2025 and 2035. The lower range of the estimate would depend on greater efficiency in AI models and hardware, and faster decarbonisation of the UK's energy grid (a timeline that feels optimistic given current infrastructure constraints).

Officials from the DSIT appear to have made the revision, first reported by Politico, after an investigation by Foxglove, an independent watchdog, and the Carbon Brief news site said they appeared to be a significant underestimate. Foxglove's head of strategy, Tim Squirrell, said: "The government has a legally binding commitment to reach net zero by 2050. This already sat awkwardly alongside its hell-for-leather embrace of a hyperscale AI datacentre buildout, which unchecked could double the electricity consumption of the entire country."

The situation has now been revealed to be much, much worse, given the fact the government doesn't seem to have done even the most basic arithmetic needed to measure the potential new carbon emissions of these datacentres. Imagine walking into a server room. The hum is constant. The air conditioning fights a losing battle against heat that radiates from racks of processors. That heat is the physical manifestation of carbon emissions, whether you count it or not.

Patrick Galey, the head of investigations for the Global Witness climate campaign, said: "We have a handful of years until our carbon budget is exhausted. To waste what little bandwidth we have left – when 750 million people worldwide lack access to electricity – assisting some of the richest men ever to hone their plagiarism bots would be a historic idiocy that future generations are unlikely to forgive today's leaders for."

The government declined to comment on the record. But the numbers speak for themselves. A 100-fold revision is not a minor correction. It suggests either a fundamental misunderstanding of the technology's energy requirements or a deliberate choice to present optimistic figures that would face less public resistance. Either way, the outcome is the same: policy decisions were made on faulty data.

Other Guardian reporting from earlier in 2026 shows the scale of individual projects. A vast new datacentre to feed Britain's rising demand for artificial intelligence could cause more greenhouse gas emissions than five international airports. Elsham datacentre in Lincolnshire is on course to cost £10bn and its 15 power-hungry computer warehouses are projected to release five times the carbon dioxide of Birmingham airport, including from take-offs and landings.

Documents estimate the datacentre would consume 3.7bn kWh of energy, with annual CO₂ emissions of 857,254 tonnes when running at full tilt. This is based on the current mix of energy sources powering the National Grid. The datacentre will also create so much excess heat that glasshouses are being proposed with capacity to produce more than 10 tonnes of tomatoes a day. That's the physical reality of AI infrastructure: it generates waste heat that requires management, and the energy to manage it generates more emissions.

Global tech firms are struggling to meet their carbon-cutting goals. By 2030, carbon dioxide emissions from AI datacentres will be six times the 2023 level, according to research by the Öko-Institut in Germany. Microsoft recently admitted that five years after it committed to becoming zero carbon by 2030, its total emissions had risen by 23% due to factors including AI expansion. This week Meta signed a 20-year deal with a nuclear power station in Illinois, while Amazon and Google are also investing in nuclear energy to fuel the race for AI dominance.

The UK has set a target to create a virtually carbon-free power system by 2030, an aim that is already in doubt amid concerns over the rising cost of the country's electricity. The letter calls for a framework for calculating the environmental impact of datacentres, as well as requiring developers to fund the construction of renewable energy generation related to their proposals. It also calls for the prevention of "greenwashing," which can include avoiding the construction of new green energy capacity by buying renewable energy certificates.

Whether the UK can actually deliver on both its AI ambitions and its net zero commitments remains an open question. The government has said it is targeting a "rapid build-out" to boost the UK's capacity for building and running AI models. But rapid build-out without accurate emissions accounting is just a faster path to the same problem. Users will pay for the electricity. The climate will pay for the carbon. Whether anyone else pays for the policy failure is the real question.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn