The UK government's Department for Science, Innovation & Technology (DSIT) has been forced to admit a staggering error in its calculations of the carbon footprint from artificial intelligence datacentres. Originally published in July 2025, the Compute evidence annex estimated that by 2035, the UK's greenhouse gas emissions from AI compute would range from 0.025 to 0.142 million tonnes of CO₂ (MtCO₂) — less than 0.05% of the UK's projected total emissions. However, a correction issued last week reveals the true picture is far bleaker: the cumulative 10-year emissions could be between 34 and 123 MtCO₂, representing 0.9% to 3.4% of the UK's projected total emissions over that period.

The discrepancy amounts to a miscalculation of approximately 100 times the original figures. While DSIT explained that the earlier numbers were annual and the revised ones cover a decade, even this justification fails to mask the scale of the error. Critics argue that such a fundamental mistake undermines confidence in the government's ability to manage the environmental impact of its aggressive AI strategy.

Carbon Brief's deeper analysis

Climate change science and policy research group Carbon Brief has gone a step further, suggesting that even the revised DSIT figures may be optimistic. The government's target of 50gCO₂/kWh by 2030 relies on a clean energy grid dominated by wind, nuclear, hydro, and solar. However, Carbon Brief and environmental campaigners Foxglove have highlighted that any reliance on gas-fired power generation — which has a carbon intensity roughly 10 times that of clean sources — could dramatically increase emissions.

Carbon Brief's calculations indicate that if gas accounts for just 5% of electricity used by datacentres, emissions would be around 3.4 MtCO₂. At 95% gas dependency, that figure rises to 68.1 MtCO₂ — nearly the annual carbon emissions of Sweden. These estimates are based on an Ofgem projection of 20GW of future datacentre electricity demand, set against a peak UK demand of 45GW in February 2026. The 20GW figure itself comes from National Energy System Operator research into future grid connection requirements, highlighting the massive scale of anticipated datacentre growth.

The context of the UK's net zero commitment

The UK has a legally binding commitment to reach net zero carbon emissions by 2050. This target sits awkwardly alongside the government's enthusiastic support for a hyperscale AI datacentre buildout. Foxglove's head of strategy, Tim Squirrell, noted: “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.” He added that the government cannot simply “rubber stamp hundreds of new datacentres” while maintaining its manifesto promise to combat the climate crisis.

The discrepancy in the DSIT figures is not an isolated incident. It reflects a broader pattern of underestimating the environmental costs of digital infrastructure. Datacentres are energy-intensive facilities that require constant power for servers, cooling, and backup systems. With the rise of AI, which demands even greater computational resources, the energy consumption of these facilities is expected to skyrocket.

Current and projected datacentre capacity

Computer Weekly's own research indicates that there is currently around 1.6GW of datacentre capacity in the UK, with just over 8GW in the planning or construction pipeline. This growth is concentrated in regions such as the north of England and Scotland, while London and the M4 corridor account for about 25% of projected capacity. The government's target of 6GW of AI-capable datacentre capacity by 2030 now appears shaky, as operators struggle to balance speed of deployment with environmental constraints.

The implications for the UK's energy grid are profound. A 20GW increase in datacentre demand would require a massive expansion of renewable energy generation capacity, along with grid upgrades and energy storage solutions. Without these, the country risks increased reliance on fossil fuels to meet demand, thereby undermining emissions reduction goals.

Tech industry response and broader implications

The tech industry has been quick to defend the role of AI in driving economic growth and innovation. However, environmental groups are increasingly calling for stricter regulations and transparency around datacentre emissions. The DSIT miscalculation has become a flashpoint in this debate, with critics arguing that the government is prioritising AI development over climate action.

Comparisons with other countries highlight the UK's unique challenges. While nations like Sweden and Norway benefit from abundant hydropower, the UK's grid is still heavily reliant on natural gas. The recent energy crisis and geopolitical tensions have further complicated the transition to clean energy. Meanwhile, major cloud providers like Amazon, Microsoft, and Google have announced their own net zero targets, but the pace of their expansion often clashes with local energy availability.

The DSIT error also raises questions about the accuracy of other government projections. If such a basic calculation can be so wrong, what else might be misestimated? The compute evidence annex was intended to inform policy decisions on AI investment, grid planning, and climate targets. A flawed foundation could lead to misguided strategies and wasted resources.

Technical and policy solutions

Efforts to mitigate datacentre emissions include improving energy efficiency, using renewable energy sources, and implementing carbon offsetting programmes. Some operators are exploring on-site power generation, such as microgrids and battery storage, to reduce strain on the national grid. However, these solutions are expensive and may not scale quickly enough to meet demand.

Policy makers are also considering carbon taxes or emissions limits for large datacentres, similar to those applied to other industrial facilities. The European Union has already introduced energy efficiency directives that affect datacentres, and the UK may follow suit. However, any new regulations must balance environmental goals with the need to remain competitive in the global AI race.

The intersection of AI, energy, and climate change is one of the most pressing challenges of our time. The DSIT miscalculation serves as a stark reminder that accurate data is essential for informed decision-making. As the UK accelerates its AI ambitions, it must ensure that its environmental assessments are rigorous, transparent, and grounded in reality.

In the coming months, further revisions to emissions estimates are likely as more data becomes available. The government has promised to improve its modelling and work with industry stakeholders. But for now, the revelation that AI datacentres could account for up to 3.4% of UK emissions over the next decade has sent shockwaves through both the tech and environmental communities.

Source: ComputerWeekly.com News