UK departments disagree on how much electricity— and carbon—AI datacentres will need by 2030, raising questions over planning for net zero.
Britain is trying to balance two competing ambitions: rapid AI growth and credible climate accounting.
Two departments, two numbers
A dispute is emerging inside the UK government over the scale of energy demand from AI datacentres—exactly the kind of planning mismatch that can ripple through climate targets. grid investment and public trust.. The Department of Science, Innovation and Technology (DSIT) has suggested AI datacentres could require 6GW of electricity by 2030.. By contrast. the Department of Energy Security and Net Zero (DESNZ) appears to be projecting far lower electricity growth for the segment. implying a much smaller climate and infrastructure burden.
For readers, the difference is not academic.. Datacentres sit at the intersection of two policy arenas: energy security and industrial strategy.. When the numbers diverge. so do the decisions about where power should come from. how quickly the grid must expand. and whether emissions calculations reflect reality on the ground.
What the disagreement means for climate targets
DESNZ is responsible for the UK’s carbon budget growth and delivery plan—its road map for reaching international climate commitments.. In January. an environmental organisation (Foxglove) asked DESNZ for an environmental impact assessment request. essentially challenging how AI datacentres were incorporated into emissions projections.
DESNZ’s response pointed researchers toward broader forecasts for Britain’s “commercial services” sector energy use. while saying it did not maintain separate projections specifically for datacentre growth.. Those broader forecasts. as described in the reporting. suggest that energy demand for the entire sector would rise by 528MW between 2025 and 2030—roughly equivalent to adding the electricity consumption of 1.7 million homes by the end of the decade.
DSIT’s compute roadmap uses a different approach and arrives at a far larger requirement.. DSIT’s plan for “AI-capable datacentre capacity” forecasts at least 6GW by 2030. meant to be delivered through multiple AI growth zones designed to attract investment.. Each zone would require at least 500MW of electricity—close to the total increase DESNZ appears to be projecting for an entire commercial sector.
This is where the tension becomes hard to ignore: if the government is publicly committed to building enough capacity for 6GW by 2030, then the question becomes whether the supporting emissions and energy calculations are calibrated to match the same reality.
How forecasts shifted and why it matters
The discrepancy came into sharper focus when DSIT. after questions were raised. revised its published figures for the carbon emissions associated with additional AI computing capacity.. Earlier projections—set out in an annex to the compute roadmap—were described as spanning between 0.025 and 0.142 million tonnes of carbon equivalent (MtCO₂). figures framed as well below one-twentieth of one percent of the country’s projected emissions.
Those low values were questioned previously, and the annex was later removed from the government website.. After renewed scrutiny. DSIT updated the numbers. presenting a wider range: cumulative 10-year greenhouse gas emissions from AI compute could range from 34 to 123 MtCO₂. roughly 0.9% to 3.4% of the UK’s projected total emissions over that decade.
In grid terms, that adjustment matters because it changes the political arithmetic of decarbonisation.. If emissions from AI compute are more significant than earlier estimates suggested. then progress depends not only on how quickly datacentres expand. but also on how fast the electricity grid gets cleaner.. DSIT’s statement indicated that. if national grid decarbonisation plans work as intended. datacentre emissions would fall toward the lower end of the updated range.
From an editorial perspective. the core story is less about whether one forecast is perfectly precise and more about whether decision-makers are working from consistent assumptions.. When internal projections move by large factors. it raises a practical concern: policies may be designed on one set of expectations. then recalibrated only after scrutiny.
The human stakes behind the spreadsheet
Datacentres are not just a technology story; they’re a real-world infrastructure load.. New power demand can tighten timelines for grid upgrades. raise cost pressures. and shape public debate over land use. local environmental impacts. and the pace of renewable deployment.. In the UK. where climate governance already involves complex carbon budget tracking. the credibility of emissions accounting affects more than policy memos—it affects how confident people feel that the transition is on track.
There is also an occupational impact. Grid planners, energy developers and local authorities need stability in demand forecasts to make timely investment decisions. If those forecasts are revised repeatedly, projects can be delayed, resources misallocated, or risks recalculated midstream.
Why “misalignment” could be structural
Researchers and advocacy groups have described the mismatch as a sign of “misalignment. ” suggesting it could reflect either incompetence or “magical thinking” about AI and big tech—an interpretation that points to a broader governance challenge: AI expansion is often treated like an economic sprint. while decarbonisation is a multi-year construction effort.
DESNZ says datacentre emissions are factored into modelling. including for carbon budget 7. and it also references work through an AI Energy Council aimed at attracting investment and supporting clean power for datacentres.. Carbon budget 7 is expected to be released this summer. which will be a key moment for how these competing narratives are reconciled.
What comes next for the UK’s AI-and-net-zero plan
A clear question now sits at the centre of UK planning: will government departments converge on a shared estimate of how much power AI compute will require, and will carbon accounting match that estimate across the entire policy chain—from grid decarbonisation to datacentre deployment?
If the updated emissions range becomes the basis for future modelling. it could reshape the urgency of clean power build-out and influence how quickly capacity constraints are treated as an AI-enabling bottleneck rather than a climate risk.. Either way. the political lesson is straightforward: without consistent numbers. the UK risks making decisions that are technically ambitious but environmentally out of step.