DSIT Projects 6GW Consumption by 2030 While Energy Department Estimates Less Than a Tenth of That
A significant and potentially damaging discrepancy has emerged between two UK government departments over the energy demands of artificial intelligence datacenters, with the Department of Science, Innovation and Technology projecting consumption of 6 gigawatts by 2030 while the Department of Energy Security and Net Zero appears to estimate less than a tenth of that figure — a gap that raises serious questions about the coherence of the government's net-zero planning and its ambition to make Britain an AI superpower.
Background
The UK government has made the development of AI infrastructure a central plank of its industrial strategy. The Department of Science, Innovation and Technology published a "UK compute roadmap" setting out plans to attract major investment in AI datacenters, positioning Britain as a global hub for the technology. The roadmap projected that AI datacenters would require 6 gigawatts of electricity by 2030 — a figure that would represent a substantial addition to the UK's total electricity demand and has significant implications for grid planning, renewable energy investment, and the government's legally binding net-zero commitments.
The Department of Energy Security and Net Zero, however, appears to be working from a dramatically different assumption — one that estimates datacenter demand at less than a tenth of DSIT's figure. The discrepancy, reported by The Guardian on 26 April, suggests that two departments with overlapping responsibilities for the UK's energy and technology future are not working from the same evidence base — a failure of coordination that could have serious consequences for infrastructure planning.
The issue is not merely academic. Electricity grid operators, renewable energy developers, and local planning authorities all make long-term investment decisions based on government demand projections. If those projections are wildly inconsistent, the result is either over-investment in grid capacity that is never needed, or under-investment that leaves the country unable to power the AI infrastructure it is trying to attract. DSIT has already revised its carbon emission projections for the AI datacenter sector upward by more than a hundredfold in response to the controversy.
Key Developments
The discrepancy came to light through analysis of published government documents, which revealed that DSIT and DESNZ were using fundamentally different assumptions about the growth trajectory of AI computing demand. DSIT's 6GW figure is based on projections of rapid expansion in large-scale AI training and inference workloads, driven by the global race to develop more powerful AI systems. DESNZ's much lower estimate appears to reflect a more conservative view of how quickly AI datacenter capacity will actually be built and operated in the UK.
The UK's data watchdog, the Information Commissioner's Office, has also been in the news this week after it emerged that its chief, John Edwards, has been off the job since February due to an HR investigation — adding to a sense of regulatory uncertainty around the UK's digital governance landscape. Meanwhile, 28,000 HMRC staff have been equipped with Microsoft Copilot following a trial that suggested the tool could save 26 minutes per worker per day — a deployment that itself raises questions about the energy implications of widespread AI adoption across the public sector.
Why It Matters
The energy demands of AI are not a niche technical concern — they are a central challenge for any government serious about both technological leadership and climate commitments. Data centres already account for approximately 1-2% of global electricity consumption, and that figure is rising rapidly as AI workloads grow. The UK's net-zero target requires a managed transition of the entire energy system, and that transition depends on accurate demand forecasting. A hundredfold revision in carbon emission projections for a single sector is not a rounding error — it is a fundamental failure of joined-up government.
This is the second time in a year that the UK government's AI strategy has been undermined by internal inconsistencies. Unlike the EU, which has developed a comprehensive AI Act with clear provisions on energy efficiency and environmental impact, the UK has taken a more permissive, sector-led approach that relies on voluntary commitments from industry. That approach may be attracting investment, but it is also creating governance gaps that could prove costly — both financially and in terms of the UK's credibility as a responsible AI power.
Local Impact
The energy implications of AI datacenter growth are felt most acutely in the communities where those facilities are built. In Scotland, where renewable energy generation is abundant, there is growing interest in attracting AI datacenter investment — but also concern about the impact on local grid infrastructure and the visual impact of large industrial facilities. In Northern Ireland, where the electricity grid is interconnected with the Republic of Ireland, any significant increase in datacenter demand in Britain could affect energy prices and grid stability across the island. For households already struggling with high energy bills — driven in part by the Iran war's impact on global oil and gas prices — the prospect of additional demand pressure on the grid is a genuine concern.
What's Next
The government is expected to publish a revised version of the UK compute roadmap later this year, which should clarify the energy demand assumptions underpinning its AI strategy. TechUK has upcoming events on regulatory frameworks for physical AI on 30 April, which may shed further light on how the industry is engaging with these questions. Watch for any parliamentary questions or select committee hearings on the DSIT-DESNZ discrepancy in the coming weeks — this is exactly the kind of issue that the Science and Technology Committee is likely to pursue.
Sources: The Guardian, The Register
