Whitehall's Clashing AI Power Forecasts Undermine Business Planning

Two pillars of the current government's industrial agenda, decarbonising the economy and positioning the UK as a global AI leader, depend on the same finite resource: grid electricity. Yet the departments charged with delivering each objective appear not to have reconciled their assumptions about how much power AI infrastructure will actually consume, as first reported by the Guardian on 26 April 2026.

The result is not merely an academic inconsistency. It is a planning failure with direct consequences for any business making capital decisions around energy procurement, site selection, or grid access.

Where the numbers diverge

The Department for Energy Security and Net Zero (DESNZ) and the Department for Science, Innovation and Technology (DSIT) have each published or relied upon projections for UK datacentre electricity consumption that do not align, according to the Guardian's reporting.

The UK's existing datacentre estate already consumes a material share of national electricity. Industry estimates have placed current datacentre power demand in the range of 4 to 5 gigawatts (GW), a figure that has grown sharply as hyperscale operators and AI-focused facilities have expanded. DSIT's AI Opportunities Action Plan, published in January 2025, committed the government to removing planning barriers and accelerating datacentre builds, implicitly accepting that demand would rise substantially.

However, the energy demand trajectories underpinning DESNZ's net zero modelling and DSIT's AI infrastructure ambitions do not use consistent baselines or growth assumptions, the Guardian reported. The gap matters because even modest differences in projected gigawatt demand translate into large divergences in required generation capacity, grid reinforcement spending, and carbon budgets.

For context, a single large AI training cluster can draw 100 megawatts (MW) or more. Multiple facilities of that scale, which the AI Opportunities Action Plan envisions, would require generation and transmission capacity equivalent to powering a mid-sized city. If one department's models assume significantly lower demand than another's, the infrastructure pipeline will be sized to the wrong number.

What the mismatch means for grid-dependent businesses

The immediate concern is not hypothetical. Grid capacity constraints are already affecting UK industrial and commercial users.

National Grid ESO data has consistently shown a connection queue running into the hundreds of gigawatts, with wait times for new grid connections stretching beyond a decade in some regions. The Electricity Networks Commissioner's report, published in 2023, found that connection delays were a binding constraint on economic growth and recommended halving typical wait times. Progress has been slow.

For SMEs and scale-ups operating in the AI supply chain, or in any sector with significant electricity requirements, the practical effect of misaligned Whitehall forecasts is threefold.

First, businesses cannot rely on government projections to inform their own energy procurement strategies. If DESNZ is modelling lower datacentre demand than DSIT expects, wholesale price forecasts derived from those models may understate future costs.

Second, site selection becomes riskier. Operators evaluating locations for energy-intensive facilities need confidence that grid reinforcement will arrive on schedule. Conflicting departmental assumptions about demand make it harder to judge which regions will see capacity investment prioritised.

Third, planning consent timelines may prove unreliable. The government has designated datacentres as nationally significant infrastructure, streamlining the planning process. But if the energy infrastructure needed to serve those datacentres has not been planned on the same demand assumptions, approvals may outpace the grid's ability to connect new sites.

Net zero versus AI ambition: a policy sequencing problem

The UK's legally binding target is to reach net zero greenhouse gas emissions by 2050, with an interim target of cutting emissions by 68% against 1990 levels by 2030. The government has also committed to decarbonising the electricity grid by 2030, a target reaffirmed by the current administration.

Meeting that target while simultaneously scaling AI datacentre capacity is not inherently contradictory, but it requires careful sequencing. New datacentre demand must be matched by new clean generation capacity; otherwise, the additional load either delays grid decarbonisation or forces reliance on gas-fired backup.

The AI Opportunities Action Plan, led by Matt Clifford, the government's AI adviser, set out ambitions to build sovereign compute capacity and attract private investment into UK datacentre infrastructure. The plan referenced billions of pounds in potential private capital. Yet the energy generation and grid reinforcement required to support that build-out sits under DESNZ's remit, not DSIT's.

The structural problem is one of policy sequencing. DSIT has moved quickly to signal planning support and demand-side ambition for AI infrastructure. DESNZ must now ensure that supply-side energy infrastructure keeps pace, using the same demand assumptions. Without a shared forecast, neither department can sequence its commitments credibly.

The government departments responsible for these two visions do not appear to have agreed on their numbers, the Guardian reported.

That observation, while simple, captures the core risk. Capital allocation decisions across the energy and technology sectors are being made against a backdrop of official forecasts that contradict each other.

What operators should watch next

Several signals will indicate whether Whitehall is moving to resolve the discrepancy.

Cross-departmental demand modelling. The clearest sign of progress would be a jointly published forecast from DESNZ and DSIT setting out agreed baseline and high-growth scenarios for datacentre electricity demand. No such document has been published to date, according to available government records.

National Grid's Strategic Spatial Energy Plan. The forthcoming spatial plan, expected to set out where new generation and transmission capacity should be sited, will reveal whether AI datacentre demand has been incorporated at realistic scale. Operators should scrutinise the demand assumptions underpinning that plan.

Connection reform outcomes. Ofgem and National Grid ESO have been consulting on reforms to the grid connection queue, including a "first ready, first connected" approach. The pace and scope of those reforms will determine how quickly new datacentre and industrial loads can physically access the grid.

Treasury signals on capital. Grid reinforcement is capital-intensive. Whether the Treasury allocates additional public capital, or creates frameworks to mobilise private investment in grid infrastructure at the scale implied by AI ambitions, will be a leading indicator of policy seriousness.

For businesses making decisions today, the prudent approach is to stress-test plans against a range of demand and pricing scenarios rather than relying on any single departmental projection. The fact that Whitehall itself has not settled on a number is, in practical terms, the most important data point available.