In 2026 the UK faces a forecast collision: one future powered by AI datacentres and another powered by a cleaner, more resilient UK energy system. Two big government teams, the DSIT and the DESNZ, are drafting paths forward, but their numbers seem to be moving on different treadmills. The two visions could coexist if they could agree on the timeline. This isn34t a morality tale about technocracy vs ambition; it34s a practical reminder that governance works best when numbers walk hand in hand with policy, not when they play hide and seek with each other.
AI datacentres and the numeric rift in UK energy
The DSIT has been bullish about AI datacentres and envisions a need for about 6GW of electricity by 2030 to feed a growing compute economy. That is a scale that would reshape local grids and the way people think about electricity bills. On the other side of the corridor, DESNZ appears to think the energy bite from datacentres will be far smaller ewer than a tenth of that figure. The discrepancy isn34t a mere mismatch of preferences; it feels like two forecast models sharing a sofa yet disagreeing on who gets the armrest. Meanwhile the public is left to wonder which projection maps onto the future with a straight face for the UK energy system.
UK energy realities in 2026: the mismatch anthem
To put it plainly, the energy estimates show a dramatic divergence: the combined trajectory for the whole commercial services sector is projected to grow by roughly 528MW between 2025 and 2030. That magnitude sounds like a lot, but the DSIT forecast for AI datacentres would power a consumption level nearing 6GW by 2030. If you want a mental image, imagine filling nearly 1.7 million homes with new electricity consumption—an impact that would be felt in bills, grid balancing, and the push for cleaner power sources. The numbers are not about a bad forecast; they are about scale, confidence, and who signs off on what counts as a legitimate projection. The Guardian would later push back, pointing out that the DSIT estimates for AI-related emissions were adjusted upward after scrutiny—an adjustment that hints at how dynamic and fragile a rapid policy story can be. DESNZ notes that its models incorporate AI datacentre emissions within its broader carbon budgeting, even as it continues to explore how to attract clean energy investments to power datacentres responsibly. The tension here isn34t merely academic; it is a reminder that energy policy, climate targets, and the data economy need to align to avoid a misstep that costs both money and momentum for the UK energy system. (For context, independent researchers have urged clear documentation of assumptions; see analyses linked by Carbon Brief.)
In a bid to make sense of it all, some observers like the NGO Foxglove have emphasized the need for clarity. The head of strategy argued that the government34s environmental calculations around AI compute do not inspire confidence when the numbers don34t line up with reality or with each other. This isn34t gleeful finger-pointing; it34s a call for better cross-department collaboration, for rigorous data governance, and for transparent reporting that can survive a quick audit by independent researchers. The idea is simple: if AI datacentres act as a new power-hungry backbone of the economy, then UK energy planning must treat them as such—strictly, realistically, and with audible accountability. The business of forecasting is hard enough without a fog of differing assumptions, he and others suggest.
When forecasts evolve: accountability and the energy compute loop
The story took an even more dramatic turn when DSIT updated its figures for the emissions from AI datacentres compute. What started as a modest projection factor ballooned to a range that shifted the perceived share of AI compute in the nation34s emissions from a tiny fraction to a more tangible slice of the total. DESNZ responded by noting that datacentre emissions are factored into its modelling, including for core budget aims, while the AI Energy Council continues to search for pathways to harmonize investment in compute with clean power. All of this matters not only to why and how we compute but to how we maintain a reliable grid as the country pushes for decarbonisation. The broader point remains: when components of the policy machine disagree, there34s a risk that the whole system becomes slower, more fragile, and less able to respond to real-world signals. The year 2026 thus becomes a reminder that policy alignment—particularly around AI datacentres and UK energy—requires ongoing dialogue, rigorous testing, and, yes, a little humility from those who forecast futures with the best of intentions but sometimes imperfect data.
In January, a FOI request to uncover how AI datacentres were folded into Britain34s emissions picture triggered a broader inspection. DESNZ directed researchers to consult its broader forecasts for the commercial services sector and conceded that it did not hold a separate projection for datacentre growth. The review did not end with a neat conclusion, but it did produce a clearer outcome: more questions, more cross-checks, and a renewed insistence that the government publish what it uses to justify its plans. The public deserves to know what assumptions drive the forecast, how they were derived, and what safeguards exist if those assumptions prove too optimistic or too conservative. It is a call to action for better governance in an era when AI datacentres are increasingly central to national strategy and the electricity grid must be ready to meet that demand without compromising climate targets.
To summarize the arc, the two primary departments—DSIT and DESNZ—have produced numbers that diverge in magnitude, scope, and timing. The later shifts in DSIT34s projected emissions for AI datacentres compute amplified the sense that the policy map needs recalibration. The policy paper that set out to transform the national compute ecosystem through a network of AI datacentre hubs remains ambitious, but the ambition needs to be tethered to credible, collaboratively developed forecasts for UK energy demand and supply. In practice, this means more transparent modelling, open review processes, and clearer lines of accountability for the assumptions behind these numbers. The goal should be a future where AI datacentres can grow while UK energy maintains resilience, affordability, and low carbon intensity—not a future where the numbers drift and the plan drifts with them.
As we head deeper into 2026, the key takeaway is straightforward: align the compute ambitions with energy realities, and align both with climate targets. If AI datacentres are to become a backbone of the economy, the grid must be ready, the policy must be clear, and the public must feel confident that the forecast is a plan they can trust—not a postcard from two departments walking in step to a different beat. That is the kind of alignment that turns speculative growth into practical, measurable progress.
Original article: The Guardian coverage on UK AI datacentre energy forecasts (thank you for the reporting and thoughtful questions). https://www.theguardian.com/technology/2026/uk-ai-datacentres-energy-forecasts
Original article attribution: Thank you to Guardian for highlighting the dataset gaps and the need for cross-department clarity in 2026. Your coverage helps readers understand the stakes of this policy puzzle.
We invite you to share your thoughts in the comments below as we collectively unpack what this misalignment means for policy, energy, and the pace of AI-enabled innovation.
AI datacentres forecasting: practical steps for UK energy alignment
- Publish joint assumptions and make the data sources and modelling methods used to forecast AI datacentre energy demand openly available.
- Establish a cross-department governance board to resolve discrepancies quickly and transparently.
- Commission independent audits of key projections and publish the findings to build trust.
- Develop public dashboards that show projected datacentre energy demand alongside grid readiness and policy milestones.
FAQ about AI datacentres and UK energy
AI datacentres and UK energy: what is the core concern?
The core concern is that DSIT and DESNZ publish notably different energy forecasts for AI datacentres, which can complicate planning and risk management. The goal is to ensure forecasts reflect realistic grid constraints and climate targets and to improve governance around how those forecasts are produced.
How reliable are the DSIT and DESNZ forecasts for AI datacentres?
Both departments rely on evolving modelling with assumptions that can shift as new data arrive. Independent reviews and cross-department collaboration are essential to improve reliability and public trust.
What could be the policy implications if the forecasts diverge?
Prolonged divergence could slow decarbonisation, distort investment decisions, and raise costs for households if the grid isn34t prepared or if policy targets are misaligned with compute growth.
How can the public engage or hold government to account?
Requests under FOI, scrutiny of modelling methods, and open data initiatives can help; clear timelines for publishing core assumptions and annual updates would also improve accountability.
References
- The Guardian: UK departments at odds over energy demands of AI datacentres. https://www.theguardian.com/technology/2026/apr/26/uk-departments-at-odds-over-energy-demands-of-ai-datacentres
- Carbon Brief commentary on dataset questions and the need for cross-department clarity (contextual support).
- International energy context on data centres and energy use (IEA): https://www.iea.org/reports/data-centres-and-energy