Every time someone asks an artificial intelligence chatbot a question, generates an image or uses an AI-powered tool, a vast network of computers works behind the scenes. Those computers need electricity, and as AI becomes more sophisticated, the infrastructure supporting it is facing growing pressure. Britain wants to become a global leader in artificial intelligence. But a recent setback involving Microsoft has exposed an uncomfortable reality: the country’s ambitions for AI may be moving faster than its ability to supply the electricity needed to power it.
On 19 October, Reuters reported that Microsoft had withdrawn from an agreement to anchor a major data centre project in Loughton, Essex, after delays to its electricity connection pushed the expected supply date into the 2030s. The project had been planned to provide 50 megawatts of AI computing capacity for Microsoft’s Azure cloud services. The setback raises a question that could shape Britain’s technology future: can the country expand its AI infrastructure without overwhelming the systems that keep its digital economy running?
The growing appetite for electricity
Data centres are the physical backbone of modern AI. They house powerful processors, store information and run the calculations behind everything from online searches to advanced AI models. Keeping these machines operational also requires cooling systems, networking equipment and reliable backup power.
The scale of that demand is already measurable. According to the Office for National Statistics, Britain’s data centres consumed an estimated 4.5 terawatt-hours (TWh) of electricity in 2024 — around 2% of Great Britain’s total grid electricity consumption of 249.2 TWh. Consumption rose by 1.3 TWh, or 41%, between 2020 and 2024.
That figure covers data centres serving external organisations, including cloud and hosting facilities, rather than every computer server operated by individual businesses. Future demand could be considerably greater. Evidence submitted to Parliament drawing on National Energy System Operator analysis projected data-centre electricity consumption of just over 20 TWh annually by 2030. However, that remains a forecast, not a guaranteed outcome: some proposed projects may never be built, and actual demand will depend on how intensively facilities are used. The challenge is not simply generating more electricity. It is getting enough power to the right locations, at the right time, through infrastructure capable of handling increasingly concentrated demand.
When the grid becomes the bottleneck
The problem is becoming visible in Britain’s electricity connection queue. On 29 July 2026, energy regulator Ofgem announced proposals to tackle speculative data-centre projects occupying space in the queue. It said applications for new electricity demand connections had surged from 41 gigawatts (GW) to 125 GW in less than a year, with data-centre projects accounting for at least 80 GW. These are requested connection capacities, not electricity already being consumed.
The scale of the requests creates a planning dilemma. If too many projects reserve capacity without being ready to proceed, viable developments can face delays. Yet building new substations, transmission lines and other infrastructure takes time, investment and coordination.
The government has already acknowledged the issue. In a policy announcement on grid connections, Energy Minister Michael Shanks said: “We will prioritise the projects, including AI (artificial intelligence) datacentres and industrial sites, that are ready and needed to deliver growth and jobs for communities across Britain.”
The policy challenge is to distinguish projects that can deliver genuine economic benefits from speculative proposals, while ensuring that essential services and other industries are not unfairly disadvantaged.
Who pays for the AI revolution?
Behind the technical debate lies a question about fairness.
Large technology companies need reliable electricity to expand their operations. Meanwhile, households and smaller businesses also depend on affordable, secure energy. Expanding the grid requires investment, and the way those costs are allocated matters to consumers. The environmental implications are equally important. If rising demand is met by additional fossil-fuel generation, it could complicate Britain’s climate ambitions. Renewable electricity can help, but generation must be available when needed, and transmission infrastructure must carry it to the places where it is consumed. Data centres also produce substantial heat and may require water for cooling, depending on their design. Their local environmental impact therefore depends on more than their electricity bills alone.
However, the picture is not entirely negative. AI and digital services can also improve efficiency, from optimising energy networks to reducing waste in industrial processes. The government’s own research into the energy consumption of digital services highlights that digital alternatives can sometimes use less electricity than their physical counterparts.
The real question is whether the benefits of AI can be delivered without allowing its infrastructure demands to grow unchecked.
Can smarter technology solve the problem?
One potential solution is to make data centres more flexible rather than simply building more power infrastructure. In a trial announced by National Grid and its partners, AI infrastructure demonstrated that it could reduce electricity demand from a cluster of 96 NVIDIA Blackwell Ultra processors by more than a third in under a minute, without disrupting critical computing workloads. The demonstration explored how data centres could adjust their power consumption in response to real-time signals from the electricity system.
Such technology could help operators reduce demand during periods of grid pressure and make better use of available capacity. Other options include locating facilities near suitable energy supplies, investing in renewable power, improving cooling efficiency and developing storage systems.
But flexible demand is not a complete solution. Some AI tasks require continuous processing, and grid operators must still ensure that electricity supplies remain reliable. Better technology will need to work alongside investment in physical infrastructure and effective regulation.
A test of Britain’s digital ambitions
Microsoft’s decision to withdraw from the Essex agreement illustrates how electricity infrastructure can become a barrier to major technology investment. It does not mean Britain’s AI ambitions are impossible, but it highlights the gap that can emerge between announcing a project and delivering the power it needs.
For policymakers, the challenge is to balance three priorities: attracting investment, protecting consumers and building a reliable, lower-carbon electricity system.
For technology companies, it means treating energy availability as a central part of AI development rather than a problem to solve after choosing a site.
And for the public, it raises a question that extends beyond data centres: how should Britain distribute the costs and benefits of a technology that is rapidly becoming part of everyday life?
AI may exist in the cloud, but the infrastructure behind it is firmly on the ground. Its future will depend not only on faster processors and more advanced algorithms, but also on whether Britain can supply the power to keep them running. The next stage of the AI revolution may therefore be decided by an unexpected factor — not how intelligent machines become, but how intelligently the country builds the energy system they depend on.
