As the UK government unveils a new plan to accelerate strategic data centres, the race to become an artificial intelligence powerhouse is raising a critical question: where will all the electricity come from?
Britain’s artificial intelligence ambitions are running into a problem that cannot be solved by better software alone: electricity. Behind every chatbot response, AI-generated image and complex computing task lies a network of power-hungry computers operating inside data centres. As demand for AI services grows, Britain faces the challenge of building the infrastructure needed to support the technology without placing additional pressure on an already constrained electricity grid.
The issue moved further up the government’s agenda on 28 October 2026, when the Treasury announced a new Data Centre Covenant as part of the UK Budget. The initiative aims to accelerate strategically important data centres, improve access to electricity connections and strengthen Britain’s domestic AI capabilities.
The announcement comes as the government seeks to balance technological growth with energy security, affordable electricity and the needs of households and other businesses.
Billions in ambition, gigawatts in demand
The scale of the challenge is becoming clearer. In its Data Centre Covenant, published on 28 October, the government estimated that Britain would need 6 gigawatts (GW) of AI data centre capacity by 2030 to protect its national interests and maintain secure access to AI computing infrastructure.
For context, a gigawatt is one billion watts of power. Six gigawatts of capacity represents a substantial and continuous electricity requirement, although actual consumption will depend on how intensively facilities operate.
The government’s estimate highlights how rapidly digital infrastructure is becoming part of the country’s wider energy planning. Data centres already support online banking, shopping, streaming, business operations and public services. The expansion of AI adds another layer of demand, particularly from facilities running advanced computing chips and large-scale AI models. However, the headline figures require careful interpretation. The government estimated that existing UK data centre electricity demand was around 2GW, according to a written statement published in Parliament on 28 October. The 6GW figure is a target for future AI data centre capacity, not a prediction that electricity consumption will simply triple. The distinction matters because capacity describes how much power infrastructure can supply, while actual electricity consumption varies with usage and operating patterns.
A crowded queue for grid connections
The problem is not simply generating enough electricity. Getting large new facilities connected to the grid has become a major obstacle. On 29 July 2026, energy regulator Ofgem announced proposals to tackle speculative data centre projects occupying space in Britain’s electricity connections queue. It said applications for new demand connections had surged from 41GW to 125GW in less than a year, with data centre projects accounting for at least 80GW of that demand.
These figures represent connection applications, not confirmed developments or electricity that must inevitably be supplied. Some projects may never be built. Nevertheless, the scale of the queue illustrates the pressure facing network planners as data centres compete with other large electricity users for connection capacity.
Ofgem’s proposed measures included a commitment fee for large data centre projects and stronger requirements to demonstrate that developments are progressing. The aim is to prevent speculative schemes from delaying viable projects that are ready to connect.
The government’s latest plan takes this challenge further. Under the Data Centre Covenant, officials intend to prioritise strategically important projects, reallocate capacity released by developments that leave the queue and identify locations where additional capacity could become available before 2030. The government also plans to develop tightly controlled arrangements for temporary on-site power generation in exceptional cases where grid connections cannot be secured in time. The policy signals that access to electricity is no longer just a technical consideration for data centre developers. It is becoming a question of national economic strategy.
Can technology help solve the problem?
Expanding electricity networks and generating capacity will take time. But a trial involving National Grid suggests that smarter management of electricity demand could help ease some of the pressure.
On 2 March 2026, National Grid announced results from a UK-first trial demonstrating how an AI data centre could adjust its electricity use in response to signals from the grid. The trial involved a cluster of 96 NVIDIA Blackwell Ultra GPUs at a Nebius data centre in London. Over five days in December 2025, researchers tested the system against more than 200 simulated grid events. The technology reduced power demand by up to 40% during testing, while critical computing workloads continued to operate.
The results suggest that data centres may not always need to operate at maximum power. Where computing workloads allow it, facilities could reduce consumption during periods of high electricity demand and increase it when the grid has more capacity available. Such flexibility could help make better use of existing infrastructure. However, the trial does not eliminate the need for additional generation and network investment, particularly as more energy-intensive AI facilities come online.
National Grid said the demonstration provided evidence that AI infrastructure could become a flexible resource for the electricity system rather than simply adding inflexible demand.
Who pays for the AI revolution?
For households and businesses, the central concern is whether the expansion of data centres could add to electricity system costs. The government’s new covenant says data centres should pay their fair share of the energy costs they create, rather than shifting those costs onto households, communities and other businesses. It also calls for responsible energy use, grid resilience and support for clean power.
There is a potential economic upside. More data centres could attract investment, create jobs and support British businesses developing AI technology. But those benefits depend partly on where facilities are built, how efficiently they use resources and whether the wider electricity system can accommodate them. The government has also committed to accelerating grid connections for strategic projects while considering the effects on local communities, water resources and energy security.
That makes the debate more complicated than a simple choice between technological progress and protecting the electricity supply. Britain must decide how to support AI growth while ensuring that its costs and benefits are shared fairly.
A race against the clock
The next four years will be crucial as the UK works towards its 2030 AI infrastructure ambitions. The government’s October Budget measures aim to speed up strategic developments, while Ofgem’s connection reforms seek to ensure that limited grid capacity is allocated to projects capable of progressing.
Yet faster connections alone will not guarantee success. Britain will also need sufficient generation, stronger electricity networks, efficient data centres and effective coordination between technology companies, regulators and energy planners. The London trial offers one possible part of the solution: making computing infrastructure more responsive to the electricity system. The government’s new covenant offers another: linking accelerated access to power with expectations around national security, sustainability and public benefit.
Ultimately, Britain’s AI future will depend on more than the computing power it can attract. It will depend on whether the country can deliver reliable electricity at the scale, speed and cost required. The question is no longer just how quickly Britain can develop AI. It is whether its electricity infrastructure can keep pace without leaving households and other businesses to bear the cost.
