The conversation around artificial intelligence in the UK has focused almost exclusively on models, datasets, and talent. These are legitimate priorities, but there is a foundational layer beneath all of them that receives far less attention, without which none of the ambition around AI development and deployment can be realised. That layer is power, and the UK does not yet have a coherent strategy for it.
AI workloads are among the most energy-intensive computing tasks ever developed at scale. The infrastructure required to support them – from the data centres that house graphics processing units (GPU) clusters to the grid connections that feed them – is under pressure that the system was not designed to handle. Addressing that pressure is not a technical footnote to the AI agenda. It is a central requirement for the UK to compete seriously in the global AI race.
The scale of the power challenge
The energy demands of AI infrastructure are substantially greater than those of the general cloud computing workloads that preceded them. A single large-scale AI training run can consume as much electricity as a small town uses in a month. As inference workloads scale across the globe and enterprise deployments such as the NHS and banking, the aggregate demand grows further still. Data centre electricity consumption globally is projected to increase significantly over the coming years, with AI workloads accounting for a growing share of that increase.
In the UK, this surge in demand is colliding with grid infrastructure that was not designed for it. Connection queues for new high-voltage supplies, which are often unsuitable for residential use and avoid using local distribution networks, stretch beyond a decade in some areas. However, reforms were introduced in early 2026 to prioritise projects with genuine strategic value and clear delivery timelines, clearing speculative applications that have blocked the queue without any realistic prospect of reaching operation. These reforms are a step in the right direction, but the underlying infrastructure investment required to serve future demand has not yet been committed at the scale needed.
Why compute without power is not a strategy
The UK government’s AI ambitions are well documented. The AI Opportunities Action Plan, published in January 2025, set out a framework for positioning the UK as a leading AI nation. It identified compute infrastructure as a priority, including plans for AI Growth Zones designed to accelerate data centre development in specific locations. The intent is right. The gap is in the power dimension of that infrastructure agenda.
Compute capacity without reliable, affordable, and sustainable power is not deployable capacity. A data centre with grid connection problems, prohibitively high energy costs, or insufficient renewable supply cannot attract the hyperscale and enterprise AI workloads that drive economic value. The UK currently has some of the highest industrial electricity prices in Europe, a structural disadvantage that could affect investment decisions regardless of how attractive other aspects of the market may be.
Closing this gap requires treating power infrastructure for AI as a national strategic priority in the same way that semiconductor supply chains and AI research funding have been treated. That means grid investment, energy pricing reform, and long-term renewable energy policy that gives operators the certainty they need to commit capital at scale.
Renewable energy and the path to sustainable AI
The sustainability dimension of AI’s power demands is significant and cannot be deferred. Training and running large AI models at scale generates substantial carbon emissions if the underlying power supply is not clean. For organisations with net zero commitments, and for governments with legally binding climate targets, this creates a direct tension between AI ambition and environmental obligation.
The data centre industry has responded to this challenge through large-scale procurement of renewable energy via long-term power purchase agreements (PPAs). These agreements provide renewable generators with the revenue certainty needed to secure project financing, enabling new wind and solar capacity to be built that would not otherwise reach financial close.
In practical terms, rather than being a burden on the grid and relying on high-carbon energy, data centres are becoming one of the more important demand-side enablers of the UK’s renewable energy buildout. By purchasing green energy at scale and investing in on-site efficiencies that lower overall carbon intensity, data centres are far from being the problem child depicted by media. They are in fact stimulating renewable generation by providing stable offtake for new wind and solar projects.
The operational challenge is managing variability. Renewable generation fluctuates with weather conditions, and AI workloads, while they can be shaped to some degree through intelligent scheduling, have minimum power requirements that cannot be reduced without affecting service availability. Battery energy storage at the facility level, combined with grid-scale storage development, is the most viable near-term solution to smoothing the mismatch between renewable supply and data centre demand.
The grid reform agenda
Grid infrastructure is the most immediate bottleneck in the UK’s AI power strategy. The connection queue problem is well understood, but the solution requires more than queue management. The transmission and distribution networks themselves need investment to handle the loads that AI infrastructure will place on them over the next decade.
Ofgem and the National Energy System Operator have both identified the need for accelerated grid investment as part of the clean energy transition. The data centre sector’s interests align closely with the broader decarbonisation agenda in this respect. Both require a grid that is larger, smarter, and more flexible than the one that exists today. Coordinating the investment requirements of AI infrastructure with the clean energy transition planning that is already underway is an opportunity that policymakers should not miss.
What a power strategy for AI looks like
A coherent power strategy for AI in the UK would have several components that:
- set clear targets for grid connection timelines for strategic digital infrastructure, with accountability for delivery,
- address industrial energy pricing through a combination of network cost reform and targeted support for electricity-intensive sectors that are critical to national economic strategy,
- provide long-term policy certainty for renewable energy procurement, enabling operators to sign the power purchase agreements that both decarbonise their operations and finance new generation capacity,
- coordinate land use, grid investment, and planning policy to ensure that AI Growth Zones are genuinely powered zones, not just designated areas on a map.
None of this is beyond the UK’s capability. The building blocks exist; private investment in data centre infrastructure is substantial and growing; renewable energy sector is expanding; and policy framework around AI is more developed than in most comparable economies. What is missing is the explicit recognition that power is not a downstream consequence of AI ambition but a precondition for it.
The risk for the UK is not falling behind in AI capability, but in failing to make that capability deployable at scale. The countries that build the most capable, most efficient, and most sustainably powered AI infrastructure will capture disproportionate long-term value from the AI transition. The UK has the assets to be one of them. A power strategy that matches its compute ambitions is the next step.
