What if the data centres straining the grid could also help balance it?

Posted on: 28 October, 2026

Slough is one of the UK’s leading data-centre locations

Artificial intelligence relies on energy-hungry data centres, placing growing pressure on the UK’s electricity grid. But could building them in the right locations and making smarter use of renewable energy turn AI from an environmental problem into part of the solution? Dr Kim Li, a lecturer in Renewable Energy and AI at the University of the Built Environment, explores how smarter planning and more flexible energy use could help turn this challenge into an opportunity.

Artificial intelligence may feel like an entirely digital technology, but every AI model depends on physical infrastructure. Behind every chatbot response, image-generation request, and automated prediction is a data centre full of servers that need electricity, cooling, and a reliable grid connection.

As AI use expands, so does its energy demand. The UK Compute Roadmap forecasts that the country will require at least 6 GW of AI-capable data-centre capacity by 2030, around three times the capacity available today.

This raises an important question: will AI expansion make it harder for the UK to achieve its net-zero ambitions, or could carefully located and flexibly operated data centres become part of the solution?

In other words, could one of AI’s biggest energy challenges also become one of its biggest opportunities?

How much pressure is AI placing on the grid?

The scale of proposed data-centre development is significant. Ofgem recently reported that applications for new demand connections increased from 41 GW to 125 GW in less than a year, with data-centre projects accounting for at least 80 GW of the total.

Importantly, these figures represent connection applications rather than confirmed electricity consumption. Ofgem believes that some proposed projects may be speculative and may never be built. Nevertheless, the increase demonstrates the scale of expected demand and the urgency of planning for it properly.

This is not only a UK concern. The International Energy Agency expects global data-centre electricity consumption to rise from approximately 485 TWh in 2025 to around 950 TWh by 2030. Although that would represent only about 3% of global electricity demand, the effect can be much greater locally because data centres are large and geographically concentrated users of power.

If several large developments seek connections in the same area, they can increase the need for new substations, transmission lines and generation capacity. Poorly planned growth could therefore increase system costs, delay other projects, and make it harder to connect housing, businesses, and low-carbon technologies.

Why location matters: from Slough to Scotland

Windfarm in Scotland

Slough provides a useful example of what can happen when digital growth and energy planning do not keep pace.

The town became one of the UK’s leading data-centre locations because of its proximity to London, strong fibre connectivity, available industrial land and, historically, substantial electricity capacity. Over time, this created a successful cluster of data centres and supporting businesses.

However, recent planning evidence from Slough Borough Council identifies electricity supply as the principal constraint on further data-centre development, with data centres and other projects having to wait for capacity.

The consequences extend beyond the technology sector. A London Assembly investigation found that rapid data-centre development in West London and along the M4 corridor had added pressure to the electricity network. In 2022, some housing developments were being offered connection dates as late as 2037.

Data centres were not the only cause. Demand is also increasing because of housing, electric vehicles, heat pumps and commercial development. Nevertheless, the experience of Slough and West London shows why data centres cannot be planned separately from communities’ wider needs.

Scottish offshore windfarm

Scotland presents a different possibility. It produces large quantities of wind power, but the network cannot always transport all that electricity to consumers elsewhere in Great Britain. When the network becomes constrained, wind farms may be paid to reduce their output while other power stations increase generation closer to where electricity is needed.

NESO reported that balancing the electricity system and managing network restrictions cost £3.125 billion in 2025/26. It estimates that locating an additional 5 GW of flexible electricity demand behind transmission bottlenecks in Scotland could save consumers up to £5 billion.

Locating some future data centres closer to renewable generation could therefore turn additional demand into an advantage. Instead of paying wind farms to reduce their output, appropriately located data centres could use some of that electricity locally.

Making electricity demand more flexible

Data centres traditionally operate continuously because many digital services must be available at all times. A hospital system, financial transaction or live AI service cannot simply be switched off whenever electricity demand rises.

However, not every computing task is equally urgent. Some activities, including parts of AI-model training, data processing, software updates and backups, may be delayed or moved without affecting users.

A flexible data centre could therefore schedule suitable computing tasks for periods when wind or solar generation is abundant. It could reduce or postpone non-urgent activity during times of peak demand and use battery or thermal storage to manage short-term changes. AI-based energy-management systems could also forecast renewable generation, optimise cooling and coordinate computing workloads with grid conditions.

This does not mean that all data-centre electricity use can be moved freely. Computing capacity has high commercial value, and delaying important workloads may not always be practical or economical. Flexibility should therefore be treated as a valuable capability that must be designed into a facility, rather than an automatic feature of every data centre.

The wider electricity system is already moving in this direction. In 2026, NESO expanded its Demand Flexibility Service so that participating consumers and businesses can be rewarded not only for reducing consumption during periods of pressure, but also for increasing consumption when surplus electricity is available.

The principle is simple: rather than continuously adjusting electricity generation to follow demand, some demand can also be adjusted to follow the availability of clean power.

What could this mean for the UK’s net-zero ambitions?

Windfarm in Scotland

Flexible data centres could use renewable electricity that might otherwise be curtailed, reduce some grid-balancing costs and encourage investment in regions with strong renewable resources. The government’s AI Growth Zones programme already reflects some of this thinking by coordinating data-centre development with access to power and regional economic opportunities.

However, data centres are not a substitute for investment in electricity networks, renewable generation and energy storage. Their wider environmental effects, including water consumption, land use, construction impacts and electronic waste, must also be considered.

There is also a question of fairness. If data centres benefit from faster grid connections or reduced electricity costs, the resulting savings and system benefits should not flow solely to technology companies. Policies must ensure that households and other businesses are not left carrying a disproportionate share of infrastructure and balancing costs.

This is why the debate matters to everyone, not only technology companies and energy-system operators. Decisions about data-centre development could influence household energy bills, housing delivery, regional investment, and the speed at which renewable generation can be connected and used effectively.

From grid burden to grid participant

Windfarm in Scotland

The key question is not simply how much electricity AI consumes, but whether its growth is planned in a way that supports or undermines the transition to a cleaner energy system.

Adding a data centre to an already constrained network could increase costs and compete with housing and businesses for limited capacity. One located near abundant renewable generation and able to shift suitable workloads could have a very different effect.

The opportunity is real, but it will not happen automatically. The UK must decide where data centres should be built, how flexibly they should operate and what responsibilities they have to the energy system and surrounding communities.

If location, flexibility and public benefit become essential requirements, part of AI’s energy problem could also become part of the solution.