AI Infrastructure and Data Centres: Who Pays for the Power Behind Artificial Intelligence?

The AI Funding Files: Following the Money, Examining the Impact

Artificial intelligence may appear to exist in a digital world, but it depends on physical infrastructure. Behind every AI-generated image, music track, chatbot response and automated service are computers, servers, data storage, networking equipment and the electricity needed to keep them running.

Much of this infrastructure is housed in data centres: specialist facilities that store, process and transmit digital information. As AI systems become more widely used, the demand for computing power may increase, putting additional pressure on electricity networks, water resources and the places where new facilities are built.

On 28 October 2026, the UK Government published its Data Centre Covenant. It sets out expectations for the sector around energy costs, water efficiency, national security, local communities and the wider benefits of investment.

The central question is not simply whether Britain should build more data centres. It is how the country can support useful innovation while ensuring that the financial, environmental and infrastructure costs are properly understood and fairly managed.

1. What is a data centre?

A data centre is a facility containing computer systems and supporting infrastructure used to store, process and distribute information. It may support online banking, streaming, shopping, public services, cloud computing, scientific research or AI applications.

Not every data centre is built specifically for AI. Some support general digital services, while others provide the high-performance computing capacity needed to train or operate advanced AI systems.

AI workloads can require large numbers of specialised processors operating together. These systems generate heat, so facilities also need cooling, power distribution, backup systems and reliable network connections.

This means that AI growth is not only a software issue. It is also an energy, engineering, planning and environmental issue.

2. How much electricity do data centres use?

A UK parliamentary statement published on 28 October 2026 estimated current UK data centre electricity demand at around 2 gigawatts (GW). One gigawatt is one billion watts of power. The statement said future requirements remain uncertain but could increase significantly as demand for computing and AI grows.

A House of Commons Library briefing published in May 2026 reported that the UK had approximately 1.6 GW of data centre capacity in 2024. It also summarised preliminary government analysis suggesting capacity could rise to between 3.3 GW and 6.3 GW by 2030, depending on policy interventions.

These figures describe different measures and should not be confused. Capacity is not the same as actual electricity consumption at a particular moment, and projections are not guaranteed outcomes.

Nevertheless, the direction of travel raises an important planning question: how will the UK provide reliable electricity for new digital infrastructure while also meeting the needs of homes, hospitals, transport, industry and the transition to cleaner energy?

Sources: Parliamentary statement on data centres, energy and water; House of Commons Library: Data centres — planning policy, sustainability and resilience.

3. Why does AI create an infrastructure challenge?

AI computing demand can be concentrated in particular facilities and locations. A large data centre may need a continuous, high-capacity electricity supply, and its connection may require new substations, cables or other network improvements.

If a project cannot obtain a grid connection in time, its developer may have to wait, change location or explore temporary power arrangements. The surrounding electricity system must also maintain security of supply and manage demand from other users.

The Government has already been reforming the grid connections process to reduce speculative applications and prioritise viable projects that are ready to proceed. Its March 2026 announcement said the queue for demand connections to the transmission network had grown by 460% in the six months to June 2025, contributing to waits of up to 15 years for some projects.

Those figures describe the connection queue and the delays facing projects at that time; they do not mean every data centre must wait 15 years.

The policy challenge is to connect projects that provide genuine value without allowing speculative demand to delay other important infrastructure, including industrial sites and essential public services.

Source: Government announcement on grid connection reforms, 11 March 2026.

4. Who pays for the electricity infrastructure?

One of the most important questions is how the costs created by new electricity demand are shared.

Connecting a major facility can involve network upgrades and other system costs. The precise costs and who ultimately pays depend on the connection, the wider electricity system, applicable rules and the arrangements made for the project.

The Government's Data Centre Covenant states that data centres should pay their fair share of energy-system costs, so households, communities and other businesses do not pick up costs created by their growth.

The covenant also argues that additional demand can help spread some fixed energy-system costs across a larger base of users. But that potential benefit depends on how new demand interacts with the electricity system, how much infrastructure must be built and how costs are allocated.

It would therefore be misleading to assume that every new data centre will automatically lower household bills. Equally, it would be misleading to assume that all infrastructure costs will automatically be passed to households.

The public needs transparent information about the costs attributable to large new developments, the contributions made by operators and the effects on the wider system.

Source: The Data Centre Covenant.

5. The race to secure grid connections

The Government intends to accelerate connections for strategically important data centres that meet the Data Centre Covenant's requirements. Measures outlined in the covenant include reserving and reallocating available grid capacity, identifying additional capacity and facilitating connections for selected projects before 2030.

For exceptionally strategic projects, the Government also plans to develop a tightly controlled framework for temporary on-site generation where other measures cannot secure the required capacity in time.

These proposals aim to help the UK develop AI infrastructure more quickly. They also raise questions about fairness and priorities.

• How will a project be judged to be strategically important?
• How will the Government compare the value of a data centre with other projects competing for limited grid capacity?
• What evidence will show that the facility is ready to proceed and can deliver the promised benefits?
• How will the effects on local communities, energy security and clean-power objectives be assessed?

Fast-tracking can be useful when a project has a clear public or economic benefit. But prioritisation should be based on published criteria and credible evidence, not simply the size of a company's investment announcement.

6. Water: the other resource behind AI

Data centres also need to manage heat. Depending on the facility and cooling technology, this may involve water use, electricity use or a combination of different systems.

The relationship between water and energy is complicated. A cooling method that reduces electricity demand may use more water, while another method may reduce water use but require more energy. The most suitable approach depends on local conditions, facility design and the available resources.

The Government's October 2026 parliamentary statement estimated that data centres currently account for less than 0.2% of non-household water consumption in England. It also warned that future growth could create more significant local pressures, particularly in areas already experiencing water stress or higher demand during hotter weather.

A small national percentage does not mean every local impact is small. Water demand is location-specific, and the effect of a new facility depends on the availability of sustainable supplies and the capacity of local infrastructure.

The Government says water companies are not required to supply data centres with water and may refuse a connection where resources are insufficient.

Developers should therefore consider water availability early, explore efficient cooling and alternative sources where appropriate, and avoid assuming that a reliable water supply will always be available.

Sources: Parliamentary statement on water and energy; Parliamentary answer on potential local water impacts, 17 September 2026.

7. The Data Centre Covenant: what is changing?

The Data Centre Covenant, published on 28 October 2026, sets out standards that data centres are encouraged to meet. Strategic projects must meet the relevant requirements to qualify for additional government action to accelerate development.

The covenant focuses on several areas:

• Energy costs: operators should pay their fair share of the energy-system costs they create.
• Grid resilience: projects should support security of supply, flexibility and clean-power objectives.
• Water efficiency: developers should minimise water demand and consider recycled or alternative sources where appropriate.
• Local communities: development should consider local resources and deliver wider benefits.
• UK jobs and supply chains: projects should contribute to employment, skills and domestic capability.
• National security and resilience: strategically important infrastructure should support the UK's ability to maintain critical digital services.
• Efficient resource use: developers should consider energy efficiency, waste heat recovery and opportunities to reuse resources.

The covenant is a policy framework, not proof that every proposed facility has already met these expectations. The effectiveness of the approach will depend on how projects are assessed, how commitments are monitored and what happens when standards are not met.

Source: The Data Centre Covenant — HM Treasury, published 28 October 2026.

8. Can data centres benefit local communities?

Data centre investment can create construction work, technical roles, supply-chain opportunities and demand for specialist services. Facilities can also provide infrastructure that supports research, businesses and digital public services.

However, data centres are highly automated. Their physical size and investment value do not necessarily translate into large numbers of permanent jobs once construction has finished.

A House of Commons Library briefing noted that a proposed £10 billion campus in Blyth, Northumberland, which could house up to ten data centres, was expected to create around 400 full-time on-site jobs.

This example illustrates why local economic claims need careful examination. A major development may involve substantial capital spending but have a more limited direct workforce than the headline investment suggests. Wider employment may arise through construction, suppliers and businesses that use the infrastructure, but those effects should be measured rather than assumed.

Communities should be able to ask what jobs will be created, what skills are needed, how local people can access those opportunities and whether the benefits will continue after construction.

9. Can data centres support cleaner energy?

AI infrastructure creates additional electricity demand, but that does not automatically mean the demand must be met by fossil fuels. The outcome depends on the generation mix, grid capacity, storage, flexibility and the timing and location of electricity use.

The Government's covenant expects data centres to support clean power and a resilient energy system. It also highlights the potential value of flexible demand and the recovery and reuse of waste heat.

Waste heat from a data centre may be useful to nearby buildings or heat networks where the temperature, distance, demand and infrastructure make reuse practical. It is not a universal solution, and any proposal must be assessed for its technical and economic viability.

Operators should publish meaningful information about their electricity use, energy efficiency, carbon impacts and contribution to system flexibility. Where they claim to use renewable electricity, the details of those arrangements matter; a headline claim should not substitute for a clear account of the facility's actual energy needs and impacts.

Source: Government statement on data centre sustainability, transparency and resource reuse.

10. Why transparency matters

The Government has encouraged data centre operators to report environmental measures annually, including total and peak water use, Water Usage Effectiveness (WUE), Power Usage Effectiveness (PUE), Energy Reuse Factor (ERF), carbon information and the facility's ability to provide electricity-system flexibility.

These measures can help regulators, local authorities, water companies and communities understand how facilities use resources. However, meaningful comparison requires consistent definitions, clear boundaries and reliable reporting.

The Government has said it will work with industry and regulators on consistent reporting arrangements and will consider a statutory framework if voluntary disclosure is insufficient.

This is an important accountability issue. Without comparable information, communities may struggle to assess environmental claims or understand the cumulative effects of multiple facilities in the same area.

Useful reporting should allow people to distinguish between a company's stated ambition, a facility's measured performance and an independently verified result.

11. What could go wrong?

• Household cost concerns: unclear allocation of infrastructure costs could undermine confidence in the fairness of the system.
• Local water pressure: several developments in a water-stressed area could create challenges even if national consumption figures appear small.
• Grid bottlenecks: rapid demand growth could compete with other essential infrastructure for connections and capacity.
• Environmental trade-offs: reducing water use may increase energy demand, or vice versa, if cooling choices are not assessed together.
• Overstated job claims: large investment figures could obscure the number and quality of long-term jobs actually created.
• Weak transparency: voluntary reporting may not provide enough consistent information for meaningful public scrutiny.
• Stranded infrastructure: facilities built on optimistic demand forecasts could become underused or commercially unviable.
• Unequal benefits: national economic gains may not be felt by the communities providing land, infrastructure or resources.

These are potential risks to monitor, not evidence that every data centre project will cause these outcomes. Good planning and transparent oversight can help identify and reduce them.

12. What alternatives should be considered?

Building more infrastructure is not the only way to meet growing computing needs. Government, industry and researchers can also consider measures that improve the efficiency of existing systems.

• More efficient AI: use models and computing methods that achieve the required result with fewer resources where practical.
• Better utilisation: improve the use of existing computing capacity before assuming that every increase in demand requires a new facility.
• Flexible demand: where workloads permit, shift non-urgent computing tasks to times or locations with more available capacity or cleaner electricity.
• Careful site selection: consider grid availability, water resources, environmental conditions and local infrastructure before development begins.
• Alternative cooling: assess water-efficient technologies and suitable recycled water sources without creating disproportionate energy or environmental costs.
• Waste heat reuse: examine whether nearby buildings or industries can make practical use of heat that would otherwise be rejected.
• Public-interest assessment: establish whether the proposed AI application solves a real problem and whether a less resource-intensive approach could work.

Efficiency does not eliminate the need for infrastructure. But it can help ensure that new capacity is developed where it is genuinely needed and that the resources used produce meaningful value.

13. Questions the public should be asking

• How much electricity will each proposed facility use at peak demand and over a typical year?
• Who pays for the grid connection and any wider network upgrades?
• How will the project affect household bills and other electricity users?
• What water sources and cooling systems will be used, and how will local impacts be assessed?
• How many permanent jobs will be created, and what training will be available to local residents?
• What environmental data will the operator publish, and will it be independently checked?
• Will the facility support clean electricity, flexibility and practical heat reuse where feasible?
• What benefits will local communities receive, and how will those benefits be measured?
• What happens if a developer fails to meet its commitments or the expected demand does not materialise?
• How will the Government balance AI infrastructure against other essential demands on land, power and water?

Conclusion: the digital future still has a physical cost

AI depends on physical infrastructure that consumes electricity, requires cooling and connects to wider networks. As Britain seeks to expand its AI capabilities, data centres will be an increasingly important part of the country's economic and digital strategy.

The Government's Data Centre Covenant recognises that growth must be considered alongside energy costs, water resources, environmental performance, national security and community benefits. Its real value will depend on how effectively these expectations are applied and monitored.

Investment in AI infrastructure should not be judged by the size of the buildings or the amount of computing power installed alone. It should also be judged by the services it enables, the jobs and skills it supports, the resources it consumes and the fairness of the costs it creates.

The question is not simply how much AI Britain can power. It is whether the country can power its digital future responsibly, transparently and fairly.

The AI Funding Files examines public commitments, published evidence and the questions that should be asked as AI policy develops. Estimates and forecasts are not guarantees, and policy announcements are not proof of results.