How much electricity do AI data centres actually use?
AI data centres in the UK now account for roughly 6% of the country’s electricity supply, and that figure has grown fast, up around a third in just two years. Every time a chatbot answers a question or an AI tool generates an image, a server somewhere is drawing power, and a lot of those servers sit in the UK.
Government estimates suggest AI data centres could use somewhere between 8% and 16% of Britain’s total electricity by 2030. That is not a rounding error. It is a genuinely new source of demand on a grid that was not built with this in mind.
Why are AI data centres becoming an energy story, not just a tech one?
Because someone has to pay for the grid upgrades that come with all that extra demand. Ofgem is currently consulting on a Data Centre Commitment Fee, a charge on large data centre developers when they accept a grid connection offer, aimed at stopping speculative projects from clogging up the queue for everyone else.
The consultation matters to ordinary households because network costs, the pipes and wires part of your bill, get shared across everyone connected to the grid. If AI data centres jump the queue without paying their fair share of the upgrade bill, that cost risk lands on domestic customers instead.
Key takeaways
- AI data centres already use around 6% of UK electricity, growing roughly a third in two years.
- Ofgem is consulting on a Data Centre Commitment Fee to stop speculative projects clogging the grid queue.
- The government’s stated aim is that AI Growth Zone sites will not push costs onto household bills.
- Around 100 data centre developers have applied for permanent gas connections, adding a separate pressure on gas demand.
Will AI data centres push up your energy bill?
The government’s official line is no, insisting AI Growth Zones, the priority sites getting fast tracked grid access, will not pass costs on to households. Energy ministers have publicly said AI expansion needs to help bring bills down, not push them up.
Independent analysts are less certain. Grid investment, whoever it is ultimately billed to, tends to show up in network charges over time, and more than 100 data centre developers have applied for permanent gas connections on top of electricity demand, which puts upward pressure on gas prices too, the same wholesale gas prices that already drive most of your winter bill.
What is Ofgem actually proposing?
Under the proposed Data Centre Commitment Fee, large developments would pay a charge when they accept a grid connection offer, giving them a financial reason to only request the capacity they genuinely intend to use. Right now, some projects request huge grid connections speculatively, then never build, while real capacity sits reserved and unused.
The consultation closes in mid September 2026, with a final policy decision to follow. It sits alongside wider work on connection queue reform, since roughly 140 data centre proposals are chasing about 50GW of grid capacity between them, close to Britain’s entire peak electricity demand.
How does electricity demand from AI data centres reach your bill in the first place?
Your bill is made up of several layers: wholesale energy costs, network costs for the pipes and wires, supplier operating costs, and various policy costs like the energy debt allowance. Network costs are the layer AI data centres touch most directly, since connecting a huge new site to the grid often means reinforcing substations, cables and transformers nearby.
Historically, those reinforcement costs get shared across everyone using that part of the network, sometimes including domestic customers, unless the developer requesting the connection pays for it directly. That is exactly the gap Ofgem’s proposed fee is trying to close, by making large data centre developers commit financially before they get a connection offer, rather than after upgrades are already under way.
None of this changes your bill this winter. It is a longer run question about who funds the next decade of grid investment as AI demand keeps climbing, and whether that cost quietly lands on household network charges by default. Ofgem’s other recent consumer protection powers show the same instinct: keep the newer, faster moving parts of the energy system from quietly loading extra cost onto ordinary bills.
Does any of this affect your own tariff choice?
Not directly, not yet. But it is worth watching, because network costs are one of the quieter components of the price cap, alongside wholesale prices and other ongoing Ofgem reforms. Roo’s view: nobody can un invent AI, so the fair fight is over who pays for the electricity it needs, not whether it happens.
In the meantime, the best move for your own bill is unaffected by any of this. Fixing a competitive tariff still protects you from cap movements either way, whatever Ofgem eventually decides about data centre connection fees.
Frequently asked questions about AI data centres and energy bills
- Do AI data centres use more electricity than normal data centres?
- AI workloads, particularly training large models, are typically far more power intensive per server than standard cloud computing or web hosting.
- Is the government trying to stop AI data centres being built?
- No. The stated aim is to manage grid access fairly and quickly, not to block investment, with priority AI Growth Zones designed to attract it.
- Could AI data centres cause power cuts?
- Ofgem’s connection queue reforms exist specifically to manage demand growth so that supply keeps pace, rather than to avoid an immediate shortage.
- Will renewable energy targets be affected by AI data centres?
- Rising electricity and gas demand from data centres adds pressure to decarbonisation timelines, which is part of why the connection and fee reforms are being pushed through now.
- How can I find out if a data centre is planned near me?
- Local council planning portals list major developments, and national grid connection queue data is published by the relevant network operator for your region.
