Data Centres' Dilution Dividend

Accelerated data centre build-out could save British bills £2bn a year by 2030

The summary

Building data centres will bring down household electricity bills. British electricity bills carry a large fixed cost, and more demand dilutes the cost that everyone else bears.

Data centres are the most effective form of new demand cost dilution because they run at high load, pay the full tariff with no industrial relief, and connect at the voltage tier where the fixed cost per site is highest. The more we build, the lower bills will be: the government's 6 GW ambition by 2030 would save British consumers around £2 billion a year. On current pace we are falling short.

The response should be greater urgency and better ideas:

1. accelerate grid-connection reform with auctions so valuable demand can pay its way and move ahead in the queue,

2. broaden the AI Growth Zone model into a standing route to consent rather than a one-off programme,

3. and enable bring-your-own-power so developers who fund their own firm generation can build ahead of the public reinforcement timetable.

The argument

The transmission network, the distribution networks, the system operator’s balancing function and a set of legacy policy liabilities recover their revenue through envelopes that the regulator Ofgem sets for a five-year price-control period. All four of those envelopes are growing sharply out to 2030: transmission from roughly £9 billion in 2026/27 to £14 billion in 2030/31 under the new RIIO-ET3 price control, balancing from around £2.5 billion today to approximately £3.2 billion, and the distribution and policy envelopes on the same upward path. Each envelope is fixed in total once set, so when demand grows the same pot is spread across more units and the rate each consumer pays per megawatt-hour falls.

If UK data centre capacity grows from 1.6 gigawatts (GW) today to 6 GW by 2030, the government’s stated ambition in the UK Compute Roadmap, that adds 45 terawatt-hours (TWh) of high-load-factor demand to a Great Britain base of 365 TWh. On the electricity system operator’s own 2030/31 envelopes, the average transmission charge dilutes by £3.54 per megawatt-hour, the balancing charge by £0.97 per megawatt-hour, and the forward-looking portion of the distribution charge by £0.50 per megawatt-hour. The saving to existing consumers is approximately £1.8 billion per year, rising to £2.2 billion if 1 GW of the build-out lands behind the Scotland-England transmission boundary (NESO’s boundary B6) and absorbs otherwise-curtailed wind. Per average household at 3,400 kilowatt-hours per year: £16 to £20.

Data centres are the strongest dilution lever in the demand pipeline. Load factors of 0.8 to 0.95, full tariff (no British Industry Supercharger relief for energy-intensive manufacturers, no British Industrial Competitiveness Scheme relief), connection at Extra High Voltage. Per connected megawatt, they draw four to five times as much energy as a typical industrial site, and they pick up a standard-tariff share of every fixed cost. The envelope is fixed by Ofgem for the five-year price-control period and reflects sunk network assets and approved reinforcement; a new data centre pays for the reinforcement its own connection causes through a connection agreement, and so does not increase the socialised envelope.

6 GW by 2030 is a stretch. It is a policy ambition rather than an independent forecast, and reaching it would require the planning, grid-connection and AI Growth Zone reforms to deliver on a scale and timeline that the system as currently configured is not capable of achieving, absent a drastic re-application of effort and new solutions. A central case of 4 GW delivers roughly three-quarters of the saving. A slippage case of 3 GW, roughly consistent with the bottom of the Compute Roadmap’s 3.3 to 6.3 GW range, delivers about half. This is why the headline is framed around acceleration: the saving scales with how much of the 6 GW actually lands by 2030, not whether the number is eventually hit. Annex A sets out the provenance of the capacity numbers.

The standard objection is that new demand raises wholesale prices. It does, at first order, but the effect is contained as long as procurement keeps pace with demand, and it fades over a two-to-four-year horizon as generators respond. The dilution benefit runs for the life of the network assets.

Annexes

1. The envelopes

Transmission charges, known in the charging methodologies as the Transmission Network Use of System (TNUoS) charge, recover the allowed revenue Ofgem sets for the three transmission owners and the national electricity system operator (NESO). Under the RIIO-ET3 price control that runs from 2026 to 2031, the total transmission envelope rises from £8.92 billion in 2026/27 to £13.63 billion in 2030/31 (NESO Five-Year View, September 2025, Table 24, p.44). Demand pays 85.76 per cent in year one and 87.87 per cent in year five. Demand revenue recovered from residual charges rises from £7.52 billion to £11.75 billion over the same five years.

Balancing charges, in the methodologies the Balancing Services Use of System (BSUoS) charge, recover the system operator’s day-to-day balancing costs. Since April 2023 (under industry modification CMP308) balancing charges fall entirely on Final Demand, as a flat rate per megawatt-hour set six-monthly and nine months in advance. Recent rates: £7.63 per megawatt-hour (April to September 2024), £11.93 per megawatt-hour (draft October 2026 to March 2027). Projected 2030/31 envelope: approximately £3.2 billion, driven by constraint costs which reached £1.9 billion in 2024/25, 71 per cent of total balancing costs (NESO Annual Balancing Costs Report 2025).

Distribution charges, formally the Distribution Use of System (DUoS) charge, recover the revenue Ofgem allows the fourteen Distribution Network Operators under the RIIO-ED2 price control, approximately £22 billion of investment over 2023 to 2028. The distribution charge has a residual component (roughly 50 per cent of the total, approximately £2.2 billion per year) that moved to fixed daily charges at the Targeted Charging Review implementation in April 2022, banded by voltage and available supply capacity for non-domestic customers, and a forward-looking component recovered through a pence-per-kilowatt-hour unit rate.

The distribution envelope cannot be stated as a clean 2030/31 number because Ofgem has not yet published the RIIO-ED3 price control (Draft Determinations due 2027, Final Determinations expected December 2027). The RIIO-ED2 five-year totex allowance of £22.2bn (2020/21 prices) is 17 per cent above average annual RIIO-ED1 outturn. The ED3 Framework Decision and Sector Specific Methodology Consultation both point to a further substantial step-up driven by clean-power 2030 and electrification, but that step-up is directional, not quantified until late 2027.

The policy-levy figure uses the Office for Budget Responsibility’s “Environmental levies” line, which aggregates Contracts for Difference, the Renewables Obligation, the Feed-in Tariff, the Warm Home Discount, the Capacity Market and the new Sizewell C Regulated Asset Base levy. The Energy Company Obligation sits outside that line as a supplier obligation; ECO4 ends 31 December 2026 and is replaced by the Exchequer-funded Warm Homes Plan, which removes approximately £1bn per year from the billpayer envelope from 2027. From 1 April 2026 the government has also moved 75 per cent of the domestic Renewables Obligation share onto the Exchequer for three years, worth approximately £2.3bn per year to domestic bills. The OBR total is gross of both transfers.

2. The dilution arithmetic

Great Britain’s final electricity demand in 2026/27 is 300 TWh (DUKES 2025 Table 5.1), already including 15 TWh from existing UK data centre capacity, most of it older colocation stock running at a 70 to 75 per cent IT load factor. Adding 4.4 GW of new-build capacity between now and 2030 (1.6 GW today to 6 GW) at an 85 per cent load factor and a Power Usage Effectiveness (the ratio of total facility power draw to the power draw of the computing equipment itself) of 1.40 delivers 45 TWh of additional annual consumption. Expanded for heat pumps and wider electrification, the 2030/31 comparison is 365 TWh without further data centre growth versus 410 TWh with it.

Transmission. Applied to the 2030/31 demand residual envelope of £11.75 billion, the implied average charge falls from £32.19 per megawatt-hour (at 365 TWh) to £28.66 per megawatt-hour (at 410 TWh), a saving of £3.54 per megawatt-hour. Across the 365 TWh of demand that would exist anyway, the national saving is approximately £1.29 billion per year.

Balancing. On a 2030/31 envelope of £3.2 billion, the rate falls from £8.77 per megawatt-hour to £7.80 per megawatt-hour, a saving of £0.97 per megawatt-hour. Across 365 TWh: approximately £354 million per year.

Distribution. Residual charges are fixed per connection and dilute through the per-site denominator effect. A hyperscale campus occupies a full Extra High Voltage Band 4 slot at approximately £1.065 million per year in fixed distribution residual in 2019 nominal terms (Ofgem, Targeted Charging Review Decision, November 2019, Table 7, North-East banding), with subsequent network-operator uplifts bringing the current figure materially higher. A 6 GW build-out covers 35 to 45 new Band 4 sites, lowering the residual allocation each existing site in the same band faces. The forward-looking portion of the distribution charge dilutes through the same per-megawatt-hour denominator effect as the transmission charge. Combined distribution saving by 2030/31: approximately £200 million per year.

Combined. £1.29 billion transmission + £354 million balancing + £200 million distribution = approximately £1.84 billion per year in fixed-charge dilution by 2030/31.

3. Why data centres dilute better than the alternatives

Three features of data centre demand matter.

Load factors sit at 0.8 to 0.95, against 0.15 to 0.25 for heat-pump households, 0.05 to 0.10 for domestic electric vehicles, and 0.30 to 0.45 for new electric arc furnaces. Per connected megawatt, a data centre draws four to five times as much energy as a typical industrial site. That multiplies the per-megawatt dilution effect.

Data centres pay full tariff. They are excluded from the British Industry Supercharger, which exempts approximately 500 energy-intensive manufacturers from 60 to 90 per cent of network and policy charges (DESNZ, Network Charging Compensation Scheme eligibility, 2024), and from the British Industrial Competitiveness Scheme that launches in 2027. Every megawatt-hour they consume picks up a standard-tariff share of fixed network and balancing costs.

They connect at 132 kilovolts or above, which places every 100-megawatt site in the highest Extra High Voltage distribution band (under the methodology known as the Extra High Voltage Distribution Charging Methodology, or EDCM). Existing sites in the same band see a proportionally lower allocation as new hyperscale sites are added.

As such, data centres are the strongest dilution lever per connected gigawatt in the current demand-side pipeline.

4. The household number

The system operator’s Five-Year View, page 8, states that the transmission cost for the average domestic household is forecast to be £93.48 in 2026/27, equivalent to 10.6 per cent of the typical annual electricity bill of approximately £882. At a typical domestic consumption of 3,400 kilowatt-hours (Ofgem Typical Domestic Consumption Values, 2026), a £3.54 per megawatt-hour reduction in the 2030/31 average transmission rate relative to a no-growth counterfactual delivers approximately £12 per household per year on transmission charges alone.

The balancing charge dilutes at the household meter in exactly the same way, because it is a flat rate per megawatt-hour and every household pays it on every kilowatt-hour consumed: the £0.97 per megawatt-hour reduction delivers approximately £3 per household. Distribution is more mixed. The non-domestic residual bands do not flow through to domestic bills, because the 2019 Targeted Charging Review set domestic residuals as a single daily standing charge rather than a banded one. The forward-looking half of distribution charges, recovered through pence-per-kilowatt-hour unit rates, dilutes on the same denominator effect and passes through to household unit rates. Of the £200 million distribution saving, approximately half reaches domestic bills, worth a further £1 per household.

Base figure: approximately £16 per household per year from fixed-charge dilution by 2030/31.

5. The constraint upside

A further effect operates if any of the build-out lands behind the Scotland-England transmission boundary, the constraint NESO labels B6. Curtailed Scottish wind in 2023 exceeded 8 TWh (NESO balancing data). A gigawatt of firm data centre load behind the boundary can absorb up to 7.4 TWh per year in principle, but perfect coincidence with curtailment events does not hold: wind capacity factors sit around 40 per cent, and constraint windows are concentrated in a subset of those hours. A realistic coincidence factor of 60 to 70 per cent delivers approximately 4 to 5 TWh of curtailment avoided per gigawatt sited north of the boundary, worth £360 million to £450 million per year at an average curtailment cost of £90 per megawatt-hour (averaged across 2023-2024 balancing actions). Spread across all Great Britain demand, approximately £3 to £4 per household.

Total household range at the 6 GW ambition: £16 to £20 per year, depending on whether 1 GW of the build-out lands north of the boundary. National saving: £1.8 to £2.2 billion per year.

6. The wholesale-price objection

Adding 4.4 GW of flat demand raises wholesale prices at first order, because it raises the running hours of marginal plant, mostly Combined Cycle Gas Turbine stations (the gas-fired power stations that set the market-clearing price for most hours of the year). If supply capacity is static, the short-run effect is approximately £5 to £10 per megawatt-hour. If supply entry keeps pace, the effect sits at the lower end of that range and fades within two to four years as generators respond: higher expected running hours lower Capacity Market bid prices (payments made to generators for being available during stress periods) and Contracts for Difference strike prices (the guaranteed prices paid to low-carbon generators under long-term government contracts), and a flat round-the-clock buyer improves the economics of firm low-carbon generation.

Supply entry has kept pace to date: offshore wind commissioning targeted at 4 to 5 GW per year under the government’s Clean Power 2030 plan, new-build firm capacity of 2 to 3 GW per year cleared through recent Capacity Market auctions four years ahead of delivery (mostly gas reciprocating engines, batteries with long-duration de-rating, and a small volume of gas power station refurbishment, on top of approximately 40 GW per year of re-contracted existing plant), and grid-scale battery storage expanding from 2 GW at end-2022 to approximately 8 GW by late 2025.

If procurement stalls, the wholesale effect is larger. A 4.4 GW data centre build-out into a frozen supply pipeline could add £15 to £25 per megawatt-hour to wholesale prices. At 3,400 kilowatt-hours per household, £25 per megawatt-hour is approximately £85 per year, which wipes out the £16 to £20 dilution saving and then some. The upper figure requires procurement to fail across the Capacity Market, the eighth round of the Contracts for Difference allocation (AR8), and the Long Duration Electricity Storage cap-and-floor scheme simultaneously.

For scale: the dilution benefit runs for the life of the assets, against which the forty-year-plus depreciation tails of transmission reinforcement and the twenty-five-year Contracts for Difference strike terms compound. The wholesale effect fades over two to four years as generators respond to the higher running hours, and in the long run flat baseload demand improves the economics of firm low-carbon generation, pulling Contracts for Difference strikes and Capacity Market clearing prices down. The policy condition is that every gigawatt of data centre demand needs to be matched by at least a gigawatt of new capacity on a timeline that brings supply online before or alongside the demand, with enough firm-equivalent share to cover the hours when wind and solar are not running.

7. Additional data centres do not add to transmission charges

The transmission envelope is set by Ofgem through the five-year RIIO price-control process. It reflects sunk network assets and approved reinforcement. It is not a direct function of the number of connected customers. New customers do not push the envelope up; they enlarge the denominator across which it is spread.

Where a new customer triggers reinforcement, the cost of that reinforcement is recovered through connection agreements under the shallowish-plus charging boundary that has applied since the Access Significant Code Review (Ofgem, May 2022). The customer pays the cost their connection causes, not the cost of the network as a whole. Industry modification CMP448 introduces a Progression Commitment Fee for generation and storage projects in the queue, a cash deposit that is forfeited if the project does not reach its milestones. Ofgem’s Demand Connections Reform Call for Input (February 2026) proposes equivalent deposits for demand. Neither mechanism shifts cost to other consumers; they shift cost onto the projects that cause it.

The dilution effect of £1.8 to £2.2 billion per year is a gift from the structure of British electricity charging. Capturing it requires the procurement side of the system to do its job. Treat data centre growth as a standalone demand question and the wholesale effect erodes most of the saving. Treat it as a paired demand-and-supply question and £16 to £20 per household per year lands.

8. Provenance of the 1.6 GW and 6 GW figures

The headline uses 1.6 GW as the “today” anchor and 6 GW as the 2030 ambition. Both numbers deserve scrutiny.

1.6 GW today. This is the Department for Science, Innovation and Technology’s (DSIT) Estimate of Data Centre Capacity: Great Britain 2024, published 1 May 2025. It is the government’s only published figure and it is explicitly scoped to co-location data centre capacity only, measured as the maximum rated computing-equipment load of co-location facilities. It excludes enterprise and hyperscale-owned facilities operated directly by Amazon, Microsoft, Google and Meta. The department does not publish those because there is no statutory requirement for operators to report capacity, and the four hyperscalers do not disclose UK-level numbers. Adding rough estimates of hyperscale self-built capacity (Amazon Web Services £8 billion committed, Microsoft £2.5 billion, Google Waltham Cross at £790 million) brings a plausible true “all-in” figure for today to around 2 to 2.5 GW. The piece uses 1.6 GW because it is the only number with a published methodology, and because the implications for dilution arithmetic are insensitive to whether the true baseline is 1.6 GW or 2.5 GW (the dilution effect runs on the growth increment, not the baseline).

6 GW by 2030. This is the government’s stated ambition in the UK Compute Roadmap (DSIT and UK Research and Innovation, 17 July 2025) and is scoped to “AI-capable” data centre capacity. It is a policy target, not an independent forecast. Reaching it requires the planning reforms in the December 2024 National Planning Policy Framework, the Nationally Significant Infrastructure Project opt-in announced 15 October 2025, the 2026 National Policy Statement for data centres, and the AI Growth Zones programme (first site at Culham, second at Blyth, over 200 applications received by April 2025) to deliver planning consents, grid connections and construction on a scale and timeline that the system as currently configured is not capable of achieving, absent a drastic re-application of effort and new solutions. The accompanying Compute Roadmap evidence annex (pp.13-14) gives a wider range of 3.3 to 6.3 GW for total UK data centre capacity by 2030 depending on policy interventions, and states that even 6.3 GW may not be enough to meet demand.

Which figure is right. The 3.3-6.3 GW range and the 6 GW AI-capable target are both preliminary analysis from the same document, presented to the House of Commons Library (research briefing CBP-10315, 3 November 2025) without independent methodology. The Library cites both figures as-is. Anyone treating the Library’s range as independent validation of the 6 GW ambition is reading the same analysis twice through different covers.

The system operator’s own 2030 projection, published independently of the department, has data centre electricity consumption rising from 5 TWh today to 22 TWh by 2030 (Clean Power 2030 Annex 1, 5 November 2024, p.4). Reverse-engineered at 85 per cent load factor and a computing-facility power ratio of 1.40, that implies approximately 3 GW of data centre capacity by 2030, sitting at the bottom of the Compute Roadmap range. NESO is the system operator and its demand forecast is the number its own network planning is built on. If the system operator is right, the dilution saving is approximately £900 million per year by 2030, about half of the 6 GW headline figure.

Scenario range. The mechanism is identical across outturns; only the denominator effect scales.

2030 capacityIncrement vs todayAdded TWhNational savingPer household3 GW (system operator forecast)1.4 GW14 TWh~£600m£5 to £74 GW (central case)2.4 GW25 TWh~£1.1bn£10 to £126 GW (Compute Roadmap ambition)4.4 GW45 TWh£1.8 to £2.2bn£16 to £20

9. Commercial and industry estimates

Three non-government sources on UK data centre capacity.

techUK, Foundations for the Future: how data centres can supercharge UK economic growth (4 November 2024): states there are approximately 450 data centres in the UK. Does not publish a total UK megawatt figure. This is the industry trade body’s flagship report and it deliberately anchors on site count rather than capacity, which makes it difficult to cross-check against the department’s 1.6 GW co-location number.

JLL, EMEA Data Centre Report Q3 2024 (13 June 2024): gives London-only capacity of 1,048 MW operational, plus 475 MW in development and 684 MW in the planning process, a total London pipeline of approximately 2.2 GW. London is estimated to hold 80 to 90 per cent of the UK co-location market, which implies a UK all-in figure on the same basis of approximately 1.2 to 1.3 GW operational today, broadly consistent with the 1.6 GW government number once hyperscale self-built is added.

CBRE, UK Data Centres Outlook (various editions): their periodic UK outlooks publish operational-supply figures of approximately 1.2 to 1.4 GW and development pipelines of 1.5 to 2.0 GW, again broadly consistent with the government number on co-location and not inclusive of hyperscale self-built.

Hyperscale self-built. Amazon, Microsoft, Google and Meta do not publish UK capacity. Public investment commitments through 2028 total more than £20 billion. No public source aggregates these into an operational-capacity number. Industry estimates put current hyperscale self-built UK capacity at 0.4 to 1.0 GW, with a pipeline that could add 1.5 to 3.0 GW by 2030.

No source publishes a consolidated “UK all-in” figure. Every published number has a scope caveat: the government number is co-location only, JLL and CBRE are commercial-market focused and similarly exclude hyperscale self-built, techUK uses site count, and the Compute Roadmap uses “AI-capable” with no formal specification. The piece uses the two numbers with the clearest provenance (1.6 GW from the government estimate today, 6 GW from the Compute Roadmap ambition by 2030) because those are the numbers with published methodology.

Sources

1. DSIT and UK Research and Innovation, UK Compute Roadmap, 17 July 2025 (6 GW AI-capable capacity ambition by 2030).

2. DSIT, UK Compute Roadmap: evidence annex, 17 July 2025, pp.13-14 (3.3 to 6.3 GW range for total UK data centre capacity by 2030).

3. DSIT, Estimate of Data Centre Capacity: Great Britain 2024, 1 May 2025 (1.6 GW co-location capacity).

4. NESO, Clean Power 2030 Annex 1: Electricity demand and supply analysis, 5 November 2024, p.4 (22 TWh data centre demand by 2030).

5. NESO, Five-Year View of TNUoS Tariffs for 2026/27 to 2030/31, Version 3.0, 18 September 2025, Tables 23 and 24 (pp.42-44) and Executive Summary (pp.7-8).

6. NESO, Final TNUoS Tariffs for 2026/27, Version 1.0, January 2026, p.6.

7. NESO, 2025 Annual Balancing Costs Report, June 2025.

8. House of Commons Library, Research Briefing CBP-10315, Data centres: planning policy, sustainability, and resilience, 3 November 2025.

9. techUK, Foundations for the Future: how data centres can supercharge UK economic growth, 4 November 2024.

10. JLL, EMEA Data Centre Report Q3 2024, 13 June 2024.

11. Ofgem, Final Determinations for RIIO-ET3, December 2025.

12. Ofgem, Final Determinations for RIIO-ED2, December 2022.

13. Ofgem, Typical Domestic Consumption Values, 2026 (3,400 kWh per year).

14. Ofgem, Targeted Charging Review Decision, 21 November 2019, Table 7.

15. Ofgem, Access Significant Code Review Decision, May 2022.

16. Ofgem, Demand Connections Reform Call for Input, February 2026.

17. DESNZ, Network Charging Compensation Scheme, 2024.

18. DESNZ, Review of Electricity Market Arrangements, July 2025 decision and subsequent consultation on long-term demand contracts.

19. CMP308 (balancing charges to Final Demand, April 2023).

20. CMP361 (balancing charges six-monthly fixed tariff).

21. CMP448 (generation and storage Progression Commitment Fee, live January 2026).

22. DUKES 2025 Table 5.1 (GB final electricity consumption).