2.6% of the sites, 12.3% of the energy: data centres in the UK’s Climate Change Agreements scheme

What 214 counts, and what it does not

214 is a count of facilities carrying the sector name 'Data Centres' in the Environment Agency's Reduced Rate Certificates register for certification period 7. It is a count of members of one voluntary tax-discount scheme. It is not a count of data centres in the UK, and it cannot be combined with any other data centre count.

• The two halves of the headline pair are measured at different dates. The facility share is over the 8,285 facilities certified for certification period 7, in a register file generated on 31 August 2026 and published under a landing page dated 18 September 2026. The energy share is over target period 6, which ran from 1 January 2024 to 31 December 2024, when the Environment Agency's biennial report records 3,332 target units covering 8,685 facilities in the scheme. Each share is internally consistent within its own file. They are not two cuts of one population, and this is the weakest joint in the metric.
• 100,323.47 GWh is the energy the scheme reported, not the scheme's whole energy. Nine of the 55 TP6 sector rows carry no energy figure. Five of those nine are blank throughout and have no CP7 register counterpart. The other four — Cement, Egg Processing, Kaolin and Ball Clay and Tyres, 20 certified facilities between them — report emissions but not energy. The Environment Agency does not say why those four in particular are blank; it states a general rule, that it “deliberately excluded energy consumption data for sectors with fewer than 3 distinct operators to avoid disclosure of commercially sensitive information”, and the attribution of the four blanks to that rule is ours, not the publisher's. Note that five further sector rows are blank for energy and are not attributed to it. Those four sectors are 1.30 per cent of reported TP6 emissions, so the effect on the data centre share is small, and it runs downwards: the true all-scheme share would be slightly below 12.29 per cent.
• Data centres rank 9th of 45 on energy per certified facility. The 4.76 ratio is a real and large gap, but the sector is not the most disproportionate in the scheme and the study should not be read as saying so.
• Target period 6 is a twelve-month period covering calendar 2024, so 100,323.47 GWh and 12,327.58 GWh are annual figures, and 57.6 GWh per certified data centre is an annual figure.

What was already known

techUK, which holds the umbrella agreement for the data centre sector, states on its Climate Change Agreements page that “Over 170 data centre sites currently participate in the scheme.” The page is dated 14 November 2023. It does not say what a “site” is, or the date the count is as at, and the identical sentence appears in Internet Archive captures from 7 December 2023 to 16 May 2026, so it has not been revised since it was published. techUK also published a company count in September 2017, for target period 2: “129 facilities from 57 target units, representing 42 companies.” That figure describes a period that ended on 31 December 2016 and is not set against anything here. The Environment Agency's own narrative report for target period 6 does not mention data centres at all. We found no published analysis that sets the data centre sector's share of the scheme's facilities against its share of the scheme's energy.

What this study adds

• A reproducible facility count from the statutory register, with a stated definition and a stated date, rather than a site count that defines neither: 214 rows carrying the sector name 'Data Centres' in the CP7 register, in a file generated on 31 August 2026 and published under a landing page dated 18 September 2026.
• The energy side, which exists only in a separate workbook, joined to the facility side through a published sector crosswalk.
• The pair itself: facility share against energy share for the same named scheme.

Our 214 is not a correction of techUK's “over 170”. The two use different definitions at different dates, so neither can be checked against the other. Prior work cited in full: techUK, Climate Change Agreements (CCA) for data centres, a resource page (14 November 2023, as dated on the page); techUK, Climate Change Agreement (CCA) for Data Centres — Target Period Two: Report on Findings (September 2017); CETEx, Centre for Economic Transition Expertise, London School of Economics and Political Science, Submission to the UK government consultation “Risks and opportunities to the sustainability of data centres in the UK” (July 2026); Environment Agency, Climate Change Agreements biennial progress report for 2023 and 2024 (18 December 2025). Two prior-art sources could not be checked. The House of Commons Library briefing on data centres and the Environmental Audit Committee's written evidence both sit behind a challenge that returned an error to our identifying user agent; we stopped rather than working around it. The Commons Library is the most likely place a competing UK figure would appear, so the prior-art sweep has a known hole there. The finding is worded as what we found, not as what exists.

The pair: 2.58 per cent of the facilities, 12.29 per cent of the reported energy

The register gives facility counts and no energy. The sector performance workbook gives energy and no facility counts. Neither file joins to the other, and the publisher does not join them. Joined through the crosswalk published with this study, the 214 certified data centre facilities are 2.58 per cent of the 8,285 on the register, while the 12,327.58 GWh the sector reported for target period 6 is 12.29 per cent of the 100,323.47 GWh reported across the 46 sectors that carry an energy figure. The energy share is 4.76 times the facility share.

By reported energy the sector ranks 2nd of 46, behind Food and Drink at 20,426.67 GWh and ahead of Chemicals at 10,141.42 GWh. By facility count it ranks 8th of 49. It is worth saying plainly what this is not: on energy per certified facility the sector ranks 9th of the 45 register sectors that carry energy, behind Glass, Steel, Compressed Gases, Wood Panels, Paper, Motor Manufacturing, Metal Packaging and Chemicals. Glass and Steel are a handful of very large plants. What is unusual about data centres is the combination: a large number of facilities and a high energy figure per facility.

By reported energy, data centres rank 2nd of 46

Scroll the table sideways to see every column →

The twelve largest of the 46 sectors that carry a target-period-6 energy figure, from the TP6 sheet of the Environment Agency’s Sector Performance Data workbook. Target period 6 ran for twelve months, 1 January 2024 to 31 December 2024.

By energy per certified facility, they rank 9th of 45

The same table read the other way is the guard-rail on the finding above. Eight sectors report more energy per certified facility than data centres do, and the two at the top are a handful of very large plants: Glass is a single certified facility and Steel is 5. Nothing here says data centres are the scheme’s most energy-intensive sector, because on this measure they are not.

Scroll the table sideways to see every column →

* fewer than ten certified facilities: Glass (1), Steel (5), Wood Panels (5). Percentages on fewer than ten facilities are marked in the published dataset. Thirteen of the 49 register sectors have fewer than ten certified facilities, one of them a single facility. No claim here rests on any of them. The scheme mean is 12.11 GWh per certified facility, being the 100,323.47 GWh the scheme reported over all 8,285 certified facilities, including the 20 whose sector energy is withheld. If you are sizing real equipment rather than reading a register, the per-machine end of this is in our server power density study and the server power consumption calculator.

What the scheme is, and what 214 counts

Climate Change Agreements are voluntary. An operator joins through a sector association, which holds an umbrella agreement; the operator's own underlying agreement sets an energy-efficiency target on its target unit, not on its sector. Certified facilities qualify for a discount on the Climate Change Levy: 92 per cent on electricity and 89 per cent on gas in the rates published from 1 April 2024. Meeting the target is not a condition of keeping the discount. An operator that misses its target may instead pay a buy-out fee — £25 for each tonne of CO2e above its target emissions in target period 6 — and stay certified, and the Environment Agency records that 1,766 target units, 54.5 per cent of them, missed their target period 6 target. “Certified” on the register used here therefore means certified for the discount, not passed an efficiency test. The Environment Agency administers the scheme for the whole of the UK on behalf of the Department for Energy Security and Net Zero, and must publish a list of certified facilities under regulations 9(1) and 9(2) of the Climate Change Agreements (Administration) Regulations 2012, revised on the last working day of each month. That list is the register used here.

The eligible activity for the data centre sector, code DATC, is “the leasing or licensing of a data facility which is being used as a data centre”, where a data facility needs regulated temperature and humidity, a back-up supply and an electricity supply of at least 200 kW. On its face that is colocation rather than a facility an operator runs for its own workloads. The register nonetheless carries operators that read as enterprises running facilities for their own workloads rather than as colocation providers, so the activity may be read more broadly in practice; we note the tension and do not resolve it. The scheme was also closed to new entrant facilities from 1 November 2018 until a new-entrant window that closed on 30 September 2025, so building completed across that period could not join the scheme while it was shut. It is not shut now: under the February 2026 statutory guidance, “new entrants can apply to be considered for inclusion in an existing umbrella agreement between 1 January and 31 August from 2026 to 2029 inclusive”.

Scroll the table sideways to see every column →

Quoted definitions in this table are from Climate Change Agreements Operations Manual, February 2022, LIT7911, Appendix A: Eligible processes. Environment Agency. OGL v3.0. The relief percentages are HM Revenue and Customs guidance on Climate Change Levy rates; the buy-out fee and the target-unit pass rate are the Environment Agency’s biennial progress report for 2023 and 2024. 214 is a count of facilities carrying the sector name 'Data Centres' in the Environment Agency's Reduced Rate Certificates register for certification period 7. It is a count of members of one voluntary tax-discount scheme. It is not a count of data centres in the UK, and it cannot be combined with any other data centre count.

The bakery comparison, with the definitions attached

The scheme's two largest sectors by facility count are Supermarkets — which inside this scheme means food processing in supermarket stores — with 2,032 certified facilities, and Bakers — craft baking, expressly not industrial baking — with 1,796. Together they reported 1,940.24 GWh for target period 6, against 12,327.58 GWh from 214 data centres: 6.35 times as much from about an eighteenth as many facilities — 3,828 certified facilities against 214.

Neither sector name means what it appears to mean, and the comparison is misleading without the definitions. The Environment Agency's operations manual defines the Bakers sector, code NAMB, as craft baking: a facility producing “bread, morning goods, flour confectionery and savoury products … using traditional bakery skills, rather than ‘industrial’ type baking processes”. The Supermarkets sector, code FDFS, is “a facility which treats and processes materials intended for the production of food products in supermarket stores” — the food preparation inside the store, not the store's electricity. 507.37 GWh is not what the UK's supermarkets use.

Scroll the table sideways to see every column →

Definitions quoted from Climate Change Agreements Operations Manual, February 2022, LIT7911, Appendix A: Eligible processes. Environment Agency. OGL v3.0. The scheme's Bakers sector is craft baking, defined against ‘industrial’ baking, and its Supermarkets sector is food processing inside supermarket stores, not a supermarket's electricity. The bakery comparison is meaningless without those definitions and they are carried with it wherever it appears.

How many operators, and why there is no single answer

The register publishes an operator name against every facility. Counting the distinct names behind the 214 data centre facilities gives a different answer for every defensible rule, so all five are published side by side and none is used in the headline.

The instability has two visible causes, pulling in opposite directions. Some operators file each data hall as a separate company, so counting registered entities overstates the number of distinct businesses. And the register carries the same registered name under more than one spelling, so the raw count overstates it again: several names do this, one of them under three spellings, and the two largest cover 15 facilities between them and 11 between them. Rule 5 goes further than spelling and merges on the first word of a name, which fuses that 15 with a separately registered company into a single holding of 45 — and, by the same rule, fuses unrelated companies that happen to share a first word. That is why rule 5 is published as an upper bound rather than an estimate. Top-ten concentration moves from 44.4 per cent under the raw strings to 51.9 per cent under the strictest rule we treat as defensible, a swing of 7.5 points on a choice of rule, and to 60.3 per cent under rule 5's deliberate over-merge.

Scroll the table sideways to see every column →

All five rules run over the same 214 certified data centre facilities on the certification period 7 register, and every one of them is a column in the published dataset, so each can be recounted from the CSV. The range is the finding: no single operator count is published and none appears in the headline or the title. Deliberately strict, and ours rather than the publisher's. It errs towards withholding. The full operator column remains available from the Environment Agency's own file under OGL v3.0.

What this study does not do

No comparison is made between target periods. Target period 6 ran for twelve months; target periods 1 to 5 each ran for twenty-four, and there was no target period at all in 2023. A TP5 to TP6 comparison would show a near-halving that is period length, not a fall in consumption. The Environment Agency itself warns that “simple comparisons in their energy and emissions must be approached with caution” because of differences in reporting scope, throughput and methodology, and scheme membership changes between periods.

No attempt is made to locate the 214 facilities. The Environment Agency withholds facility addresses for six sectors, including data centres, for reasons of national security and site security, and every data centre row carries a placeholder in place of an address. Nothing here is converted into a number of homes or a sum of money, and no figure is used to estimate anything outside the scheme.

Deliberately not published

• Any operator or company count in the headline. Top-ten concentration runs from 44.4 per cent under the raw strings to 51.9 per cent under the strictest rule we treat as defensible — a 7.5-point swing on a choice of rule — and to 60.3 per cent under rule 5, which is published as an upper bound rather than an estimate.
• The facility churn or turnover finding, entirely. There is no archival snapshot of the register anywhere against which to bound it.
• Any comparison between target periods. TP6 is twelve months, TP1 to TP5 are twenty-four, and there was no target period in 2023.
• Any use of an energy-per-facility figure to extrapolate beyond the scheme.
• Any attempt to locate the 214 facilities, infer their locations from operator names, or look up addresses elsewhere.
• Any conversion of the relief into a sum of money or a number of homes.
• A map. There are no addresses to put on one, and constructing them would defeat the publisher's security withholding.

Where the official sources disagree with one another

Scroll the table sideways to see every column →

What could make these figures wrong, or read wrongly

1. 214 is a count of facilities certified under one voluntary scheme, not a count of data centres in the UK. It must not be multiplied by, netted against, or compared with any external data centre count.
2. The two halves of the headline pair are measured at different dates. The facility share is over the 8,285 facilities certified for certification period 7, in a register file generated on 31 August 2026 and published under a landing page dated 18 September 2026. The energy share is over target period 6, which ran from 1 January 2024 to 31 December 2024, when the Environment Agency's biennial report records 3,332 target units covering 8,685 facilities in the scheme. Each share is internally consistent within its own file. They are not two cuts of one population, and this is the weakest joint in the metric.
3. 100,323.47 GWh is the energy the scheme reported, not the scheme's whole energy. Nine of the 55 TP6 sector rows carry no energy figure. Five of those nine are blank throughout and have no CP7 register counterpart. The other four — Cement, Egg Processing, Kaolin and Ball Clay and Tyres, 20 certified facilities between them — report emissions but not energy. The Environment Agency does not say why those four in particular are blank; it states a general rule, that it “deliberately excluded energy consumption data for sectors with fewer than 3 distinct operators to avoid disclosure of commercially sensitive information”, and the attribution of the four blanks to that rule is ours, not the publisher's. Note that five further sector rows are blank for energy and are not attributed to it. Those four sectors are 1.30 per cent of reported TP6 emissions, so the effect on the data centre share is small, and it runs downwards: the true all-scheme share would be slightly below 12.29 per cent.
4. Those 20 facilities count in the facility denominator but contribute nothing to the energy denominator, so the two denominators are not perfectly coextensive. They are 0.24 per cent of the register.
5. The energy figures cover only energy subject to a Climate Change Agreement. In the Environment Agency's words, the report “only covers the energy use that is subject to a CCA, which may not be all the energy used in facilities (for example, fuels used by those regulated by the UK Emission Trading Scheme, UK ETS, are excluded)”. These are not site totals.
6. Target period 6 is a twelve-month period covering calendar 2024, so 100,323.47 GWh and 12,327.58 GWh are annual figures, and 57.6 GWh per certified data centre is an annual figure.
7. Data centres rank 9th of 45 on energy per certified facility. The 4.76 ratio is a real and large gap, but the sector is not the most disproportionate in the scheme and the study should not be read as saying so.
8. The two workbooks name their sectors differently, so the join runs through a hand-built crosswalk published with this study: 42 exact matches, 8 renamed or split, 5 performance sectors with no CP7 counterpart. Data Centres, Bakers and Supermarkets each match one-to-one on an exact string, so no headline figure depends on a crosswalk judgement.
9. The scheme's Bakers sector is craft baking, defined against ‘industrial’ baking, and its Supermarkets sector is food processing inside supermarket stores, not a supermarket's electricity. The bakery comparison is meaningless without those definitions and they are carried with it wherever it appears.
10. Distinct-operator counts for the 214 facilities range from 60 to 87 across five rules and top-ten concentration from 44.4 to 60.3 per cent. The range is the finding. No single operator count is published, and no operator figure appears in the headline.
11. The register publishes an operator name for every facility. In the farming sectors that name can be an individual's. The published dataset carries an operator name only where the string contains a registered-company or public-body token; 236 of 8,285 rows are withheld by that screen, and all 214 data centre rows pass it. The screen is ours, is deliberately strict, and is reproducible from the published dataset: the rows it withholds are the rows where operator_as_published is empty.
12. Facility addresses are withheld by the publisher for six sectors — Data Centres, Eggs and Poultry Meat, Meat, Pigs, Poultry Meat Processing and Poultry Meat Rearing — for reasons of national security and site security. That is 1,148 facilities, including all 214 data centres. No address of any kind is carried into the published dataset, for any sector.
13. Both source files are overwritten at the same URL every month with no archive and no version history. From the next Environment Agency refresh, the live URLs will not reproduce these figures. The byte counts, md5 and sha256 of the exact files used are published in the source list; the exact files themselves are republished with this study under the Open Government Licence v3.0, so the whole analysis stays checkable after the sources have moved on; and the analysis script refuses to run against different bytes.
14. The register's own internal timestamp decodes to 31 August 2026, while the data.gov.uk landing page says the dataset was updated on 18 September 2026. The latest certification start date anywhere in the file is 26 August 2026, which is consistent with a file generated on 31 August. Both dates are real and mean different things; the file used here was generated on 31 August 2026 and published on the landing page dated 18 September 2026.
15. Four official sources give four different sector counts for the same scheme: 49 sector names on the register's Data sheet, 48 sector associations on its Details sheet, 55 sector rows on the TP6 sheet and 53 claimed on the sector performance landing page. Each figure here names which of those it uses.
16. The sector performance landing page is stale against its own data: it still describes a scheme running to 31 March 2025 and “four biennial target periods between 2013 and 2020”. It is not used for sector counts or scheme dates.
17. Percentages on fewer than ten facilities are marked in the published dataset. Thirteen of the 49 register sectors have fewer than ten certified facilities, one of them a single facility. No claim here rests on any of them.
18. The workbook states that energy and emissions were “rounded down to 2 decimal places for presentation purposes”, but the stored cells carry full precision and are used as stored. Every figure is rounded once, at presentation.
19. The data centre sector is recorded as meeting its target period 6 target while reporting more absolute energy than in its base year, because the target is an efficiency measure adjusted for membership change while the energy columns are not. The two figures are not comparable, neither is set against the other anywhere in this study, and no growth in data centre energy use is measured or claimed here.
20. Two prior-art sources could not be checked. The House of Commons Library briefing on data centres and the Environmental Audit Committee's written evidence both sit behind a challenge that returned an error to our identifying user agent; we stopped rather than working around it. The Commons Library is the most likely place a competing UK figure would appear, so the prior-art sweep has a known hole there. The finding is worded as what we found, not as what exists.
21. Certification is not a pass mark. An operator that misses its target may pay a buy-out fee — £25 for each tonne of CO2e above target emissions in target period 6 — and stay certified: the Environment Agency records 1,766 target units, 54.5 per cent, failing their TP6 target. The 8,285 facilities counted here are facilities certified for the discount, not facilities that met an efficiency target.
22. The scheme covers the whole of the UK, not Great Britain: the Environment Agency administers it UK-wide on behalf of the Department for Energy Security and Net Zero, and 177 of the 8,285 certified facilities carry Northern Ireland postcodes. None of the 177 is a data centre facility, for a mechanical reason rather than a geographic one: data centre rows carry a placeholder in place of an address, so they cannot appear in a postcode-derived count at all. The register does not say where the 214 are.

How the two files were joined

Two Environment Agency files were mirrored with their hashes and joined. Facility counts come from the Reduced Rate Certificates register for certification period 7, published under regulations 9(1) and 9(2) of the Climate Change Agreements (Administration) Regulations 2012. Reported energy comes from the Sector Performance Data workbook for target period 6. The publisher does not join these two files. Both are censuses of what is published, not samples.

Sources, with the exact bytes used

Scroll the table sideways to see every column →

Both source files are overwritten at the same URL every month with no archive and no version history. From the next Environment Agency refresh, the live URLs will not reproduce these figures. The byte counts, md5 and sha256 of the exact files used are published in the source list; the exact files themselves are republished with this study under the Open Government Licence v3.0, so the whole analysis stays checkable after the sources have moved on; and the analysis script refuses to run against different bytes.

How to check this study

Everything is recomputable from the two mirrored workbooks alone. Re-download both from the source URLs and check the byte counts and hashes in the source list first: the Environment Agency overwrites both files monthly at the same URL with no archive, so a later copy will not reproduce these figures and a hash mismatch is expected after the next refresh. Every figure below can then be recomputed from the published CSV alone, with the recipe given against it; the two rankings and the crosswalk can be rebuilt from the mirrored workbooks the same way.

Scroll the table sideways to see every column →

Contains public sector information licensed under the Open Government Licence v3.0. The attribution statements the publisher specifies are “© Environment Agency copyright and/or database right 2016. All rights reserved.” for the Reduced Rate Certificates register and “© Environment Agency copyright and/or database right 2017. All rights reserved.” for the Sector Performance Data workbook; both are reproduced as published. The published dataset contains no personal data. Facility addresses are not carried for any sector. Operator names are published only where the string carries a registered-company or public-body token; 236 rows are withheld by that screen. All 214 data centre operator strings are bodies corporate. Downloads: full dataset, CSV (8,285 rows, 24 columns) · README, column definitions and attribution · every figure on this page as JSON. The register publishes an address for most sectors and withholds it for six, data centres among them. No address is carried into our published dataset for any sector. Because both Environment Agency files are overwritten monthly at the same URL with no archive, the exact source bytes this study was computed from are republished here under the Open Government Licence v3.0, so the whole analysis stays checkable once the live URLs have moved on: Reduced Rate Certificates register, CP7 (549,952 bytes, md5 3d6832c081de879062e2fb482d618106) · Sector Performance Data, TP1 to TP6 (65,086 bytes, md5 3a66e1102eb8ac67faad2df3032dfa6a).

Questions about these figures

Where to take this next

This study measures a register, not a building. If the question behind your visit is how much electricity your own equipment draws, what a rack of it needs, or what an estate costs to run and to cool, the tools below work from published server specifications rather than from scheme returns. The large-site electricity study is the closest companion to this one: it measures metered electricity by local authority from Department for Energy Security and Net Zero data, on a different base and for different years, and reaches data centres from the other direction.

More Servnet research · UK business e-waste statistics 2026, also built from Environment Agency data · Server consolidation calculator · Power and cooling tools