Summary
- Computer Weekly estimates approximately 1.5 million tonnes of annual operational CO₂e from a surveyed UK data centre estate rated at around 1.5GW.
- Its 14.6GW planning pipeline could add 12 million tonnes annually under an average-grid scenario or 38 million tonnes under a gas-based scenario.
- Different capacity estimates, regional grid modelling and embodied emissions assumptions prevent these figures from being treated as measured industry totals.
A new investigation has estimated that Britain’s existing data centres account for approximately 1.5 million tonnes of annual operational greenhouse gas emissions, while identifying a much larger potential carbon burden associated with proposed developments.
The assessment, published on 13 October by Computer Weekly, combines a survey of operating facilities with construction pipeline information, electricity carbon intensity data and assumptions about computing equipment utilisation.
It identifies an existing estate of approximately 1.5GW across 90 to 110 sites and a development pipeline comprising 173 projects that have entered planning but are not yet built. Published capacity figures are available for enough of those proposals to identify more than 14.6GW of prospective development.
The researchers estimate that the existing facilities produce approximately 1.5 million tonnes of carbon dioxide equivalent annually when assessed against Britain’s average electricity carbon intensity in 2024.
Applying the same approach to the entire identified development pipeline produces a potential additional 12 million tonnes annually. An alternative calculation assuming the additional electricity is supplied by gas-fired generation produces approximately 38 million tonnes.
Both pipeline estimates are hypothetical outcomes, not measured emissions or forecasts of capacity already approved for operation. The calculations depend on proposed projects reaching completion and operating under the assumed electricity and utilisation conditions.
Capacity estimates depend on the methodology
The investigation’s estimate of approximately 1.5GW for the existing estate differs from other published assessments of UK data centre capacity.
DataCentral previously reported CoStar Analytics’ estimate of 3.5GW of installed capacity, alongside approximately 1GW under construction. The figures arise from different inventories and counting methods and should not be presented as directly comparable.
Computer Weekly based its count on non-domestic Energy Performance Certificate records for England, Scotland and Wales, checking individual sites and estimating capacity from published specifications or building area where necessary.
The report then models annual consumption rather than assuming that every facility draws its rated capacity continuously. Its central assumption is that sites use an average of 60% of rated capacity, with a sensitivity range of 45–70%.
Facility overheads are incorporated through power usage effectiveness, or PUE, which measures total facility electricity consumption relative to the energy supplied to IT equipment. The investigation uses PUE values of 1.3–1.7 for existing facilities and 1.15–1.4 for newer developments.
These assumptions materially influence the results. Capacity expressed in megawatts is not equivalent to electricity consumption in megawatt-hours, and PUE must be applied consistently with the measurement boundary used for the original capacity figure.
Combining the assumptions with an annual electricity carbon intensity of 125g per kilowatt-hour produces the study’s central estimate of 1.5 million tonnes. Changing the utilisation and efficiency assumptions produces a range of approximately 0.8 million to 2.5 million tonnes.
The National Energy System Operator independently reports a 2024 national average electricity carbon intensity of 125g CO₂/kWh. It is an historical national average, not a fixed emissions factor for every location or future year.
Government metering statistics provide another measurement basis. The Department for Energy Security and Net Zero estimates that qualifying data centres consumed 4.5TWh of electricity in Great Britain during 2024, approximately 2% of grid consumption.
That dataset does not cover every category of data centre. Differences in facility coverage, capacity definitions and estimated utilisation prevent it from being reconciled directly with the investigation’s model without more detailed site-level information.
Regional electricity and future generation
Geography changes the carbon estimates because electricity generation and network conditions differ between regions.
Computer Weekly’s modelling uses regional carbon intensity estimates ranging from approximately 22g per kilowatt-hour in North East England to 254g in South Wales. These regional figures are modelled estimates rather than independently metered regional generation emissions.
When the existing estate is weighted by its geographical concentration around London and the M4 corridor, the study’s operational emissions estimate rises from approximately 1.5 million to 1.9 million tonnes annually.
The proposed pipeline has a different distribution, including substantial developments in Scotland and North East England. The researchers estimate a lower average regional electricity carbon intensity for that pipeline, although a significant proportion remains concentrated in areas with higher modelled emissions.
Regional averages do not establish the emissions associated with supplying one additional large electricity consumer. The electricity generated in response to increased demand may differ from the generation mix already serving the system.
To examine that distinction, the investigation uses an alternative gas-based intensity of approximately 400g per kilowatt-hour. The resulting 38-million-tonne annual figure provides a contrasting scenario rather than a prediction that every new data centre will operate exclusively on gas.
Future emissions will also depend on how rapidly the electricity system decarbonises, whether projects secure reliable grid connections, and which proposed facilities ultimately receive investment and planning approval.
Embodied emissions add another calculation
Electricity use is not the only source of emissions associated with data centre development. Buildings, electrical installations, cooling equipment and servers require materials and manufacturing before the facility enters operation.
The investigation estimates embodied emissions of approximately 3.5 million tonnes for the existing estate and 31 million tonnes for the proposed pipeline. These figures represent estimated emissions associated with construction and equipment, not recurring annual electricity emissions.
The researchers annualise those impacts using assumed asset lives of 40 years for buildings, 15 years for mechanical and electrical equipment, and four years for servers.
Under that approach, annualised embodied emissions add approximately 40–50% to the modelled operational footprint. The result depends on assumed equipment manufacturing impacts and replacement cycles, particularly for servers that may be refreshed several times during a building’s life.
Such lifecycle calculations should not be added unqualified to the UK’s territorial emissions inventory. Manufacturing emissions may arise outside Britain, whereas national territorial statistics record emissions produced within the country’s borders.
The investigation therefore provides scenarios for the potential scale of data centre emissions rather than an audited sector inventory.
More consistent disclosures of actual electricity consumption, commissioned IT capacity, equipment replacement and electricity supply arrangements would allow future assessments to move beyond assumptions about rated capacity and project pipelines.
Until then, the gap between published capacity estimates and the uncertainty surrounding proposed developments remain central limitations on calculating the sector’s current and future carbon footprint.
