Summary
- ONS analysis identified 239 externally serving data centres in 2024, up from 221 in 2020.
- Estimated electricity consumption reached 4.5TWh in 2024, around 2% of Great Britain grid electricity.
- An expanded digital-infrastructure methodology puts 2025 investment at £11.2bn rather than £3.6bn under the narrow telecoms measure.
The Office for National Statistics is reworking how the UK’s digital infrastructure is measured, highlighting a statistical problem familiar to the data-centre industry: the physical assets behind cloud computing and AI are spread across buildings, electrical equipment, servers, software, networks, and multiple industrial classifications.
A new ONS methodology article says data-centre investment and output cannot currently be isolated cleanly within the National Accounts because facilities cross both industry and asset boundaries. A purpose-built data-centre building may sit in one investment category, its servers and network hardware in another, and software and databases elsewhere.
Using data matched with the Department for Energy Security and Net Zero, the ONS identified 239 data centres serving external organisations in 2024, up from 221 in 2020. It estimates those facilities consumed around 4.5TWh of electricity in 2024, approximately 2% of the 249.2TWh supplied through the Great Britain electricity grid.
That electricity estimate was 1.3TWh, or 41%, higher than in 2020. The figures do not encompass every type of server room or privately operated enterprise facility, but they put a more concrete national-account frame around an infrastructure category whose visibility has increased rapidly with cloud and AI investment.
A companion ONS research paper goes further by testing a broader definition of digital infrastructure investment. Under the existing narrow approach, focused largely on telecommunications infrastructure, UK investment in 2025 is estimated at about £3.6bn. The experimental wider definition produces a figure of £11.2bn.
The difference comes from recognising a larger collection of physical and intangible assets, including data-centre buildings, information and communications technology hardware, network equipment, software, databases, and spectrum. The ONS stresses that the broader measure is statistics in development rather than a replacement for the settled National Accounts framework.
That caveat is important, but so is the direction of travel. Data centres increasingly behave like large industrial infrastructure projects: they require land, substations, power distribution, cooling plant, backup systems, structural shells, fibre, security, and substantial inventories of computing hardware. Treating the sector primarily through traditional telecommunications statistics risks missing much of the capital being deployed.
The same classification problem affects policy. Ministers considering grid connections, planning, critical national infrastructure, or industrial strategy need to understand whether investment is going into network assets, buildings, imported servers, domestic construction, software, or electricity infrastructure. Each has different economic and supply-chain consequences.
The ONS work adds a different layer to attempts to put numbers around Britain’s rapidly expanding data-centre estate by asking not simply how many megawatts or buildings exist, but where their economic activity appears inside official measurement.
The research also points to gaps still to be closed. Cooling systems, electrical distribution, server racks, and other facility-specific assets may require additional work before the statistical treatment reflects the way modern data centres are actually designed and financed.
That will become more pressing as facilities grow larger. A multi-hundred-megawatt AI campus can combine property development, power infrastructure, specialist mechanical and electrical plant, imported accelerators, software, and network investment in a single programme worth billions of pounds. Conventional sector boundaries struggle to describe that package.
The ONS has not replaced its official investment statistics with the £11.2bn expanded measure. Its work does, however, establish data centres as a distinct measurement problem within the digital economy — one rooted as much in buildings and electricity as in software and telecommunications.
