Mapping the UK's data centres build-out: implications for digital sovereignty

Data centres are the ‘backbone’ and ‘back-office’ of the digital economy: they host the compute, storage, and networking capabilities that make cloud, artificial intelligence (AI), and digital services delivery and trade possible. Their location, ownership, and energy footprints have become issues relevant not only for industrial and local development policies, but also in discussions on Foreign Direct Investments (FDI) strategy and digital sovereignty.

This blog unpacks the location, typology and energy use of data centres in the UK, and draws implications for the impact of future investments in digital infrastructure in the UK, rather than relying on foreign investments in data centres in the UK. Building on previous work, and on the Baxtell database, we show that data centres are predominantly located in England, are top-heavy by installed capacity and are almost entirely owned or controlled by a small number of global players.

The UK currently (up to the end of 2025) hosts 348 live data centre sites with about 2.0 Gigawatt (GW) of operational electrical capacity. Another 0.87 GW is under construction and a further 10.88 GW is in the announced or permitted pipeline. If the whole pipeline was commissioned, live capacity will grow by a factor of roughly 5.8 in a few years. The median site in terms of GW capacity was built in 2024; this is a young and rapidly expanding stock.

The market looks fragmented at first glance — 117 distinct operators — but the picture changes once we weight by megawatt (MW). The top ten groups account for around three-fifths of operational MW (Baxtell, 2025). Five carrier-neutral1 platforms (Vantage, VIRTUS, Ark, Equinix, Digital Realty) together with hyperscale owner-operators such as Amazon AWS and NTT dominate installed capacity, even as the long tail of smaller colocation, real-estate and managed-service providers populates the count of sites. Sixty-five per cent of sites are carrier-neutral colocation; only sixteen are hyperscale. This is a capital-intensive industry in which scale, grid access and engineering capability create durable advantage — a structure that policy should treat as a concentrated infrastructure market, not a competitive service sector.

The geography is even more revealing. Up to the end of 2025, England hosts 313 of 348 sites — about 90 per cent of the UK stock — with only 14 in Scotland, 12 in Wales and 9 in Northern Ireland. London alone accounts for 218 sites: nearly two-thirds of the national total. This London-centric agglomeration is consistent with the broader pattern documented in CITP Briefing Paper 16 (The Uneven Geography of Digital Infrastructure): digital infrastructure gravitates towards interconnection nodes, low-latency finance, and major service hubs. The UK's own Crown Dependencies sit at the top of this list. Regulatory arbitrage is part of the story of where data is housed, though further analysis is needed to draw conclusions with regard to the UK’s exposure to this.

This matters for reasons that should worry the trade and industrial policy communities equally. First, the energy and emissions implications are non-trivial, and they are local. EPRI (2024)2 projects global DC electricity demand to more than double by 2026; in Ireland, the International Energy Agency (IEA) expects data centres to absorb close to a third of national electricity consumption in the same year. The UK's own pipeline points in the same direction, with the Southeast and London already grid-constrained. Recent work by Bonfiglioli et al. (2025)3 using US commuting zone data finds that localities more exposed to AI experience faster CO² emissions increases, that scale effects dominate any compositional gains, and — crucially — that power plants near data centres shift towards greater fossil-fuel use. Annual renewable Power Purchase Agreements do not address this; only hourly-matched, additional clean energy does. The headline of “net-zero cloud” hides a reality that is heavily and yet increasingly based on local fossil fuel energy.

Second, the employment case for data centres is far weaker than it appears. US evidence is sobering: about $10 billion in data centre deals received at least $811 million in public tax abatements, in exchange for an estimated 837 permanent jobs — more than one million dollars of public subsidy per job (Alvarez et al., 2026)4. Forthcoming work on France (Fontanelli et al. 2026)5, which looks at 280 French commuting zones between 2011 and 2022, finds positive effects on construction and advanced business services, but concentrated in metropolitan areas and skewed toward high-skill intellectual occupations. The capital-to-labour ratio of a hyperscale data centre is exceptionally high. Once built, a billion-euro facility may employ only a few dozen technicians. Selling data centres as instruments of levelling-up — or as substitutes for genuine regional industrial policy — is not supported by the (so far scant) evidence.

Third, the fiscal stakes are high. The intensity of data centre location seems to be particularly high in tax havens, likely led by profit shifting (Papadakis and Savona, 2024)6. A leading digital-services exporter like the UK loses tax base when FDI attraction and location siting decisions are driven by arbitrage rather than by economic and environmental merit. Industrial policy that competes on relaxed environmental standards or on tax giveaways risks subsidising the very dynamics it claims to counter.

Fourth, the debate around EU digital sovereignty risks leaving the UK misaligned with its largest trade partner. The EU has just launched its digital sovereignty package, with the stated aim of spurring EU investments in digital infrastructures and reducing dependence on non-EU providers. The package entails several distinct instruments. For instance, the proposed Cloud and AI Development Act sets the EU on a path to roughly triple data-centre capacity over the next five to seven years, conditional on sustainability and sovereignty requirements. The AI Continent Action Plan and the AI Gigafactories initiative channel public co-financing into very-large-scale compute. In sum, the EU is increasingly treating cloud and data-centre capacity as a strategic industry, with EU-resident control, sustainability conditionality, and interoperability as core objectives.

For the UK, the trade-offs are not symmetric across instruments. Alignment on sustainability disclosure, switching and interoperability rules would reduce friction for digitally-delivered services exports, help preserve the data adequacy decision, and keep open case-by-case participation in EuroHPC and AI Gigafactory consortia (Lehdonvirta et al.,2025)7. The strictest EU sovereignty variants and a “buy European” tilt would, however, constrain the US hyperscalers — Amazon AWS, Microsoft Azure, Google Cloud — that operate a large share of UK live and pipeline capacity, and could slow down the UK AI Strategy. Non-alignment preserves regulatory autonomy but raises double-compliance costs, and the prospect of the UK as a regulatory island between EU sovereign-cloud rules and US cloud-export controls. Selective alignment — close on sustainability, switching and interoperability, reserved on sovereignty and procurement — is a plausible middle path, but requires a coherent strategy to avoid yet another instance of alienating the EU in the midst of a renewed effort towards a UK-EU reset.

The picture that emerges is of a country about to undergo a step-change in compute capacity. It would be advisable to empirically inform the framework guiding data centres planning, energy, FDI and industrial-policy objectives to be fully equipped to govern it. Overall, what follows from this, both for UK and EU policy, is not a brake on digital infrastructure but a proper infrastructure policy for it.

Four steps are within reach.

(i) Mandatory environmental impact assessments for new sites above a clear capacity threshold – say 10 MW - covering energy, water, embodied carbon and cumulative grid load.

(ii) Environmental disclosure statements as a precondition for permits: hourly energy use, water withdrawal and consumption, and the share of capacity contracted to third parties, published on a standardised basis.

(iii) Conditional public support: subsidies, planning consents, and grid-connection priority tied to 24/7 hourly-matched renewable [energy] procurement with additionality verification, mandatory waste-heat recovery for connection to district heating, which implies a process of heating recycling akin to a circular economy model. In addition, an independent institution should assess the projected employment outcomes of attracting data centre FDI or allowing domestic investments, unlike the unconditional incentives that produced the US figures above.

(iv) A National Data Centre Strategy that brings relevant government department such as the Department for Science, Innovation and Technology (DSIT)– which owns the UK's digital and AI strategy, including data-centre policy (designated as Critical National Infrastructure under DSIT in September 2024) and digital regulation - the AI Safety Institute; the newly established AI Economics Institute8; the Department for Energy Security and Net Zero, the Ministry of Housing, Communities and Local Government around the same table. The National Data Centre Strategy should treat data centres and cloud services as the territorial, energy-intensive industrial infrastructure it has become — including explicit scenario analysis for AI-driven growth in the Climate Change Committee’s Seventh Carbon Budget.

The UK is at an inflection point and could become a benchmark for the EU in terms of a smart national data centre strategy, and a reliable partner when it comes to EU digital sovereignty. Treating data centres as weightless cloud, rather than as the heaviest piece of new industrial infrastructure since the gas networks, is a political choice — and so is the alternative.