Last Updated on August 8, 2026 by Karl Thompson
This article is about what happens when a hidden infrastructure stops being hidden — not because it fails, but because it starts competing for something else people want. Britain’s data centres were built quietly on industrial estates for twenty years and attracted almost no attention. They became a public issue the moment it emerged that in parts of west London, the electricity they had been allocated was electricity that could not be given to new homes. That collision is the subject of this article, and it raises a question none of the previous four have: what happens when two legitimate claims are made on the same finite resource, and the decision is made by a queue rather than by a policy?
This article is about what happens when a hidden infrastructure stops being hidden — not because it fails, but because it starts competing for something else people want. Britain’s data centres were built quietly on industrial estates for twenty years and attracted almost no attention. They became a public issue the moment it emerged that in parts of west London, the electricity they had been allocated was electricity that could not be given to new homes. That collision is the subject of this article, and it raises a question none of the previous four have: what happens when two legitimate claims are made on the same finite resource, and the decision is made by a queue rather than by a policy?
This is part of How Modern Society Works, a series on the hidden systems that shape everyday life. Each article takes something ordinary and works outwards to the system that produced it.
Why West London Couldn’t Build Homes
Imagine two planning applications arriving at roughly the same time in the same corner of west London.
The first is for a residential development: a few hundred flats on a brownfield site in Hillingdon or Ealing, precisely the sort of scheme every housing target in London depends on.
The second is for a data centre on the M4 corridor: a windowless steel-clad building, few permanent staff, no shop front, a security fence.
Both need an electricity connection. In one part of west London during the early 2020s, the local distribution network had run out of capacity to give new connections to either.
Here is the part that matters. There was no mechanism by which anybody could decide that homes should come first. As the Greater London Authority explained to developers at the time, electricity providers are not permitted to prioritise housing developments over other building projects. Connections are allocated on a broadly first-come, first-served basis, by a regulated utility applying rules that make no distinction between a block of flats and a server hall.
The result, in 2022, was that developers in Hillingdon, Ealing and Hounslow were told that grid capacity for new connections might not be available until 2030 — with some projects informed they could be waiting until 2037. The three boroughs had delivered around 11% of London’s housing supply in 2019–20. The GLA’s own briefing note attributed the constraint to a rapid influx of data centres along the M4 corridor, which it said use electricity “equivalent of towns or small cities”.
The catastrophe did not fully materialise. The GLA worked with National Grid and Ofgem on short-term fixes, and by early 2025 more than 12,000 homes had been connected. But several projects were set back, and the London Assembly’s Planning and Regeneration Committee, reporting in December 2025 under the title Gridlocked, concluded that energy capacity had become a real constraint on both housing delivery and economic growth — with London’s electricity requirements forecast to grow by at least 200%.
This is the sharpest available illustration of a point that runs through the whole of this book. Infrastructures are not simply hidden supports beneath social life. They are rival goods. What one takes, another cannot have. And when nobody has decided how to ration them, the rationing happens anyway, by whichever mechanism is already in place.
What Is Inside a Data Centre?
The technical description can be brief, because the interesting questions are not technical.
A data centre is an industrial building whose product is computation. Inside are rows of racks holding servers; above and around them, cooling systems removing the heat those servers generate; beneath the floor, power distribution; outside, transformers, backup batteries and diesel generators capable of starting within seconds of a grid failure.
The inputs are electricity, silicon and data. The output is processing. There is no loading bay for finished goods because there are no finished goods.
What makes these buildings distinctive as industrial facilities is the extraordinary density of their power draw. A conventional warehouse of similar footprint might need a few hundred kilowatts. A hyperscale data centre needs tens of megawatts, continuously, with no seasonal variation and no overnight lull, and it needs that connection guaranteed to a standard that ordinary industry does not require. It is, from the grid’s point of view, less like a warehouse than like a small aluminium smelter that never switches off.
Slough now hosts more than thirty operational data centres with around one gigawatt of total capacity, including roughly 675 megawatts of hyperscale facilities. That is genuinely comparable to the output of a large power station, concentrated in a single town on the western edge of London.
Why Are So Many Data Centres in Slough?
The clustering is not accidental, and the reasons would have been familiar to a nineteenth-century industrialist.
Slough sits immediately west of London, adjacent to Heathrow, on one of the densest fibre corridors in Europe, with fast, low-latency connections to the City’s financial infrastructure. It had a large trading estate with industrial land, existing heavy power connections and a planning regime — a Simplified Planning Zone — designed to make development straightforward.
Once a handful of operators established there, others followed, for reasons that compound. Financial institutions want their systems physically close to the systems they trade with, because at these speeds distance is latency. Cloud providers want to be where the customers already are. Network operators run more fibre to where the traffic is, which makes the location more attractive still.
Alfred Marshall described this dynamic in the 1890s as industrial localisation: firms cluster because clustering creates shared infrastructure, a specialised labour pool and a thickening of the connections between them, each of which lowers costs for the next arrival. The mills gathered at the coalfields and the canals. The data centres gathered at the substations and the fibre.
The comparison with Manchester is worth making but should not be pushed too far, because the differences are as instructive as the similarities. A Victorian mill town’s infrastructure was visible, employed thousands locally, and reorganised the entire social life of the place around itself. Slough’s data centre cluster is behind fences on an industrial estate, and most residents could not identify a single one of the buildings.
That invisibility is not modesty. It is a consequence of the fact that the product leaves through a cable.
How Are Data Centres Classified in Planning Law?
Here is a small administrative fact with very large consequences.
In the English planning system, data centres are generally classified under use class B8: storage and distribution. The same category as a warehouse.
The classification is not absurd on its face — a data centre stores things and distributes them, and when the category was framed nobody was building server halls. But the effects are substantial. A local authority assessing a B8 application is working with a framework built around lorry movements, floor area and employment density. It is not equipped to ask about a fifty-megawatt grid connection, because warehouses do not have those.
The London Assembly’s Gridlocked report recommended creating a distinct planning use class for data centres precisely so that authorities could take what it called a more strategic approach, alongside a specific data centre policy in the London Plan and local area energy plans in every borough.
This is worth pausing on, because it is an example of something that recurs across infrastructure and rarely gets noticed. Geoffrey Bowker and Susan Leigh Star argued that classification systems are never neutral descriptions of the world; they are infrastructures in their own right, which quietly determine what can be seen, counted, questioned and refused. A category built for one purpose continues to operate long after the world has changed, and things that fall inside it inherit its assumptions.
For roughly two decades, a building that draws the power of a small city has been assessed by a system designed for a shed. The consequences of that fell, eventually, on people trying to get a flat built in Hillingdon.
Do Data Centres Create Jobs?
The standard criticism of data centres is that they consume enormous resources and employ almost nobody. It is worth testing that, because the industry disputes it and the argument turns on how you count.
A report published in January 2026 by Kao Data, a data centre developer, argues that Slough’s cluster supports more than 14,000 jobs and contributes over £30 million a year in local business rates, that the Simplified Planning Zone framework generated £18 million for local councils between 2014 and 2024, and that data centres replaced Slough’s declining manufacturing employment on something close to a one-to-one basis, including around 8,000 construction jobs between 2010 and 2025.
Two things should be said about these figures, and both matter.
The first is that they are not obviously wrong. Business rates are real money that a local authority actually receives. Construction employment during a decade-long building boom is real employment. The claim that a town which lost manufacturing found a replacement industrial base is a serious claim, and Slough’s story is not the story of a place that got nothing.
The second is that the 14,000 figure is a direct, indirect and induced total produced for an industry body by the industry, and multiplier-based employment figures are among the most elastic numbers in economic analysis. The direct operational headcount of a hyperscale data centre — the people who work in the building once it is running — is typically in the low hundreds at most, and often far fewer. Construction jobs are real but temporary and largely held by workers who move on to the next site. Induced jobs are an estimate of what those workers spend locally.
The honest summary is this. Data centres generate substantial capital investment, meaningful business rates, a large temporary construction workforce and a small permanent one. Whether that constitutes a good deal for a town depends entirely on what the alternative use of the land and the grid connection would have been — which is precisely the question the planning system was not asking, because it was assessing a warehouse.
Note also who produced the evidence. In the absence of an official statistical series on data centre employment, the numbers in public circulation come from developers and trade associations with an obvious interest in the answer. This is not an accusation of bad faith. It is an observation about what happens when a sector grows faster than the state’s capacity to measure it: the sector supplies the measurements.
How Much Water Do UK Data Centres Use?
One popular criticism deserves closer scrutiny, because the evidence does not support the strong version of it.
Data centres in the American southwest have become genuinely contentious over water, and for good reason: in hot, dry regions, evaporative cooling consumes very large volumes, and it competes with agriculture and households in places where water is scarce.
Britain is not that place, and the numbers reflect it. A survey conducted by techUK in collaboration with the Environment Agency found that 64% of commercial data centres in England use less than 10,000 cubic metres of water a year — less than a typical leisure centre — with only 4% using more than 100,000 cubic metres. Around half the surveyed sites use waterless cooling systems entirely. The reason is straightforward: Britain’s climate provides free cooling for a large part of the year, so there is little commercial reason to evaporate water.
The House of Commons Library’s briefing on data centres records the industry’s argument that UK consumption is much lower than in the US for exactly this reason, alongside the counter-argument from environmental campaigners that the transparency around consumption remains poor. Both points can stand. The best-informed critical position is not that data centres are draining England’s reservoirs, but that peak abstraction during hot weather in already water-stressed regions is poorly measured and poorly planned for, and that new facilities should be required to disclose cooling method and peak-day demand.
I include this because a book arguing that hidden systems deserve scrutiny has an obligation to apply that scrutiny in both directions. On water, in Britain, the popular story is weaker than the evidence. On electricity, it is stronger.
The State Notices
Government has responded to all of this in two ways that pull in opposite directions.
In September 2024, data centres were designated part of the UK’s critical national infrastructure — the same status as energy, water and telecommunications networks, bringing them within the state’s resilience and security architecture. This is a formal acknowledgment that the country cannot function without them, and it arrived roughly two decades after it became true.
At the same time, government policy has been to accelerate their construction. The AI Opportunities Action Plan proposed AI Growth Zones, designed to speed planning decisions and improve grid connections for new AI infrastructure, with locations identified including Culham in Oxfordshire, Teesside, Newcastle and sites in north and south Wales — and a queue of local authorities from Dumfries and Galloway to Doncaster campaigning to be designated.
The geography here is the interesting part. The zones are deliberately not in the Thames Valley, because the Thames Valley has run out of power. Commercial development is already spreading to north London, Hertfordshire, Essex, Oxfordshire and the Docklands for the same reason. The constraint that produced the housing conflict in west London is now redistributing the industry across the country.
There is a recognisable historical rhythm to this. Governments once competed to attract mills, then ports, then car plants, then call centres, and now compute. Local authorities bid for the investment, offer planning certainty and hope for rates and jobs. What has changed is not the logic of regional development policy but the input the industry needs: not a deep-water harbour or a motorway junction, but a grid connection.
Rival Goods
The recurring image in this series has been of infrastructure as something beneath everyday life — buried, orbiting, warehoused, unnoticed. The data centre chapter complicates that image in a way the others do not.
Infrastructures do not merely sit under society. They occupy things. Land, spectrum, water abstraction licences, grid capacity, planning officer time — all finite, all allocated by some mechanism, and the mechanism is usually not one that anybody designed for the purpose.
In west London the mechanism was a queue at a distribution network operator, applying rules that forbade it from preferring homes to servers. That rule is not a scandal; it exists to stop utilities making arbitrary political judgements about whose project matters. But its effect was to convert a profound question of social priority — housing or computation? — into an administrative sequence. The question was answered without ever being asked.
This is what makes the episode useful. For twenty years, data centres were invisible in exactly the way infrastructure is supposed to be. They became visible not through failure but through collision — the moment their consumption of a finite resource showed up as somebody else’s shortage. That is a different mechanism of revelation from the one we have seen in previous chapters, and it may be the more common one from here on.
Because the collision will recur. Electricity demand in London is forecast to rise by at least 200%, and possibly far more, over the coming decades, driven by data centres, heat pumps and electric vehicles simultaneously. Every one of those is a policy the government supports. They will compete for the same wires.
Slough is often described as the Manchester of the digital age, and there is something to it. But the more exact parallel may be this. Manchester’s mills did not merely produce cloth. They reorganised the water, the land, the labour and the air of the place around themselves, and the reorganisation was contested at every step by people who lived there. The contest is what made industrial society legible to the first sociologists.
Data centres have reached the same stage. They are no longer only a hidden infrastructure. They have become a planning dispute, which is to say a political one — and that is usually the point at which a society starts to understand what it has built.
There is a further question this article has deliberately left open. Somebody decided to build these things here rather than somewhere else, and somebody put up the money. In west London the allocation was made by a queue at a network operator. In the fibre network that connects these buildings to everything else, it was made by a spreadsheet — and the results were, if anything, stranger.
References
| Claim | Source | URL | Why it supports the claim |
|---|---|---|---|
| Grid capacity constraints in Hillingdon, Ealing and Hounslow halted new housing connections from 2022; GLA worked with National Grid and Ofgem; over 12,000 homes connected by early 2025; London electricity need forecast to grow 200%+ | London Assembly Planning and Regeneration Committee, Gridlocked: how planning can ease London’s electricity constraints, December 2025 | https://www.london.gov.uk/who-we-are/what-london-assembly-does/london-assembly-press-releases/data-centres-delay-housing-delivery | Committee’s own published summary of findings |
| Recommendation for a new planning use class for data centres; data centres currently in use class B8; call for a London Plan policy and local area energy plans | Inside Housing, reporting the same committee report | https://www.insidehousing.co.uk/news/growth-of-power-hungry-data-centres-delaying-new-housing-in-london-committee-warns-94977 | Sets out the committee’s recommendations and the B8 classification |
| Electricity providers are not permitted to prioritise housing over other building projects; developers told supply may be unavailable until 2030 | BBC News, reporting the GLA letter to developers | https://feeds.bbci.co.uk/news/uk-england-london-62346322 | Quotes the GLA letter directly on both points |
| Some projects told they could wait until 2037 for a connection; three boroughs delivered c.11% of London housing supply | Datacenter Dynamics and BBC coverage of the GLA note | https://www.datacenterdynamics.com/en/news/report-home-building-to-halt-in-west-london-due-to-data-center-power-demands/ | Reports the 2035/2037 timescales and the housing supply share |
| Slough: 30+ operational data centres, c.1GW total capacity, 675MW hyperscale; 14,000+ jobs (direct, indirect, induced); £30m+ annual business rates; £18m from the Simplified Planning Zone 2014–2024; c.8,000 construction jobs 2010–2025 | Kao Data, The Quiet Revolution: How Data Centres Remade Slough and Secured the UK’s Digital Future, January 2026 | https://kaodata.com/discover/news/new-report-reveals-the-positive-contribution-data-centres-have-made-to-slough-creating-14-000-jobs-and-30m-annual-contribution-and-recommends-the-uk-follows-the-towns-lead | Industry-published report; figures attributed to it in the text as such |
| 64% of English commercial data centres use under 10,000 m³ water/year, less than a typical leisure centre; only 4% over 100,000 m³; around half use waterless cooling | techUK with the Environment Agency, Understanding Data Centre Water Use in England, as reported | https://www.itpro.com/infrastructure/data-centres/majority-of-english-data-centers-use-less-water-than-a-typical-leisure-center-as-operators-embrace-new-cooling-methods | Reports the survey’s headline findings and methodology |
| Data centres designated critical national infrastructure in September 2024; industry and campaigner positions on water transparency | House of Commons Library, Data centres: planning policy, sustainability, and resilience (CBP-10315) | https://commonslibrary.parliament.uk/research-briefings/cbp-10315/ | Parliamentary briefing stating the CNI designation and summarising both sides on water |
| AI Growth Zones identified at Culham, Teesside, Newcastle and north and south Wales; new capacity moving to north London, Hertfordshire, Oxfordshire, Docklands and Essex because Thames Valley power is limited | CBRE, UK Data Centres Outlook 2026 | https://www.cbre.co.uk/insights/books/uk-real-estate-market-outlook-2026/data-centres | Names the growth zone locations and the displacement of new build |
| Data centre sector’s own rebuttal of the west London blame | Computer Weekly | https://www.computerweekly.com/news/252523510/Datacentre-sector-hits-back-at-claims-that-West-London-electricty-grid-capacity-crunch-is-its-fault | Records the industry response and quotes the GLA briefing note |
Note on sourcing. The Slough employment and business rates figures are the best available but come from a report commissioned and published by a data centre developer, and the chapter says so in the text rather than presenting them as neutral. There is no official statistical series on UK data centre employment. The London Assembly’s Gridlocked report is cited here via the Assembly’s own press summary; the full report is available from the same site and should be quoted directly if you want page-level references.
Reading
- Geoffrey C. Bowker and Susan Leigh Star, Sorting Things Out: Classification and Its Consequences (MIT Press, 1999)
- Alfred Marshall, Principles of Economics (1890), Book IV, ch. X, ‘The Concentration of Specialized Industries in Particular Localities’
- Langdon Winner, ‘Do Artifacts Have Politics?’ (1980)
- Friedrich Engels, The Condition of the Working Class in England (1845) — on Manchester as an industrial landscape
How Modern Society Works
Series hub · Next: Britain’s Hidden Fibre Network →