Section 1
The scale of the machine
Every time an AI model generates text, creates an image, or processes data, it runs through physical servers3. Those servers require three tangible resources: physical land to sit on, massive electric power to run, and millions of litres of water to cool down3.
Beyond "Carbon": The Three Material Pillars
Focusing only on carbon allows tech companies to claim net-zero status while quietly draining local water tables or monopolizing electricity grids3. We track land, water, and power together because a low-carbon energy source can still consume millions of litres of water and crowd out local infrastructure3.
Section 2
The Land Around You: Blocked homes and empty hangars
Data centres do not integrate into communities. Unlike traditional developments that bring shops, cafes, offices, or community hubs, a data centre is functionally a vast, windowless hangar surrounded by perimeter fencing and security patrols. They take up significant local acreage while generating far fewer permanent local jobs than developers typically claim9: in the UK, hyperscale sites are usually found to employ 20-50 people on an ongoing basis, against developer estimates that run into the thousands once construction and projected "indirect" jobs are added in10.
Preventing homes from being built
The most immediate impact on UK residents is housing7. In West London (Hillingdon, Hounslow, and Ealing), 29 data centre developments have reserved so much electrical grid capacity that the Greater London Authority warned new housing connections could be blocked until 20357. Across the UK, grid bottlenecks linked to data centre requests have held up construction on roughly 50,000 homes7.
Havering: Green belt vs. local carbon
The proposed £14.7bn East Havering Data Centre Campus on green belt land at North Ockendon demonstrates the trade-off11. While offering a corporate ecology park, planning documents admit it will emit 1.2 million tonnes of CO2e annually-consuming nearly 20 times Havering borough's entire carbon budget by 2038, and almost 1% of the UK's11. Local residents' groups describe it as a direct loss of farmland for zero local community utility.
Brick Lane: A housing scheme, reduced to a data centre
In July 2026, the government overruled Tower Hamlets Council and approved a £500m redevelopment of the Truman Brewery estate including a data centre-despite the council's own plan for the site promising around 800 homes24. The approved scheme delivers just 44 homes, six of them for social rent, in a borough with more than 30,000 households on the social housing waiting list25. It's a direct illustration of what a site could otherwise have been used for.
Section 3
The Water You Drink: Environmental justice and bills
Cooling high-density AI servers requires vast quantities of potable water1. During English droughts and hosepipe bans affecting over 27 million people12, data centre corridors continue drawing heavily from domestic water infrastructure. In Slough - Europe's largest data centre hub, with at least 32 sites currently running13 - Thames Water opened a formal investigation back in 2022 after a single proposed site applied to draw 25 litres of water per second14. This isn't a one-off problem: officials on the government's own AI Energy Council have since flagged lengthy water connection timelines as an ongoing, unresolved bottleneck for new data centres across the UK15.
Official UK Government Warning: GDSA Report
As highlighted in our March 2026 newsletter, the UK government’s own Government Digital Sustainability Alliance (GDSA) acknowledged the direct environmental justice threat of AI water consumption:
“The siting of water-intensive infrastructure in already water-stressed areas, particularly those with existing socio-economic vulnerabilities, raises significant environmental justice concerns. This practice could disproportionately impact local communities, potentially leading to increased water bills, rationing, or reduced environmental flows, thereby transforming a technical water management issue into a matter of social equity.”
Section 4
The Cost of Electricity: Grid strain and fuel poverty
The UK electricity grid faces unprecedented pressure from tech expansion4. Data centre developers have requested 125 gigawatts of grid connections-almost triple the UK's entire peak national power demand of around 45 gigawatts15.
Why this lands hardest on Black UK communities
Data from the Joseph Rowntree Foundation shows that Black British households are more than twice as likely to live in poverty compared to white households (39% vs 18%), with nearly two-thirds (64%) falling behind on basic household bills5. This fits a wider pattern: people from Black, Asian, and ethnic minority communities in Britain are, on average, more affected by climate and environmental crises than white British people, largely because of existing income and housing inequality6. When data centre growth strains the grid and drives up systemic energy delivery costs, it creates a regressive financial burden on the households least able to pay.
Ireland: A warning on grid takeover
By 2023, data centres in Ireland consumed 21% of all metered electricity-more than every urban and rural home in the country combined. Grid operators paused new connections in Dublin until 2028. The UK is currently following the same trajectory.
Section 5
Every prompt multiplied: The cost of daily AI
While model training uses massive energy bursts, 80–90% of AI's total electricity use happens during "inference"-the everyday prompts generated by millions of users3.
Section 6
Global diaspora links: From the UK to Mississippi
The physical supply chain powering UK AI data centres connects directly to Black communities across the globe.
To supply electricity to the UK grid and designated "AI Growth Zones," North Yorkshire’s Drax power station burns wood pellets imported from the US South16. Its primary supplier operates a processing plant in Gloster, Mississippi-a town with a 71% Black population where nearly 40% of residents live in poverty17. Federal research shows wood pellet plants are more than twice as likely to be located in predominantly Black and poor communities18. Gloster residents filed a federal lawsuit against Drax over air pollution violations in 202519. The energy powering UK servers draws directly on environmental harm experienced by the African diaspora abroad.
Section 7
Taking on Big Tech: Regulation, organising, and collective power
While Big Tech corporations appear monolithic, community action, regulation and accountable systems can chip away at their power. Here are some strategies we have seen:
1. Organising and Litigation in the US and UK Context
US activists have leveraged state-led lawsuits and jury systems to force settlements (e.g., the up to $18 billion Meta settlement with a coalition of state attorneys general over child safety20). In the UK, public outrage can be used to make Big Tech expansion politically painful for elected representatives in key parliamentary constituencies. Social action groups such as the UK's Foxglove legal have worked with other organisers to scrutinise data centre plans and have won concessions against the government8.
2. Good Legislation Beats One-Off Settlements
Local council budgets remain under sustained pressure after over a decade of austerity, with most English councils still facing real-terms funding gaps21, leaving regulators timid when facing tech lawyers. Structural legislation tends to outlast settlements: even the record $18 billion Meta settlement let the company avoid trial and admission of guilt, where analysts estimated it faced $100-150 billion in exposure, and left its core surveillance-advertising business model untouched23. One key lever is anti-monopoly regulation: the European Commission recently designated ChatGPT, Reddit, and Roblox under the Digital Services Act (DSA) as Very Large Online Platforms or Search Engines, its strictest regulatory tier2. Pushing the UK's Competition and Markets Authority (CMA) to adopt similar oversight could provide a structural brake on unvetted data centre expansion.
3. Demanding Seats at the Table
The UK's official AI Energy Council was established to plan energy infrastructure for AI, yet its founding members were drawn entirely from tech firms (Google, Microsoft, Amazon, ARM) and energy companies and regulators (NESO, EDF, Scottish Power, National Grid, Ofgem)22. Across four meetings between April 2025 and April 2026, that pattern has held: not one community, housing, or civil society representative has attended15. Shouldn't there be community representation on all national and local data centre planning boards?
| Actor | Accountability Actions |
|---|---|
| Community Groups | Mobilise local planning objections around housing, water bills, and land use ("eating the elephant one mouthful at a time"). |
| Government & CMA | Enforce strict anti-monopoly rules (similar to the EU DSA); integrate housing impact checks before granting data centre power connections. |
| Local Councils | Adopt mandatory planning clauses requiring full water/air disclosures and zero diesel generator emissions. |
| AI Energy Council | Open mandatory governance seats for affected local communities and civil society groups. |