Buckinghamshire AI Datacentre Approval to Be Quashed: What Government Officials Need to Do Next
The government has conceded that planning permission for a proposed hyperscale AI datacentre by the M25 in Buckinghamshire should be quashed after admitting its reasons for not requiring an environmental impact assessment were inadequate.
Angela Rayner had previously overruled the local council to approve the scheme in line with a drive to accelerate private AI investment. Her successor, Steve Reed, told the court the decision contained a "serious logical error" and accepted that permission should be set aside.
What happened
- The project, known as the West London Technology Park, is a 72,000 sq m datacentre on former landfill in Iver, promoted as attracting up to £1bn in foreign direct investment.
- Environmental groups challenged the approval over carbon and water impacts. The government conceded and agreed the permission should be quashed.
- Campaigners including Rosa Curling of Foxglove and Sonja Graham of Global Action Plan argued ministers leaned too heavily on developer assurances and underplayed energy use and water demand.
- Greystoke, the developer, declined to comment.
Why this matters for government
Datacentres were designated critical national infrastructure in September 2024, reflecting their centrality to AI and the digital economy. Peter Kyle called them "the engines of modern life".
At the same time, the UK had around 1.6GW of datacentre capacity in 2024, forecast to grow up to fourfold by 2030. Without credible assessment and enforceable conditions, approvals will be vulnerable to legal challenge-delaying investment and undermining public trust.
Immediate actions for planning and policy teams
- Require full EIA where thresholds are approached or cumulative impacts are plausible. Record clear reasoning if screening out. Keep the audit trail tight.
- Scope energy demand with realistic AI workloads. Include training and inference load profiles, peak demand, and 24/7 consumption-not just averages.
- Evidence grid capacity. Obtain statements from the transmission operator and DNO on connection dates, reinforcement needs, and curtailment risk.
- Interrogate water use. Compare cooling options (air, closed-loop, adiabatic), WUE targets, recycled water plans, and drought contingencies. Require a water neutrality plan where relevant.
- Set carbon requirements. Specify guarantees for additional renewable supply (e.g., 24/7 time-matched PPAs), on-site generation, storage, and demand response participation.
- Capture waste heat. Assess local district heating viability; require feasibility and, where viable, a delivery obligation or contribution.
- Secure enforceable conditions. Use KPIs such as PUE, WUE, and annual operational emissions, with monitoring, reporting, and penalties for non-compliance.
- Address embodied carbon. Demand whole-life carbon assessment and low-carbon materials strategy for build and fit-out.
- Respect green belt tests. Apply a rigorous alternatives assessment and the "very special circumstances" test. Prioritise brownfield and co-location with existing grid assets.
- Use legal agreements. Through Section 106 and similar, bind delivery of offsite renewables, grid upgrades, biodiversity net gain, skills, and community benefits.
What to fix in process and policy
- Standardise metrics. Require PUE, WUE, annual MWh, peak MW, carbon intensity (hourly if possible), and water source breakdown in all submissions.
- Cumulative impact assessment. Map existing and proposed datacentres within the same water catchment and grid zone; evaluate combined demand.
- Cooling and siting rules. Prefer air or closed-loop systems where water stress is high; avoid siting that depends on potable water.
- Transparent public interest test. Balance CNI status with local effects using a published rubric to reduce challenge risk.
- Decision templates. Update ministerial and inspector letters to document EIA reasoning and alternatives in plain, defensible language.
Signals to watch
- Revised decision letter and EIA screening guidance updates from the department.
- New or refreshed national policy on datacentre siting and utilities alignment.
- Grid connection queue reforms that affect delivery timelines and conditions.
What stakeholders are saying
Rosa Curling of Foxglove argued the approval was "fundamentally wrong" and overly credulous of big tech's claims. Sonja Graham of Global Action Plan said this "embarrassing climbdown" could have been avoided with proper scrutiny, noting public concern about pressure on water and power.
Practical checklist for the next datacentre application
- Confirm EIA screening determination and publish the full rationale.
- Obtain independent peer review of the energy and water chapters.
- Secure written grid and water undertaker statements with dates and capacities.
- Mandate 24/7 clean energy procurement or equivalent carbon performance standard.
- Condition heat recovery, biodiversity net gain, and construction carbon limits.
- Document green belt alternatives and why each was ruled out.
- Set monitoring and reporting cadence with public dashboards.
- Pre-agree enforcement mechanisms for KPI breaches.
Authoritative guidance
- UK Government: Environmental Impact Assessment guidance
- UK Government: Critical National Infrastructure overview
Build internal capability
If your team is being asked to assess more AI infrastructure, upskilling on AI fundamentals and infrastructure impacts will speed up better decisions. A practical starting point is this curated set of role-based programs: AI courses by job.
Bottom line
Accelerating datacentre delivery and meeting environmental standards are not mutually exclusive. They depend on solid EIAs, enforceable conditions, and clear evidence on energy and water. Get the process right upfront, and you reduce court risk while keeping strategic investment on track.
Your membership also unlocks:
