Scotland mandates environmental review for £5bn Cato AI data centre

The Scottish Government's intervention establishes a new regulatory floor for hyperscale infrastructure, specifically targeting facilities exceeding 50MW. For the streaming and AI sectors, this signals that the era of rapid, low-friction data centre expansion in the UK is shifting toward a more rigorous planning phase where environmental costs are weighed against economic gains. As streaming platforms increasingly rely on AI-optimized compute power, these mandates could extend project timelines and increase pre-construction costs for infrastructure providers like ILI Group. Watch for whether Fife Council adopts the proposed moratorium on all hyperscale applications, which would further stall regional capacity growth.

The Cato AI data centre proposal sits within a broader wave of hyperscale infrastructure investment across the UK, where planning authorities are increasingly scrutinizing energy and water demands. In July 2026, Water UK warned MPs that single proposed data centres are requesting up to 3 million litres of water a day for cooling, roughly what 3,500 homes consume daily, while the Environment Agency's National Framework for Water Resources explicitly excludes data centre demand from its long-term supply forecasts. The UK currently has around 500 data centres, a number expected to grow by 20 per cent over the next five years to meet AI computational demands, according to iNews reporting on the sector's expansion and regulatory gaps. ILI Group's Auchtertool site, if approved, would add significant capacity to Scotland's grid and water infrastructure at a time when national planning frameworks have not yet integrated such demand. The regulatory tightening around the AI data centre opposition reflects a wider push for mandatory environmental reporting across the UK. The Royal Academy of Engineering published a report calling on ministers to make reporting on energy consumption, water consumption, and carbon emissions mandatory for all data centres, with Prof Tom Rodden warning that without such action, AI development could cause irreparable environmental damage. The EU's Common Union Rating Scheme already requires data centres to report water consumption, and a UK government appendix on AI water footprint recommended mandatory location-specific reporting for all facilities above 1MW, aligned with EU standards. Scotland's new 50MW threshold for environmental review effectively creates a domestic equivalent, closing the loophole that allowed Fife Council to initially approve the Cato application without full scrutiny. Technical data on water consumption underscores why the Scottish Government intervened. A March 2026 WRc report for MOSL found that data centre potable water consumption in England stands at approximately 1,879,000 cubic metres per year, with 65 per cent concentrated in just six facilities. The report concluded that sector growth driven by AI expansion is not yet integrated into water resource planning, and that the mismatch between five-year planning cycles and rapid data centre development makes it difficult to anticipate demand in water-stressed regions. Thames Water has stated that a large data centre can use between 4 and 19 million litres of water per day, equivalent to the daily demand of 50,000 households. For streaming platforms that depend on -driven encoding, recommendation engines, and content delivery optimization, these infrastructure constraints could influence where and how quickly new compute capacity becomes available across the UK market.