Water UK says AI data centre plans ignore where the water comes from

This is less a pure demand story than a permitting and cost-of-capital story. If water becomes the binding constraint, UK data-center projects can slip 6-18 months, which matters more than a small capex overrun because the revenue start date moves out while financing costs keep accruing. That favors incumbents with pre-secured utility rights and disciplined developers, and it penalizes greenfield land banks and speculative capacity in water-stressed corridors.

The cleaner second-order beneficiaries are not the obvious cloud names but water-efficiency and thermal-management suppliers: XYL, PNR, and VRT. If operators respond with closed-loop cooling, filtration, and reuse systems, that shifts spend from pure land/power accumulation toward infrastructure that improves water intensity per MW; that can support margin expansion for these vendors over the next 1-3 years. By contrast, colocation platforms with aggressive UK expansion plans risk seeing IRR compression even if topline guidance looks intact.

The contrarian risk is that the market treats this as a headline issue rather than an earnings issue unless planners actually block sites. If the government fast-tracks recycled-water infrastructure or local utilities win allowed-return capex to support new load, the bottleneck could convert into incremental spend instead of lost growth. The thesis is falsified if EQIX/DLR and UK-linked developers continue to win permits without water-related delays, or if management starts disclosing materially lower water intensity per rack than expected.