GridteamAI

Quarterly Sector Report

Volume Gap 39

Water Utilities · Industrial Demand + DC Cooling

by GridteamAI

UK · Industrial Demand · Data Centres · Cooling · AIGZ England & Wales · Q3 2030

The Cooling
Constraint.

AI compute, cooling water and the AIGZ map — where data-centre demand lands inside the WRMP24 framework, and what the catchments will bear.

gridteamai.com · book@gridteamai.com v2.1 · Q3 2030 Water · Volume 5

UK Industrial Water Demand + Data Centre Cooling · Executive summary05

00 · Executive Summary A two-page brief

Cooling water is the fastest-growing industrial demand — and a site-selection constraint.

UK industrial water demand is approximately 15–20% of public water supply plus direct abstraction — and within it, data-centre cooling is the single fastest-growing segment, concentrated in the five AI Growth Zones and the historical Thames Valley cluster.

The Thames Valley — Slough, West London, the M4 corridor — is the primary UK cluster and the highest water-stress region (the WRSE deficit), where cooling water is now a binding site-selection variable. The four headroom AIGZs — Northeast, North Wales, South Wales, Lanarkshire — were designated with water availability explicitly in view.

The cooling-technology trajectory runs from evaporative (high water, low energy) toward hybrid, dry and immersion / liquid cooling — the last increasingly required by AI GPU density. AI-compute growth is the single largest uncertainty in WRMP29 demand forecasting (draft 2028, final 2029).

The Q3 2030 picture

GridteamAI base case: WRMP29 industrial demand forecasting uses AI-compute low/medium/high scenarios; AIGZ designation redistributes some demand to headroom regions; Thames Valley new-build shifts to hybrid, dry and immersion cooling; non-potable supply development grows in constrained catchments.

The shape of this volume. Section 1 sets out the headline figures. Section 2 develops twelve findings, each confidence-annotated. A chapter-weight pull at p. 14 anchors the argument. Section 3 steps through the cooling load and the AIGZ map. Section 4 covers WRMP forecasting and demand. Section 5 covers cooling technology, risks and the outlook. Section 6 makes recommendations, with a one-page tear-out at p. 25.

Executive summary · ends1 · The headline figures · p. 07

1 · The headline figures09

Figure 1 · AI Growth Zones × water regionsDSIT · WRMP24

The cooling-water map: one constrained cluster, four headroom zones.

The Thames Valley historical cluster sits in the WRSE deficit region, where cooling water is a binding site-selection variable. The four headroom AIGZs — Northeast, North Wales, South Wales, Lanarkshire — were designated with water availability in view; Culham sits in the WRSE context.

A matrix of the five AI Growth Zones against their water regions. Thames Valley, the historical UK cluster (Slough, West London, M4 corridor; Thames Water and Affinity, WRSE) shown in ink as the highest UK water-stress region where cooling water is a binding site-selection variable. Culham AIGZ in Oxfordshire (UKAEA fusion plus AI compute; Thames Water and Affinity, WRSE) shown ink-dashed in the WRSE deficit context, with SESRO and WRMP29 demand integration. Four headroom AIGZs shown in teal: Northeast (Tees Valley plus Northumberland, Northumbrian Water, WRN headroom, Kielder and CCUS Track 1 adjacency); North Wales (Anglesey and Trawsfynydd, Welsh Water, WRWales headroom, Wylfa nuclear adjacency); South Wales (Port Talbot and Newport, Welsh Water, WRWales headroom, hydrogen cluster); and Lanarkshire Central Belt (Scottish Water, Gap 42 devolved, Scottish renewable supply adjacency). A footer notes AIGZ designation redistributes data-centre concentration away from the WRSE constraint, and that WRMP29 forecasts AI compute scenario demand.

Fig. 1 / Section 1

Thames Valley (ink) is the binding constraint; four headroom AIGZs (teal) have material capacity. Designation follows water as well as grid.

Quarterly Sector Report · Gap 392 · Twelve findings · p. 10

2 · Twelve findings on industrial demand10

2 · Twelve findings on industrial demandQ3 2030

Twelve findings. Each carries a confidence level.

HIGH — verified against WRMP24, DSIT AIGZ designation, Defra non-household policy or the Tranche 1 Gap 19/23 baseline. MEDIUM — named UK industry coverage (techUK, Data Centre Alliance) cross-referenced to a primary source. LOW — GridteamAI scenario projection, marked as such.

Findings 01 – 12 · summary

01Industrial demand is 15–20% of UK water — and a heterogeneous, cluster-concentrated category.
02Data-centre cooling is the single fastest-growing industrial demand segment.
03Demand splits between public water supply and direct abstraction, regulated differently.
04The Thames Valley cluster sits in the most water-stressed UK region (WRSE).
05Four of the five AI Growth Zones sit in headroom regions by design.
06Cooling-water consumption per MW varies sharply by cooling technology.
07AI compute is the single largest uncertainty in WRMP29 demand forecasting.
08The cooling-technology trajectory runs toward hybrid, dry and immersion.
09Non-potable supply is the growing option in constrained catchments.
10Abstraction reform tightens direct-abstraction availability for industry.
11Climate change raises cooling load and cuts supply yields concurrently.
12Implications for undertakers, regulators, DC operators and Government.

Findings 01 – 09 follow in detail · pp. 10 – 12 · 10 – 12 summarised on the matrix

Quarterly Sector Report · Gap 392

112 · Findings 01 – 03

2 · Findings 01 — 03The programme · the legacy · the targets

Finding 01 · HIGH

Industrial demand is 15–20% of UK water — heterogeneous and cluster-concentrated.

UK industrial water demand is a heterogeneous category — power generation, manufacturing, chemicals, food and beverage, paper and pulp, refining and petrochemicals, semiconductors, pharmaceuticals and life sciences, and the rapidly-growing data-centre cooling load. In aggregate it is approximately 15–20% of UK public water supply consumption plus direct abstraction. Demand is concentrated in cluster locations (Teesside, Humber, Merseyside, Grangemouth, the Cambridge corridor, the Wales semiconductor cluster), major urban regions, the Thames Valley data-centre cluster, and port, LNG and hydrogen sites. The fastest-growing segments through 2030–2040 are data-centre cooling, semiconductors and life sciences, hydrogen production and CCUS.

Finding 02 · HIGH

Data-centre cooling is the single fastest-growing industrial demand segment.

Within industrial demand, data-centre cooling is the single fastest-growing segment, AIGZ-concentrated and dominated by the Thames Valley, Greater Manchester, North and South Wales and Central Belt Scotland. Cooling load tracks IT load: as AI-compute density rises, cooling demand rises with it. The historical UK installed base is predominantly evaporative cooling, the most water-intensive technology, concentrated in the Thames Valley — the highest water-stress UK region. This is why cooling water has moved from a secondary site-selection consideration to, in the WRSE region, a binding constraint on where new compute can land.

Finding 03 · HIGH

Demand splits between public water supply and direct abstraction, regulated differently.

Industrial demand splits between public water supply (PWS — supplied via undertaker networks, billed at non-household tariff, forecast in the WRMP non-household category) and direct abstraction (abstracted under EA or NRW licence, regulated independently of WRMP forecasting). Power generation, chemicals, refining and some large manufacturing typically use direct abstraction; data centres, semiconductors, pharmaceuticals, life sciences and food and beverage typically use PWS. England’s non-household water market opened for retail competition in 2017 (Wales separate); customers choose a retailer while wholesale supply stays with the W+S undertakers. The split matters because abstraction reform (Gap 36) tightens one channel while WRMP forecasting governs the other.

2 · 01 – 03Findings 04 – 06 · facing

Purchase access

The CoolingConstraint.