Summary
- Ofwat's draft 2026 cost-change determinations include £477m for growth pressures associated with housing and data centres.
- United Utilities has proposed water infrastructure supporting development including data centres in east Manchester.
- The draft process forces developers and utilities to reconcile speculative campus pipelines with the physical capacity required in water networks.
Ofwat has provisionally allowed £477m of additional water-sector expenditure for growth pressures including data centres, putting digital infrastructure demand directly into regulated network investment.
The allocation forms part of £3.4bn of additional spending proposed through the regulator’s 2026 cost-change draft determinations for 13 water companies. The wider package covers asset health, contaminants, growth, major schemes, and expenditure that could not be settled fully during the previous price review.
The £477m is not a dedicated data centre fund: it covers infrastructure required by both housebuilding and data centre development. United Utilities received the largest overall proposed allowance, at £995m, and has included new infrastructure intended to support development including data centres in east Manchester.
Cooling choices reach beyond the data centre fence
Electricity has dominated discussion of UK data centre constraints, but large campuses also interact with water treatment, distribution, drainage, wastewater systems, and fire protection. The size of that demand depends heavily on the mechanical design selected for each facility.
Evaporative cooling can reduce electricity consumed by mechanical refrigeration under suitable conditions, but it requires make-up water as heat is rejected through evaporation. Dry systems reduce operational water demand but can require more fan or chiller energy and larger areas of external plant.
Liquid cooling does not remove that trade-off. Direct-to-chip systems carry heat efficiently away from processors, usually through a closed internal circuit, but the heat still has to leave the building. The secondary side can ultimately reject it through dry coolers, chillers, evaporative equipment, or a combination of systems.
As AI raises rack densities, the volume of heat that must be moved through those systems increases even when IT hardware becomes more efficient per computation. Water companies therefore need to understand not just how many data centres are proposed in a region, but what cooling technologies they expect to use and how quickly each phase will be occupied.
The development pipeline makes forecasting difficult. A 200MW campus may be presented at its ultimate masterplan scale while the first operational building uses a fraction of that capacity. Some schemes will lose planning battles, encounter grid delays, or never secure a customer, yet utilities need years to plan major treatment and network upgrades.
Utilities cannot build on headline megawatts alone
That creates a familiar infrastructure dilemma. Underbuilding can leave housing, industry, and digital projects competing for constrained networks, while building too far ahead of speculative development risks loading unnecessary expenditure onto customers.
Developers can reduce that uncertainty by providing realistic phasing, credible grid connection dates, cooling specifications, and contracted demand rather than relying only on ultimate campus capacity. A project that cannot secure electricity is unlikely to consume its planned water volume, while a customer-backed development nearing construction presents a very different network requirement.
Ofwat’s cost-change consultation remains open until 24 September, with final determinations due after the regulator reviews responses. The current figures can therefore change before companies receive approval to spend.
East Manchester provides a useful case study because data centre demand there sits alongside a wider programme of housing and commercial growth. Water infrastructure has to serve all of those loads rather than being designed around a single industry, which makes allocation and cost recovery more complex.
The same issue is likely to recur elsewhere as data centre clusters move into markets where utilities have not historically planned for industrial-scale cooling demand. A location can appear attractive on land and electricity while lacking spare capacity in another system required to support construction and operation.
Planning authorities are also paying closer attention to water strategies. Developers are increasingly being asked to quantify consumption, demonstrate measures to reduce potable water use, and explain how cooling behaves during hot weather or drought conditions.
Those requirements can feed directly into design. Where water capacity is limited, a developer may favour dry heat rejection, accept a higher electrical penalty, or use hybrid systems that consume water only under specified ambient conditions. The resulting choices affect PUE, WUE, capital cost, plant footprint, and grid demand together.
Ofwat’s draft allowance gives those trade-offs a regulated financial value. Data centre growth has become large enough to appear explicitly in water network investment decisions, alongside housing and other regional development. The final determinations will show how much of that proposed expenditure survives consultation and where new capacity is ultimately built.
