Water infrastructure for AI data centres: a Middle East & GCC design guide

Mark Cullens

Aug 21

9 min read

Aerial view of a large data-centre campus with rooftop cooling equipment and water retention ponds

Direct answer

Water infrastructure for an AI data centre is the complete system that secures, conditions, stores, distributes, monitors and discharges water across the site. It includes utility connections, tanks, treatment equipment, pumping, pipework, cooling interfaces, backup supplies, drainage and controls. Designing only the treatment skid leaves major risks at the boundaries between those elements.

The infrastructure must follow the selected cooling architecture. Direct-to-chip, rear-door and immersion systems generally circulate fluid through closed technology and facility loops. Routine water consumption arises mainly where the final heat-rejection system uses cooling towers, evaporative condensers or hybrid wet operation. A site using dry rejection may need little process make-up, but it still requires water for domestic services, cleaning, commissioning, emergency activities and any humidification or adiabatic functions.

The design objective is therefore not maximum treatment or maximum water reuse. It is a maintainable water system that supports cooling availability through normal operation, phased growth and credible utility failures while meeting source, health, environmental and discharge requirements.

Establish the water architecture from the thermal design

Water design should begin after the project team has defined how heat moves from the IT equipment to the outdoor environment. Rack cooling, coolant distribution units, facility-water temperatures, chiller selection and heat rejection determine which circuits are closed, which consume water and which require the tightest chemistry.

A clear architecture identifies every boundary: the incoming utility connection; raw and treated storage; treatment trains; distribution headers; cooling-tower or evaporative make-up; closed-loop filling and dosing; domestic and cleaning services; backup connections; blowdown; membrane reject; drainage; and final discharge. It should also show ownership at each interface, including equipment supplied by the cooling, water, mechanical and IT vendors.

This system view prevents common gaps. A treatment plant may produce the correct quality but lack the pressure needed at the cooling plant. A storage tank may provide adequate volume but refill too slowly after an outage. A backup connection may deliver incompatible chemistry. These are infrastructure failures rather than treatment-process failures, yet they can reduce cooling capacity just as quickly.

Quantify demand and quality at every point of use

The site water balance should distinguish average, peak, start-up, maintenance and failure conditions. Annual consumption alone cannot size a summer make-up main, establish storage autonomy or confirm whether a sewer can accept simultaneous blowdown and membrane cleaning waste.

Each demand should be assigned a flow profile and quality requirement. Cooling-tower make-up may tolerate conditioned reclaimed water. Closed facility loops may need demineralised or tightly controlled fill water but only in small intermittent volumes. Domestic services require an approved potable supply. Cleaning, irrigation or construction uses may have separate standards. Combining these duties into one highest-quality system can waste energy and chemicals; serving them from one low-quality network can expose sensitive equipment.

The balance must reflect the IT load ramp. A campus may operate one data hall while later phases are under construction, producing low-flow conditions and long residence times in oversized pipes or tanks. Conversely, a future phase can exceed the firm capacity of the original utility connection even when space has been reserved for another treatment module. Hydraulic and water-quality models should therefore cover the build sequence, not only the final masterplan.

Secure an appropriate source portfolio

Source strategy is part of site selection. Candidate supplies may include municipal potable water, treated sewage effluent, industrial reclaimed water, brackish groundwater where permitted, desalinated water, district systems or tanker connections. Their value depends on capacity, pressure, quality, interruption history, connection programme, tariff and legal availability.

Two connections are not necessarily independent. They may share a treatment works, power supply, pipeline corridor or pumping station. The resilience assessment should trace common causes of failure and confirm whether the backup source can meet the required peak flow for the specified duration.

Historical water-quality data should cover seasonal and operational variation. A utility classification describes the intended use of the supply; it does not establish compatibility with the site's materials, cooling chemistry or target cycles of concentration. Representative sampling, complete ionic analysis and event records are required before final treatment guarantees are agreed.

Commercial and programme factors also matter. Land rights, pipeline routes, road crossings, elevation, utility approval and connection lead time can determine whether a theoretically attractive source is deliverable. The site-selection record should state which capacities are committed, which are assumptions and which require mitigation.

Separate technology, facility and heat-rejection circuits

The water infrastructure should preserve physical and hydraulic separation between sensitive technology cooling, closed facility loops and open heat-rejection systems. Reclaimed or variable-quality water would normally enter an appropriate make-up or heat-rejection circuit, not server cold plates, manifolds or immersion equipment.

At every heat exchanger or coolant distribution unit, the design should define temperatures, pressure regimes, approach, materials, leak consequences and water-quality responsibility. The relative pressure of adjacent circuits influences which fluid crosses if a heat exchanger fails. Leak detection and isolation sequences should reflect the consequence for both IT equipment and environmental containment.

Closed loops still need infrastructure for filling, expansion, dosing, sampling, filtration, air removal, drain-down and controlled disposal. Their low routine consumption can make these services easy to overlook. Temporary construction water should not be introduced without a flushing, passivation and acceptance plan because debris or unsuitable chemistry can compromise long-term performance.

Open cooling circuits require separate provisions for make-up control, side-stream filtration, chemical dosing, blowdown, drift and biological management. Keeping the duties explicit avoids the mistaken assumption that one generic water-treatment package can serve every interface.

Match treatment to each water-quality duty

Treatment selection should work backwards from the required product-water envelope. The process may use screening and equalisation; clarification or media filtration; cartridge filtration or ultrafiltration; softening; activated carbon; reverse osmosis; disinfection; and final pH, corrosion or scale conditioning. Not every stage belongs on every project.

Suspended solids, dissolved salts, hardness, silica, chloride, sulphate, nutrients, organic matter and microbiological activity affect different parts of the infrastructure. Ultrafiltration can protect equipment from particles but does not remove salinity. Reverse osmosis can reduce dissolved constituents but creates concentrate, consumes energy and requires suitable pretreatment and cleaning arrangements. Blending or lower cooling-tower cycles may provide a better whole-life result than extensive membrane treatment.

Separate quality tiers can improve efficiency. A central plant may produce general cooling make-up, while a small polishing unit supplies closed-loop fill or laboratory needs. Package boundaries should include analysers, chemical systems, backwash, clean-in-place facilities, ventilation, drainage and safe maintenance access rather than only the principal vessels and membranes.

Design hydraulics, storage and redundancy as one system

The hydraulic design must deliver the required flow and pressure from the source boundary to the most remote user under normal and failed equipment states. It should account for elevation, pressure zones, simultaneous demands, control-valve authority, minimum velocities, transient events and future branches.

Storage has several functions. Raw-water tanks buffer source interruption and equalise incoming quality. Treated-water tanks provide response time after a treatment fault. Break tanks can create separation between utility and site systems. These purposes should not be combined without checking how one operating event affects the others.

Tank volume should be based on defined scenarios: peak cooling duty, loss of the utility main, one treatment train unavailable, an off-spec source event and the time required to establish backup supply. Usable volume must allow for freeboard, low-level protection, sediment zones and unavailable capacity during cleaning. Refill rate is as important as nominal volume.

Redundancy should remove single points of failure without creating excessive idle equipment or stagnant pipework. Duty/standby pumps, sectionalised headers, modular treatment trains and maintainable isolation can preserve useful capacity during planned work. The design should state which component failures are tolerated and what cooling mode remains available in each case.

Control chemistry, materials and microbiological risk

Materials and water chemistry must be selected together. Chloride, sulphate, pH, hardness, alkalinity, silica, oxidants and temperature influence corrosion and scale. Mixed metals, elastomers, coatings, tower fill and heat-exchanger surfaces can impose different limits on the same circuit.

For evaporative systems, cycles of concentration reduce blowdown but increase the concentration of dissolved constituents. Conductivity is a useful control signal, although it cannot identify every scaling, corrosion or biological risk. Periodic ionic analysis, corrosion monitoring, deposit inspection and microbiological testing remain necessary.

Cooling towers and other aerosol-generating equipment require an explicit Legionella water-management programme. It should define competent ownership, control limits, monitoring, drift control, inspection, cleaning, disinfection and response to loss of control. Reclaimed-water treatment does not replace these public-health duties. The final programme must comply with the requirements that apply in the project jurisdiction.

Material protection also extends to idle and future systems. Partially commissioned loops, redundant branches and tanks awaiting later phases can stagnate or corrode. Preservation, flushing and recommissioning procedures should be included in the infrastructure operating philosophy.

Integrate monitoring, controls and operational data

Water infrastructure should operate as a coordinated utility rather than a collection of local skids. The control narrative needs to connect source availability, tank levels, treatment status, pumping, make-up demand, backup supplies and discharge capacity.

Online measurements commonly include flow, pressure, level, conductivity, pH, turbidity and disinfectant residual. Hardness, silica or organic indicators may be added where their risk and response time justify them. Critical instruments need calibration, validation and a defined response to failure; an unreliable analyser should not leave a diversion valve in an ambiguous state.

The building management or supervisory-control system should calculate the water balance, treatment recovery and abnormal losses. Trends should retain raw measurements, alarms, operator actions and calibration history. Metering at each major user allows the team to distinguish genuine cooling demand from leaks, uncontrolled blowdown or construction consumption.

Remote access and vendor support must follow the site's cybersecurity and change-control requirements. Water alarms should be prioritised by their effect on cooling autonomy so operators can see the time available before redundancy or capacity is threatened.

Plan for phased capacity and modular expansion

AI campuses are rarely built at their final load on day one. Water infrastructure should operate efficiently at early turndown and expand without interrupting live halls. The masterplan needs reserved land, utility corridors, tie-in points, electrical capacity, drainage routes and crane or maintenance access.

Modular treatment trains and pumping sets can align capital expenditure with the IT ramp. Standardised modules simplify procurement, spares and training, but the raw-water envelope and discharge constraints must remain site-specific. Copying a module without rechecking those boundaries can transfer a design that is reliable at one location into unsuitable conditions at another.

Oversized first-phase systems can create poor pump efficiency, unstable dosing and excessive residence time. Variable-speed pumping, recirculation, sectionalised tanks and minimum-flow strategies should be assessed against the actual load sequence. Temporary construction demands and the contamination risk from unfinished areas require separate controls.

Resolve discharge, recovery and residuals

Every treatment and cooling decision creates a corresponding residual. The infrastructure plan should identify cooling-tower blowdown, reverse-osmosis concentrate, filter backwash, softener regenerant, clean-in-place waste, tank washout, laboratory drains and chemical-containment discharge.

Hydraulic capacity and chemical acceptance must be checked together. Combining two individually acceptable streams can exceed sewer limits for conductivity, specific ions, pH, temperature or instantaneous flow. Equalisation, segregation, neutralisation or off-site disposal may be required.

Recovery should be pursued where it improves the complete water and salt balance. Reusing blowdown or concentrate can reduce intake, but repeated concentration may shift the limiting constituent to another system. A recovery project is credible only when the final residual has an approved destination and the added treatment can be operated reliably.

Turn the design into procurement and commissioning evidence

The design basis should become measurable supplier obligations. Guarantees need to state the source envelope, product quality, capacity, turndown, recovery, availability, utility consumption, chemical use, waste flows and performance at normal and worst-case conditions.

Interface schedules should identify battery limits, electrical and control responsibilities, drainage, chemical delivery, communications and the data required from the cooling-equipment suppliers. Factory testing can prove control logic and equipment operation; site testing must demonstrate performance within the installed hydraulic system.

Commissioning should include flushing, cleaning, passivation where required, analyser calibration, simulated failures, duty/standby changeover, storage tests and water-quality challenges. A reliability run provides better evidence than a single performance snapshot. Final handover should include operating procedures, training, critical spares, maintenance plans, baseline laboratory results and validated alarm limits.

Apply a practical Middle East and GCC delivery roadmap

Early screening should confirm the thermal architecture, preliminary demand, possible sources, discharge routes and site constraints. Concept design then develops the water balance, source envelope, quality tiers, treatment philosophy, storage autonomy, hydraulic arrangement and control narrative.

Targeted testing is appropriate where fouling, membrane recovery, biological stability or unusual chemistry cannot be resolved from available data. Detailed design should close utility agreements, process guarantees, transients, hazards, layouts, maintainability, commissioning and operator readiness.

For a regional portfolio, owners can standardise data requirements, interface definitions, modular capacities, controls and acceptance tests. Climate, source chemistry, utility reliability and discharge permissions should remain in site-specific appendices rather than being hidden inside a generic standard.

How Crescent Engineering can support the project

Crescent Engineering can translate the cooling concept and water evidence into an integrated infrastructure design basis, treatment package, control philosophy, commissioning plan and lifecycle support model. Engagement can begin with a site water balance and source assessment, then progress through process design, equipment supply, installation, training and aftermarket service.

To start, provide the cooling schematic, IT load ramp, design-weather conditions, available water analyses, source capacities, plot constraints, backup philosophy and known discharge limits.

Crescent Engineering — Water Treatment

https://www.crescent-me.com/water-treatment

References and further reading

Final equipment selection, public-health controls, utility requirements and discharge compliance must be confirmed for the specific project and jurisdiction.