Air cooling, evaporative cooling and liquid cooling: a Middle East & GCC strategy guide
Direct answer
Air cooling, evaporative cooling and liquid cooling are not competing labels for a single decision. They describe different stages of the heat path. An AI data centre may use liquid at the rack, a closed facility-water loop across the building and either dry or evaporative equipment outdoors. The correct design is the complete chain that removes the required heat at the site’s climate, water constraints and availability target.
In the Middle East and GCC, high rack density makes liquid capture increasingly valuable, while high ambient temperatures make final heat rejection demanding. Evaporation can improve thermal efficiency but consumes water; dry rejection reduces routine water demand but may need more area and electricity at peak conditions. Hybrid systems trade between the two.
Selection should therefore use hourly climate and load modelling, clearly separated fluid circuits, defined water quality, credible failure modes and lifecycle cost. A product name alone does not establish the energy, water or resilience outcome.
Follow the heat path from chip to atmosphere
Cooling design becomes clearer when the team maps every step taken by the heat. The path begins at processors, memory, power electronics and other IT components. Heat then moves into room air, a cold plate, a rear-door heat exchanger or an immersion fluid. It passes through one or more facility circuits before reaching chillers, dry coolers, cooling towers or another final rejection device.
Each transfer has a temperature difference, pressure requirement and equipment boundary. Those conditions determine fan and pump energy, heat-exchanger area, chiller lift and the opportunity for compressor-free operation. They also determine which fluid is close to electronics and which circuit can safely accept variable-quality make-up water.
A project that evaluates “air versus liquid” without drawing this complete chain can miss the decisive issue. Liquid capture at the rack does not automatically make the campus water-cooled, and air-cooled rooms can still reject heat through a water-cooled chiller plant.
Understand the role of air cooling
Traditional air cooling uses server fans and room-side equipment to carry heat away from the racks. Good air management—containment, controlled pressure, sealed bypass paths and suitable supply temperature—reduces recirculation and unnecessary fan work. It remains practical for lower-density spaces and for the portion of a hybrid rack load not captured by liquid.
At higher density, the required air volume and local distribution become difficult. Fan energy, acoustic conditions, floor or ceiling space and the consequences of a failed airflow path can constrain the design before total building cooling capacity is reached. A single average kilowatts-per-rack figure can hide local hot spots.
Air cooling should therefore be retained where it is effective, not assumed as a universal default. The design needs rack-level load information, airflow containment performance and a clear allowance for residual heat from liquid-cooled equipment.
Define where liquid cooling begins and ends
Direct-to-chip systems transfer heat into coolant at cold plates. Rear-door units place an air-to-liquid heat exchanger at the rack boundary. Immersion systems place components in a dielectric fluid and transfer heat through an external exchanger. These approaches have different fluid compatibility, leak consequences, service procedures and proportions of captured heat.
Most deployments use a cooling distribution unit between the technology coolant loop and the facility-water system. The CDU controls temperature, pressure and flow while preserving separation between fluids. Its heat exchanger, controls, pumps and redundancy are therefore critical to IT availability.
The design must state the coolant composition, materials, cleanliness, dissolved-gas limits, filtration, fill procedure and acceptable temperature classes. It should also identify who owns each boundary: server supplier, rack integrator, CDU vendor, mechanical contractor or operator. Ambiguous responsibility can leave sampling points, expansion capacity or leak detection outside every package.
Compare dry, evaporative and hybrid heat rejection
Dry coolers and air-cooled chillers transfer heat to outdoor air without routine evaporation. Their capacity falls and electrical demand can rise as ambient dry-bulb temperature increases. Equipment size, approach temperature, fan power, noise and available plot area are important GCC constraints.
Cooling towers and evaporative condensers use evaporation to approach the outdoor wet-bulb temperature. They can lower condensing temperature and improve thermal performance, but require make-up, treatment, blowdown, drift control and microbiological management. Their water demand follows heat load, weather and cycles of concentration.
Hybrid equipment can operate dry for part of the year or at lower load, then introduce wetting or evaporation when needed. The value depends on the actual control sequence. A hybrid label does not prove low annual consumption if the system remains in wet mode for most high-load hours.
Heat-rejection selection also affects plant topology. Distributed roof equipment can shorten pipe routes but complicate maintenance, access and future expansion. A central plant can simplify water treatment and operator attention, although common headers and constrained replacement routes require careful redundancy. These practical differences should be included in the comparison alongside calculated efficiency.
Model GCC climate and load together
Option comparison should use representative hourly dry-bulb, wet-bulb and humidity data alongside the IT load profile. The model should include early phases, partial load, design peaks and future rack densities. Annual totals alone do not show whether the system maintains capacity through the hottest sequence.
For each hour, the calculation should estimate cooling capacity, compressor energy, fan and pump energy, evaporation, blowdown and any limits on supply temperature. Water quality affects the achievable cycles of concentration, so the source analysis belongs in the same model.
Sensitivity cases are essential. A warmer-than-historical year, degraded heat exchanger, unavailable cooling cell, restricted water supply or higher liquid-cooling capture ratio can change the preferred option. The result should be a range with stated assumptions rather than one apparently exact consumption figure.
Establish the water and energy trade-off
Water and energy performance should be reported together. Reducing evaporation may increase dry-cooler fan power or chiller lift; improving thermal efficiency through evaporation may increase water withdrawal and consumptive use. The balance varies by season, load and source.
Power usage effectiveness and water usage effectiveness provide useful portfolio signals, but they need boundary definitions and supporting operational data. A low site WUE can conceal water consumed in electricity generation or off-site treatment, while a single PUE value can conceal peak electrical capacity required in emergency dry mode.
The business case should include utility connection capacity, water and power tariffs, treatment chemicals, membrane replacement, blowdown disposal, equipment area and carbon objectives. Where water is scarce or politically sensitive, resilience and social licence may outweigh a narrow operating-cost result.
Water accounting should distinguish withdrawal from consumption. Cooling-tower make-up is withdrawn at the site, but only the evaporated portion is consumed locally; blowdown leaves as a managed liquid stream. Reclaimed-water use can reduce reliance on potable supply without eliminating treatment energy or discharge obligations. Clear terminology allows operators, regulators and customers to compare options on the same basis.
Protect the technology and facility loops
Technology coolant and closed facility water are high-value assets even though their routine make-up is small. Chemistry affects corrosion, deposition, electrical risk, elastomers and heat-transfer performance. Introducing untreated utility or reclaimed water directly into these circuits is inappropriate.
The infrastructure should provide controlled filling, expansion, air removal, filtration, dosing, sampling, drain-down and disposal. Pressure regimes at heat exchangers influence the direction of leakage, while detection and isolation sequences should protect both electronic equipment and the building system.
Commissioning cleanliness is particularly important. Fabrication debris, temporary water and oxygen introduced during phased works can damage small channels or accelerate corrosion. Flushing, passivation, quality acceptance and preservation of idle branches should be documented before live equipment is connected.
Design open evaporative circuits properly
An open cooling circuit concentrates the constituents in its make-up water as pure water evaporates. The permissible cycles depend on hardness, alkalinity, silica, chloride, sulphate, suspended solids, microbiological activity and equipment materials. Conductivity control alone cannot manage every limit.
Treatment may include filtration, softening, membrane processing, disinfection, corrosion and scale inhibitors, and side-stream solids removal. The correct train works backwards from the tower chemistry and the final blowdown route. More treatment is not automatically better if it produces a difficult concentrate or fragile operating regime.
Cooling towers also require a Legionella water-management programme with competent ownership, monitoring, cleaning, drift control and response limits. Public-health obligations remain in force whether the make-up is potable, desalinated or reclaimed.
Plan for mixed-density and phased campuses
Many campuses will contain conventional air-cooled halls, hybrid racks and high-density liquid-cooled zones. The mechanical arrangement should serve these loads without forcing every circuit to the coldest temperature. Separate temperature tiers or heat exchangers can reduce chiller lift and preserve flexibility.
Early phases create turndown challenges. Large central plants may operate inefficiently below their design point, while oversized pipes and tanks produce long residence times. Modular pumping, treatment and heat-rejection cells can align capacity with the IT ramp if tie-ins and control logic are planned in advance.
Technology change should be treated as a scenario, not a prediction. Reserved pipe routes, electrical capacity, plant space and connection standards allow a future change in liquid-cooling share without overbuilding every first-phase asset.
Evaluate reliability and maintenance
Reliability analysis should follow functional cooling paths. Redundant chillers do not protect the load if both depend on one treatment skid, one make-up header or a shared control panel. Similarly, multiple CDUs may share a facility-water bottleneck.
Maintenance access, isolation and replacement routes are part of availability. A component is not practically redundant if servicing it requires shutting the common system. Chemical delivery, tower cleaning, filter backwash and safe access to roof equipment need space and operating procedures.
The failure matrix should state the remaining thermal capacity after loss of each critical component and source. It should also include the electrical impact of switching to dry or backup operation. Operators then know whether an incident changes efficiency, redundancy or supported IT load.
Supply-chain resilience deserves attention as well. CDU pumps, proprietary controllers, tower gearboxes, membrane elements and water-quality analysers can have very different lead times. Critical-spares selection should be tied to the failure analysis, with preservation and periodic testing for stored components. Standardising selected equipment across campus phases may shorten recovery without forcing every cooling technology into one design.
Specify performance at system level
Procurement specifications should define conditions at the boundaries of each package: entering fluid temperature, leaving temperature, flow, pressure drop, heat duty, ambient condition, fouling allowance, fluid quality, turndown and allowable power. Without consistent conditions, supplier efficiencies cannot be compared.
Guarantees should cover normal, peak and failed-equipment states. For wet systems, include make-up and blowdown volumes at the stated chemistry. For dry and hybrid systems, include fan and compressor power at GCC design conditions. For CDUs, define capacity, control stability, leak response and fluid cleanliness.
Factory tests establish equipment functionality, but integrated site testing proves the heat path. Commissioning should demonstrate mode changes, duty/standby transitions, control coordination, alarm handling and performance at representative load.
Use a practical option-selection process
Begin with rack loads, permitted temperatures, heat-capture assumptions, design weather, water sources, plot constraints and availability requirements. Develop complete heat-path options rather than comparing isolated products.
Model each option at hourly and design conditions, then test water restriction, equipment failure and future load cases. Compare electrical capacity, annual energy, water use, treatment residuals, maintainability, programme and lifecycle cost using the same boundaries.
Record why the selected architecture remains acceptable under uncertainty. The final design basis should show which assumptions require confirmation and which changes would trigger a review. This creates a defensible bridge from concept selection to supplier guarantees and operation.
The recommendation should identify the conditions under which a different option would become preferable. A change in utility water allocation, rack-density forecast, electrical connection capacity or heat-reuse opportunity may alter the result. Recording those decision thresholds makes the study useful throughout design instead of freezing an early concept after its assumptions have changed.
How Crescent Engineering can support the project
Crescent Engineering can connect the selected cooling architecture to its required water qualities, treatment trains, storage, pumping, controls and discharge arrangements. Support can include water balances, source analysis, modular equipment, integration with cooling suppliers, commissioning and operator training.
Useful starting information includes rack-density forecasts, proposed liquid-cooling technology, facility-water temperatures, design-weather data, candidate heat-rejection systems and complete source-water analyses.
Crescent Engineering — Water Treatment
References and further reading
US Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design
US Department of Energy — Cooling Water Efficiency Opportunities for Federal Data Centers
ASHRAE Handbook — Data Centers and Telecommunications Facilities
US Environmental Protection Agency — Water Reuse for Industrial Applications
Final cooling temperatures, fluid specifications, public-health controls and performance guarantees must be confirmed with the relevant equipment suppliers and authorities.
Comments