Evaporative cooling in a data center uses the physics of water evaporation to remove heat from air or from a refrigerant loop. As water changes state from liquid to vapor, it absorbs a large amount of thermal energy from the surrounding medium, lowering its temperature without the electrical cost of a compressor-based chiller. The result, in the right climate, is one of the most energy-efficient cooling approaches available to facility operators.
The Physics Behind the Efficiency
The core principle is latent heat of vaporization. When water evaporates, it carries heat away from whatever it was in contact with, whether that is ambient air or a metal heat-exchanger surface. This process requires no compressor and no refrigerant circuit, so the only significant electrical draw is from fans and water pumps. That difference in parasitic load is what allows evaporative systems to achieve Power Usage Effectiveness (PUE) figures closer to 1.0 than conventional chiller-based designs.
Across the full spectrum of data center air cooling methods, compressor-free operation is evaporative cooling’s defining advantage. A facility running adiabatic cooling in an appropriate climate can realistically target a PUE in the low 1.1x range during cooler months, with evaporative assist kicking in during warmer periods to keep PUE from climbing above 1.2x or 1.3x.
The process is sometimes called adiabatic cooling because no heat is added or removed by an external source during the phase change. The air or fluid simply exchanges sensible heat for latent heat as water evaporates into it, or through contact with an evaporatively cooled surface.
Direct vs Indirect Evaporative Cooling
There are two main configurations, and the distinction matters enormously for data center applications. Direct evaporative cooling puts water in direct contact with the air stream entering the facility. Indirect evaporative cooling keeps that water entirely separate, using a secondary heat exchanger to transfer the cooling effect without adding moisture to the IT equipment’s environment.
| Attribute | Direct Evaporative Cooling | Indirect Evaporative Cooling |
|---|---|---|
| Mechanism | Water evaporates into the supply air stream directly | Water cools a secondary surface or fluid; no contact with IT air |
| Adds humidity to IT air | Yes, raises relative humidity of supply air | No, supply air humidity is unaffected |
| Water use | Higher, proportional to the volume of air treated | Lower, since evaporation occurs on the secondary side only |
| Best climate | Hot and dry; humid climates reduce effectiveness sharply | Broader range; more viable in moderately humid regions |
Direct evaporative cooling is simpler and cheaper to install, but it carries a real contamination risk. Outdoor air drawn through wetted media can introduce particulates, biological matter, and minerals into the IT hall unless thorough filtration precedes it. The elevated humidity also needs to stay within the ASHRAE A1-A4 envelope (roughly 20-80% relative humidity with a 15 degrees Celsius dew point upper limit for Class A1) to avoid condensation on circuit boards and corrosion over time.
Indirect systems sidestep those concerns entirely. A plate heat exchanger or rotary thermal wheel transfers cooling energy from the wet side to the dry side without ever mixing the two air streams. You get the energy savings without the humidity exposure, which is why most enterprise-grade deployments now default to indirect or two-stage indirect configurations.
Adiabatic Assist on Dry Coolers and Economizers
A common deployment pattern is not full evaporative cooling but adiabatic pre-cooling fitted to an existing dry cooler or air-side economizer. A spray nozzle system or evaporative pad wets the incoming air before it reaches the dry cooler, dropping air temperature by several degrees Celsius. That lets the dry cooler reject more heat at the same fan speed, or maintain adequate cooling at reduced fan power, without rebuilding the cooling plant.
Water is consumed only when ambient temperature rises above a set threshold, typically somewhere between 18 and 25 degrees Celsius depending on the design, so seasonal consumption remains modest in temperate climates. During cooler months the system runs as a conventional free-cooling economizer with no water draw at all. Operators who track Water Usage Effectiveness (WUE) alongside PUE find this hybrid approach keeps both metrics acceptable, rather than optimizing one at the expense of the other.
PUE vs WUE: The Central Tradeoff
This is where the honest conversation about evaporative cooling gets complicated. Full adiabatic cooling systems can cut mechanical cooling energy by 70-90% compared to a conventional chiller plant, which matters for both operating cost and carbon footprint. But every liter of water that evaporates is consumed, not returned to source. Evaporated water is gone, and in water-stressed regions that is no longer an abstract concern.
The WUE metric, defined as annual water consumption (in liters) divided by IT energy consumption (in kilowatt-hours), has moved from a secondary figure to a primary sustainability indicator. A facility with a PUE of 1.15 but a WUE of 3.0 L/kWh is trading one resource problem for another. Whether that trade is acceptable depends on local water availability and the regulatory environment the operator works within.
The U.S. Department of Energy has published guidance on data center water efficiency as part of its Better Buildings Challenge, and the conversation has grown more prominent as large facilities face community opposition over municipal water drawdown during drought conditions. Operators now typically commission site-specific water risk assessments alongside energy modelling for any large-scale evaporative deployment.
Climate Suitability: Where It Works and Where It Fails
Evaporative cooling’s effectiveness is directly constrained by wet-bulb temperature, which combines air temperature and humidity into a single figure representing the lowest temperature achievable through evaporation. The lower the wet-bulb temperature relative to the dry-bulb (actual air) temperature, the more cooling potential exists.
Hot and dry climates, desert regions in the US Southwest, parts of the Middle East, and highland areas in Africa and South America, offer the widest psychrometric range. Facilities here can run adiabatic systems aggressively and still maintain safe supply temperatures, with energy savings most dramatic year-round.
Humid climates, including coastal regions, monsoon zones, and much of Southeast Asia, compress that range to the point where direct evaporative cooling becomes largely ineffective. Indirect systems retain more value, but the advantage narrows considerably; chiller-based or liquid cooling architectures tend to deliver better total cost of ownership in these locations.
Temperate climates, including the UK and northern Europe where many hyperscale campuses have been built, sit in a workable middle ground. Free cooling hours are plentiful, adiabatic assist handles the minority of hot days, and the combination produces solid annual PUE figures without extreme water draw. The climate does much of the work, which is why the Nordic data center industry has been particularly aggressive in publishing low PUE figures.
How Hyperscalers Use Evaporative Cooling
The largest operators have shaped how the industry thinks about adiabatic cooling, both in terms of scale and in terms of the scrutiny they now face. Meta, Google, and Microsoft have all operated facilities that use evaporative cooling as a primary or assist strategy, and their public sustainability reports have made the WUE figures behind those low PUE numbers visible in ways that were not common a decade ago.
Some hyperscale facilities in dry climates have historically consumed millions of liters of water per day during peak summer conditions. That figure draws attention when local municipalities are managing drought restrictions. The public reporting obligation has pushed several operators to rethink their cooling mix, reserving evaporative approaches for moderate-density rows and shifting densest GPU clusters to liquid cooling, where heat density per rack makes air-side approaches physically impractical anyway.
The decisions made by data center infrastructure operators now increasingly reflect both PUE and WUE targets, not just uptime tiers, when choosing cooling architectures for new campuses. The shift toward rear-door heat exchangers, direct liquid cooling, and immersion sidesteps the water consumption problem entirely, since those approaches transfer heat into a closed loop rejected via dry coolers with no evaporative loss. For AI infrastructure specifically, where rack densities can exceed 100 kW, adiabatic cooling is often physically insufficient regardless of the WUE tradeoff.
Frequently Asked Questions
What is evaporative cooling in a data center?
Evaporative cooling uses the latent heat absorbed during water evaporation to cool incoming air or a secondary heat-exchange fluid. Because the process requires no compressor, it consumes far less electrical energy than chiller-based cooling. It is most effective in hot, dry climates where the gap between dry-bulb and wet-bulb temperatures is large, giving evaporation maximum thermal work to perform.
What is the difference between direct and indirect evaporative cooling?
Direct evaporative cooling introduces evaporated water directly into the supply air stream, raising its humidity as it lowers its temperature. Indirect evaporative cooling uses a heat exchanger to transfer cooling energy from the wet side to the dry supply air without the two streams ever mixing. Indirect systems are preferred in data centers because they avoid humidity and contamination exposure to IT equipment, at the cost of slightly lower peak efficiency and higher capital expenditure.
How much water does data center evaporative cooling use?
Water consumption depends on climate, system type, utilization, and whether the design is full evaporative or adiabatic assist only. Facilities in hot, dry climates running full evaporative cooling consume substantially more water than those using evaporative assist only during peak summer days. Operators track this with the WUE metric (liters of water per kilowatt-hour of IT load). Sites in temperate regions with primarily free-cooling operation tend to report low annual WUE figures because evaporation is only active for a fraction of the year.
Is evaporative cooling energy efficient?
Yes, when the climate suits it. Evaporative systems can deliver facility PUE figures in the low 1.1x range because they require no compressor-driven refrigeration cycle. The electrical load is limited to fans and pumps. For operators with access to low-cost renewable electricity, that energy saving also translates directly to a lower carbon footprint per unit of compute. The energy efficiency advantage is most pronounced in climates where evaporative cooling can operate year-round without supplemental chiller support.
When is evaporative cooling a bad choice?
Evaporative cooling performs poorly in humid climates where wet-bulb temperatures are high, leaving little psychrometric headroom for useful work. It is also a poor fit for ultra-high-density AI or GPU racks where heat loads exceed what air-side cooling can physically manage. In water-stressed regions, the WUE cost may outweigh the PUE benefit from a regulatory standpoint, and facilities with strict particulate controls may find that even indirect systems require more filtration than the energy savings justify.
