Price Check Guarantee

Price Check Guarantee

Nationwide Installations

Nationwide Installations

Next Day Deliveries

Next Day Deliveries

Free Expert Advice

Free Expert Advice

Free Site Survey
Contact Us
25/09/2026

Keeping Data Centres Cool with Evaporative Cooling

The size of your data centre affects which type of cooling will be most effective and energy efficient for you. The heat load within any IT facility is roughly equal to the IT load, measured in kilowatts (kW). In larger server rooms and data halls, the cooling needed to keep the facility running can run to several hundred kW or more, and be one of the highest running costs. Therefore making your cooling systems as efficient as possible cam dramatically reduce operating costs.

Evaporative Cooling Systems for Data Centres

In a small server room of up to around 20 racks, traditional refrigeration such as air conditioning may be the best option. One or two computer room air conditioners (CRACs) or split units provide a cool airflow around the room, and an external heat exchanger moves the heat outside the building.

For medium to large data centres, evaporative cooling can be more cost-effective and energy efficient. It is popular because it can remove large amounts of heat using very little energy.

Conventional refrigeration uses compressor-driven chillers to pump heat out of a building. Compressors are power-hungry, and data centres relying on traditional cooling typically achieve a power usage effectiveness (PUE) of 1.5 to 2 or more. PUE is total facility energy divided by the energy used by IT equipment, so a perfect score is 1.0.

In contrast, an evaporative system mostly needs pumps and fans to wet a medium and move air through it, so uses far less electricity. Data centres running evaporative and/or free cooling can achieve a PUE as low as 1.1. Free cooling uses naturally cool outdoor air or water to lower temperatures, reducing the need for evaporative or mechanical cooling.

Evaporative cooling does not need to, and in most cases cannot, reach the low temperatures of refrigeration-based air conditioning. It doesn’t have to: ASHRAE recommends server inlet temperatures of 18-27°C, and running at the warmer end of that range improves energy efficiency, although it makes the space less comfortable for the engineers who work in it.

Evaporative cooling pairs well with a free-cooling strategy. On cool days, typically in autumn and winter, outside air alone can do most of the work. On warmer spring and summer days, evaporative cooling takes over. Evaporative cooling also scales well. Its components, including evaporative pads, misting systems and cooling towers, are mechanically simpler than chillers with fewer moving parts, so they can be easier to maintain and more reliable.

Types of Evaporative Cooling

There are two main of evaporative cooling in use:

  • Direct Evaporative Cooling: passes air through wet pads and straight into the data centre or data hall. It is very efficient, but it adds humidity, which has to be managed.
  • Indirect Evaporative Cooling: this cooling uses evaporation to cool one airstream, which then cools the data hall air through a heat exchanger. The indoor air never gets wetter, and outside dust and pollutants stay out.

In larger or liquid-cooled facilities, cooling towers do the same job on water loops rather than air.

Practical Usage Considerations

The first is water use. More water could be, for some evaporative systems, be consumed on hot days than on cooler ones. This is a growing concern in drought-prone areas and in communities, including those in protected landscapes, that oppose large new data centres.

The second is humidity. Evaporative cooling works best in drier climates, such as Arizona in the US or Spain in Europe, and less well in humid countries such as Singapore. In the UK’s generally cooler climate, many sites run on free cooling for most of the year and use evaporative cooling only on the hottest days. As summer temperatures rise, will they have to rely more on evaporative cooling? Potentially, yes as free cooling from lower outdoor temperatures becomes less available during the year.

Thirdly, as with any water-based cooling system, the water must be treated to prevent scale, corrosion and Legionella, as minerals build up when water evaporates.

Water Usage Concerns for Data Centres

Over the last couple of years, the public has become far more aware of data centres and the role they play in modern life. Local objections first centred on electricity supply, but arguments against new data centres now increasingly focus on water. The AI boom is triggering a wave of construction as summers get hotter and drier.

Evaporative cooling is more energy efficient than refrigeration, but opponents worry about its impact on local water supplies. Some of the water evaporates in the cooling tower, carrying heat away, and must be replaced. More is lost through regular flushing (known as blowdown) to clear out mineral deposits.

The volumes can be significant. In UK terms, a mid-sized 15 MW data centre (with open evaporative cooling) may use 300 to 500 million litres a year, equivalent to the usage of 2,000 to 3,500 UK households.

For more information see: https://committees.parliament.uk/writtenevidence/164825/html/

Data centres also have a hidden, indirect footprint. Power stations also use water, mainly for cooling. In the US, data centres were responsible for an estimated 211 billion gallons of indirect water use through electricity generation in 2023, around twelve times their on-site use.

How Thirsty are AI servers?

AI’s power demands are high and growing. AI chips run hotter and pack more computing power into each rack than traditional servers. Growing AI demand will therefore put extra pressure on local resources, both electricity and water.

In August 2025, Google reported that a median Gemini text prompt uses about 0.26 millilitres of water. This may not seem like much, but it is only on-site cooling, not the water used to generate the electricity powering the data centre. The issue is scale: billions of queries a day, 24/7, plus the development and training of ever-larger AI models and agents. As AI data centres scale, so does the need for cooling.

Why Data Centre Location Matters

Overall, data centres are not massive water users. Water UK estimates that UK data centres use around 6.6 million litres a day on average, just 0.14% of the 14 billion litres put into public supply daily. In the US, data centres use about 2% of national water consumption.

https://www.congress.gov/crs-product/R49057

The problem is timing and location. Evaporative cooling needs the most water on the hottest days, exactly when households, farms and power stations also hit peak demand. Water UK estimates UK data centres’ average peak demand at about 14 million litres a day, and says water companies have received requests from individual developers for up to 21 million litres a day.

For more information see: https://committees.parliament.uk/writtenevidence/164766/html/

The UK is now facing this directly. More than three-quarters of UK data centres are in the already water-stressed east and south-east of England. A 2026 Global Action Plan study found 84% of proposed UK data centres are planned for areas the Environment Agency classes as water-stressed. The Environment Agency’s analysis of metered data from around 200 sites found a clear summer peak in demand. Meanwhile, England faces a projected public water supply shortfall of nearly 5 billion litres a day by 2050.

There is a nuance, though. UK operators are not yet required to report their water use, and most UK sites use little. One industry analysis found 67% use less than 1,000 m³ a year, while the six largest users account for about 65% of the sector’s consumption.

How Data Centres Are Recucing Water Usage

As with almost every data centre technology, cooling system manufacturers are innovating fast in response to market demand. Several approaches now let data centres use far less water:

  • Closed-loop Cooling: circulates water through sealed pipes and reuses it, cutting consumption by up to around 70% compared with open evaporative systems
  • Direct-to-chip and Immersion Cooling: brings liquid right to the processors; NVIDIA’s newest racks can be cooled with water at 45°C, warm enough that chillers are not needed
  • Non-potable Water: such as treated wastewater or harvested rainwater, keeps data centres off the drinking supply; some US towns now insist on it, and Water UK has called for it to be required where possible
  • Smarter Siting: in cooler, wetter places allows free-air cooling for much of the year
  • Heat Reuse: sends waste heat into district heating networks instead of evaporating it

The industry tracks progress with Water Usage Effectiveness (WUE): litres of water per kilowatt-hour of IT energy. Amazon, for example, reports cutting its WUE from 0.25 to 0.12 L/kWh since 2021.

Summary

Data centres use water in two ways: directly, by evaporating it to cool servers, and indirectly, through the power stations that supply their electricity. The indirect share is usually larger and less often mentioned. Cutting water use often means using more electricity, and vice versa. The real questions are where a data centre is built, how it is cooled, and whether it draws on drinking water in a region that is already running dry. In the UK especially, we cannot fully answer those questions yet, because operators do not have to report their water use. Mandatory water reporting, like the rules the EU has already introduced, would be a sensible first step.

For more information on the data centre cooling technologies supplied by Server Room Environments, please contact our Projects team.

Blog main image 3538816 1790501666

Related blog posts

15/05/2026
Next Article
A Server Rack Guide : Selecting The Right Server Racks and Data Cabinets
Blog box fixed 3515121 1784468450

In any IT environment, server rack and cabinet choices are fundamental to the overall physical layout, hardware reliability and uptime. The choice of rack type, cabinet dimensions, power distribution, and cooling approach all have direct consequences for airflow management, IT hardware, uptime, working life, and the ability to scale infrastructure over time.

Read more ...