Nature Positive Role of the Technology Sector 2025

Water use effectiveness (WUE) comparison across types of heat rejection  FIGURE A5 Low end of range High end of range2.91 2.91 0.00 0.000.70 0.022.10 0.802.80 0.00 Free air cooling Filtered outside air directly used for cooling without additional refrigeration Some water may be used in winter for humidificationOutside air passed through a water-saturated medium to be directly used for coolingDirect evaporative coolingIndirect evaporative coolingMechanical coolingCooling tower Indoor air or liquid cooled through a heat exchanger with outside air that has passed through a water-saturated mediumEquipment such as chillers and compressors used to mechanically cool air or liquid used for coolingWater is cooled in a cooling tower through direct contact with outside air and then used for cooling Range driven by variation in regional climates and water quality – hotter, drier temperatures and lower air quality consume more waterLow end of range reflects air-cooled chiller, while top end reflects water-cooled (and can range from 2.3 to 2.8 L/kWh)Litres/kWh Note: As these cooling technologies can be used with either air cooling or liquid cooling at the server level and in combination with each other, which will impact efficiency, achievable WUE values with each technology can range beyond the bounds listed. Source: see endnote.193 Facilities seldom use only one type of cooling – liquid cooling may be used directly to cool the chips, but other equipment (such as power supplies) still needs to be air cooled.194 Closed- loop server and facility-level cooling are another promising area – while these systems still require water input during set-up, compared to evaporative liquid cooling methods they can reduce dependency on freshwater withdrawals. A joint study by Ramboll and Grundfos estimated 25% potential water savings across data centres in Europe through a combination of reduction, reuse and reclamation measures.195 Finally, data centre operators must also consider their indirect water requirements, such as from their power demand (e.g. for cooling at thermal power plants). Material inputs and land use Construction inputs for data centres can be substantial. One study indicates that nearly 130 tonnes of emissions are embedded in a typical 530m2 building’s shell, through use of concrete, steel and other inputs. The same study found that a 1 MW data centre (very small relative to typical developments today) could require 66 tonnes of copper, 15 tonnes of plastic, 33 tonnes of aluminium and 171 tonnes of steel for the electrical operations alone, on top of the building’s construction.196 For more details on the impacts of the metals in use in construction, refer to the following World Economic Forum reports: –Nature Positive: Role of the Mining and Metals Sector –Nature Positive: Role of the Cement and Concrete Sector Developing data centres also requires land – and they are typically located in urban or suburban areas to limit latency concerns and utilize existing infrastructure. Land use for data centres currently remains low, with the global footprint estimated at ~100 square kilometres, or roughly the land Nature Positive: Role of the Technology Sector 61