Understanding Liquid Cooling Technologies for Data Centers

Liquid cooling has become a hot keyword in the data center industry as traditional air‑cooling solutions struggle to achieve PUE below 1.3, prompting stricter regional PUE targets such as 1.2 in the north and 1.25 in the south.

With the digital economy accounting for 38.6 % of China’s GDP, data‑center power consumption is soaring, and improving heat‑transfer efficiency by using higher‑capacity cooling media leads to the adoption of full‑immersion liquid cooling.

Immersion cooling submerges IT equipment in coolant, offering large heat‑transfer area, low pump power, and PUE as low as 1.04, enabling single‑rack heat removal of 100 kW.

Cold‑plate liquid cooling places heat exchangers directly on components, suitable only for regular‑shaped chips, and still leaves about 30 % of heat to be removed by traditional air cooling.

Traditional air‑cooling (wind cooling) and liquid cooling address different sides of the cooling chain; combining them yields four configurations: all‑air, air‑liquid, water‑air, and water‑liquid, each with distinct PUE characteristics.

Hybrid “air‑liquid” deployments allow both air‑cooled and liquid‑cooled sub‑rooms in the same data center, balancing higher upfront liquid‑cooling costs against lower PUE (≈1.09 vs 1.4 for pure air cooling).

From an outdoor perspective, water‑evaporation cooling offers high efficiency but faces water scarcity in northern regions, prompting a shift toward dry‑cooler or air‑cooled condensers that achieve comparable PUE without water consumption.

Overall, mixing air and liquid cooling strategies provides a more sustainable, cost‑effective solution that meets regional PUE goals and adapts to water‑resource constraints.