AI data centers adopt liquid cooling to reduce water and energy use

Tech companies and cooling specialists are developing new methods to address the rising water and energy demands of AI data centers. As artificial intelligence expands, the heat generated by high-performance GPUs is pushing traditional air-cooling methods to their physical limits.

Nvidia has introduced its DSX system, which utilizes a closed-loop cooling method. This system allows liquid to flow directly near chips, which can exceed temperatures of 80°C. By allowing liquid to enter the servers at a warmer temperature of 45°C, Nvidia aims to reduce the need for constant air cooling, though experts note a trade-off where reducing water use often increases electricity consumption.

In Taiwan, startup Fiscu is developing immersion cooling solutions using synthetic oils to reduce data center power consumption by over 40%. Meanwhile, in Malaysia, the rapid expansion of digital infrastructure has created a growing market for specialized lubricants and cooling fluids to manage the massive heat output from increasing power draws.

Global water consumption for data center cooling reached 222 billion liters in 2025. Rystad Energy estimates that without adaptive measures, this figure could nearly triple to 644 billion liters by 2030.