LG CNS turns to liquid cooling as AI servers push data centers to the limit

LG CNS is bringing liquid cooling to a major data center in Korea as the rise of artificial intelligence (AI) pushes high-performance graphics processing units (GPUs) to consume more power and generate more heat. LG CNS said Wednesday that it is working with Naver Cloud to install direct-to-chip liquid cooling at its Samsong Data Center, preparing the facility to support high-performance AI services. The cooling system will support Naver Cloud’s operation of Nvidia’s next-generation Vera Rubin GPUs owned by the National IT Industry Promotion Agency, under the Ministry of Science and ICT. The Samsong Data Center, with a total power capacity of 80 megawatts, is expected to become a major AI factory hub, with the liquid-cooling system designed to handle server racks drawing more than 200 kilowatts. Direct-to-chip cooling uses cooling plates attached to GPU chips to absorb heat through circulating liquid, which transfers heat more efficiently than air. Conventional air cooling generally handles power densities of about 20 to 30 kilowatts per rack, while high-performance GPU servers are mo