LG CNS announced on August 26 that it is partnering with Naver Cloud to establish a next-generation cooling infrastructure using liquid cooling technology. As the demand for generative artificial intelligence (AI) increases, leading to higher power consumption and heat generation from high-performance graphics processing units (GPUs), the company is proactively implementing large-scale liquid cooling technology to complement existing air cooling methods in data centers.
In collaboration with Naver Cloud, LG CNS will apply direct chip cooling (DTC), a type of liquid cooling technology, at the Samsong Data Center to lay the groundwork for supporting high-performance AI services. Previously, LG CNS took over the management of the Samsong Data Center and signed a colocation agreement with Naver Cloud.
This liquid cooling infrastructure will be built to ensure the stable operation of NVIDIA's next-generation ultra-high-performance GPU, Vera Rubin, owned by the Ministry of Science and ICT's Information and Communication Industry Promotion Agency. Vera Rubin is expected to improve inference performance by approximately 3.3 times compared to the existing Blackwell Ultra.
LG CNS believes that the Samsong Data Center will become a key hub for implementing AI factories, as Vera Rubin will be fully integrated into domestic data centers based on the liquid cooling infrastructure. With a total power capacity of 80 megawatts (MW), the Samsong Data Center is poised to establish itself as one of the largest AI factories in the country.
The DTC technology being applied at the Samsong Data Center involves attaching cooling plates through which coolant flows directly to GPU chips, effectively absorbing heat. This liquid cooling method utilizes a liquid with higher heat transfer efficiency compared to air, allowing for efficient management of heat generated by densely packed GPU servers. Typically, the cooling capacity that can be managed with existing air cooling methods is limited to about 20-30 kW per rack. However, high-performance GPU server racks are now exceeding power densities of 200 kilowatts (KW). As a result, liquid cooling technologies like DTC are gaining attention as essential for implementing AI factories.
According to market research firm MarketsandMarkets, the global liquid cooling market for data centers is projected to grow from $4.07 billion in 2026 to $27.65 billion by 2033.
In response, LG CNS plans to integrate its data center design, construction, and operation (DBO) capabilities with AI infrastructure technology to implement DTC liquid cooling, GPU server operating environments, power supply systems, and AI platform operations. Naver Cloud aims to provide high-performance AI services reliably to domestic public institutions and enterprises based on its AI platform operational capabilities and extensive experience with large-scale GPU services. The two companies plan to complete the construction and demonstration of the liquid cooling infrastructure by next year.
Cho Heon-hyuk, Senior Vice President of LG CNS's Data Center Business, stated, “Data center cooling technology is becoming a key competitive advantage for the quality and operational efficiency of AI services. We will continue to build essential infrastructure to support the growth of the domestic AI ecosystem based on LG CNS's strengths in data center DBO capabilities and next-generation cooling technology in collaboration with Naver Cloud.”
Meanwhile, LG CNS is expanding its related business by advancing its AI data center construction methods and cooling technologies. In March, the company unveiled the 'AI Box,' which consolidates 576 GPUs and 1.2 MW of AI infrastructure into a container module. This model combines LG CNS's DBO capabilities with LG Electronics' cooling solutions, and plans are underway to establish an 'AI Box Campus' in Busan, housing 50 AI Boxes.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.