Ellis Group Builds Korea's First AI Data Center Using Hot Water Cooling

by BAEK SEO HYUN Posted : August 4, 2026, 08:52Updated : August 4, 2026, 08:52

Ellis Group, an artificial intelligence (AI) full-stack company, is set to establish Korea's first AI data center utilizing a 'hot water cooling' method that employs cooling water above 40 degrees Celsius. The facility will be based on 2,560 state-of-the-art B300 graphics processing units (GPUs), aiming to create an eco-friendly and high-efficiency AI computing infrastructure.


On August 4, Ellis Group announced that it has been selected as a participating company in the 'AI Computing Resource Utilization Enhancement Project' led by the Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA). The company will build and operate a modular AI data center based on the B300.


A total of 2,560 B300 GPUs will be distributed across four modular data centers, with 640 GPUs in each facility. The design has been optimized for the B300 environment, focusing on cluster structure, cooling systems, networks, and energy efficiency to meet the demands of ultra-high power and high-density environments.


The data center will implement direct liquid cooling (DLC) technology and a high-speed network suitable for large-scale GPU clusters. Notably, it will introduce the hot water cooling method, utilizing cooling water above 40 degrees Celsius, along with outdoor cooling technology. This approach aims to reduce power consumption for cooling and enhance operational efficiency, contributing to the establishment of an eco-friendly data center.


Through this initiative, Ellis Group plans to expand Korea's AI computing infrastructure and establish a foundation to meet the growing demand for high-performance GPU-based AI services.


Kim Jae-won, CEO of Ellis Group, stated, "We will provide a stable and efficient AI computing infrastructure by building a modular AI data center optimized for high-performance GPU environments. We aim to strengthen Korea's AI computing competitiveness through an optimized AI infrastructure, from GPUs to networks and storage, based on domestic technology with K-PMDC."





* This article has been translated by AI.