LG Electronics has introduced a large-capacity air-cooled turbo chiller specifically designed for artificial intelligence (AI) data centers. As the heat generation and power consumption of AI servers increase, the focus is on enhancing cooling efficiency and reducing the time required for data center construction and expansion.
On September 20, LG Electronics announced the launch of its 'air-cooled oil-free inverter turbo chiller,' expanding its cooling portfolio for AI data centers with the addition of large-capacity models.
The key feature is the company's self-developed 'oil-free magnetic bearing' technology, which uses electromagnetic force to levitate the compressor's rotating shaft, reducing mechanical friction. This design eliminates the need for oil pumps and piping, minimizing the occurrence of oil films on heat exchangers that can decrease efficiency. The part-load efficiency is over 20% higher than that of conventional inverter screw chillers commonly used in data centers.
The chiller also incorporates a 'refrigerant pre-cooling' system that allows the compressor to shut down when external temperatures are low, enabling natural circulation of the refrigerant. The company claims this method can reduce annual power consumption by approximately 30% compared to traditional water-based pre-cooling systems. The refrigerant used, R-513A, has a global warming potential (GWP) that is more than 50% lower than the previously used R-134a.
In July, LG Electronics became the first domestic company to receive NVIDIA quality certification for its 600kW cooling water distribution unit (CDU). The company is expanding its AI data center cooling business by establishing a 'chip-to-chiller' system that generates chilled water in the chiller, distributes it through the CDU, and directly cools chips with cold plates, combining air cooling and liquid cooling.
To accelerate data center expansion, LG has also introduced a modular design. This approach involves pre-assembling pumps, valves, and emergency buffer tanks in containers, which can then be connected to the chiller on-site. This allows for the addition of modules as needed, enabling a gradual increase in cooling capacity in line with server expansion schedules.
Operational stability has also been enhanced. Machine learning technology detects anomalies in cooling water flow and pressure in advance, and during power outages, it maintains cooling water circulation. After power restoration, the system is designed to recover to 100% of its rated cooling capacity within 180 seconds.
Lee Jae-sung, head of LG Electronics' ES Business Division, stated, "This new product encapsulates all the values that AI data center customers need, from overwhelming efficiency to rapid on-site deployment capabilities based on modular design. We will strengthen our leadership in the rapidly evolving global AI data center cooling market."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.