LG电子开发适配英伟达Vera Rubin的D2C冷却AI服务器机柜，计划明年量产并切入万亿韩元级市场

LG Electronics (066570.KS), which has grown into the world's largest home appliance company, is expanding its territory into a core area of artificial intelligence (AI) data center construction, leveraging its cooling and mechanical engineering technologies. Having demonstrated world-class competitiveness in the heating, ventilation, and air conditioning (HVAC) systems that manage data center heat, LG Electronics has succeeded in developing AI server racks using next-generation cooling technology that directly minimizes semiconductor heat generation, along with mechanical design capabilities accumulated through home appliance manufacturing. LG Electronics plans to break into the AI server rack market led by Taiwan, growing its data center business revenue to the scale of tens of trillions of won, completing LG Group's AI data center turnkey supply capabilities, and expanding alliances with global big tech companies.
According to industry sources on the 12th, LG Electronics' Production engineering Research Institute (PRI) developed a rack prototype in the first half of the year that meets the specifications of Nvidia's next-generation AI computing platform "Vera Rubin," and plans to begin mass production next year after completing reliability evaluations in the second half. At the LG PRI site in Pyeongtaek, Gyeonggi Province, a cluster has been built with dozens of racks arranged in parallel totaling 1.8 megawatts (MW), where AI server rack prototype production and quality verification are being carried out simultaneously.
LG Electronics envisions entering the AI server rack market to secure revenue of trillions of won and, over the medium to long term, growing the business to the scale of tens of trillions of won. According to market research firm MarketsandMarkets, the AI server rack market is expected to grow at an average annual rate of 34% over the next five years, reaching $448.6 billion (approximately 672 trillion won) by 2030.
Currently, the AI server rack market is led by Taiwanese companies such as Foxconn, Quanta Computer, and Delta Electronics, which have partnerships with Nvidia. These companies generate revenue of 1 trillion to 2 trillion won per quarter from AI server rack production alone. However, as the AI server rack supply chain has become excessively concentrated in Taiwan, a movement to diversify supply bases is spreading among big tech companies, and LG Electronics has identified such demand and moved into product development.
Home Appliance Cooling and Mechanical Capabilities Infused... Jensen Huang Praises "LG Cooling Essential"
LG Electronics was able to succeed in developing AI server racks that minimize heat generation because it has accumulated "heat-handling technology" over 58 years. After launching Korea's first home air conditioner in 1968, the company has produced compressors and heat exchangers, which are at the core of HVAC. Cooling technology once used only in homes expanded into large commercial "chillers" applied to buildings and factories, and recently the stage has broadened to AI data centers that emit enormous amounts of heat.
Three years ago, LG Electronics developed hybrid cooling technology combining air-cooling and water-cooling methods, and this year it independently developed a "direct-to-chip (D2C)" cooling device that dissipates heat by flowing coolant near the chip. The power consumption of Nvidia's latest AI accelerators exceeds 100 kilowatts (kW) per unit, and is expected to rise to more than 200 kW in the next-generation Vera Rubin. Successor products are being discussed at the 600 kW level. As the power density that AI server racks must handle rises sharply, LG's D2C technology is expected to enhance the performance of next-generation server racks.
Nvidia CEO Jensen Huang also expressed his expectations when he visited Korea last month, saying after meeting LG Chairman Kwang-mo Koo at the LG Twin Towers in Yeouido, Seoul, that "LG's cooling technology is essential for gigawatt (GW)-scale data centers."
AI server racks go beyond simple assemblies, requiring mechanical engineering capabilities as they densely integrate hundreds of chips in a limited space. Through developing refrigerators and air conditioners, LG Electronics has advanced its mechanical engineering technologies, from developing core components such as motors to thermal and fluid control, vibration and noise management, and structural design. AI server racks are seen as closely connected to home appliances in that they must assemble dozens of compute trays along with cooling, power, and communication devices into a single finished product and ensure stable performance even at high temperatures.
The industry observes that LG Electronics, which possesses both precision mechanical design and mass production capabilities, could emerge as a strong server rack player able to compete with Taiwanese server manufacturers.
LG Affiliates' Data Center Capabilities Fully Mobilized... The Final Puzzle Is the 'AI Rack'
LG Electronics views the AI rack as the final puzzle piece to complete the group-wide AI data center turnkey strategy. To target domestic and overseas AI data center markets, LG Group is advancing "One LG," a turnkey strategy that bundles together LG Electronics' chillers, LG Energy Solution's batteries, and LG CNS's data center construction capabilities. By adding server rack manufacturing to this, the group envisions providing the entire data center construction process at once. In the AI server rack business, a plan is being reviewed under which LG CNS, which has data center engineering capabilities, will participate together with LG Electronics at the center.
"The only area left empty in One LG's data center portfolio was server manufacturing, which connects Nvidia's graphics processing units (GPUs) and boards to complete the rack," an industry official said. "Combining LG Electronics' manufacturing and cooling technology with LG CNS's server design and system integration capabilities could deliver competitiveness beyond Foxconn."
As LG expands its business scope into server rack manufacturing, it has come to possess most of the capabilities needed for AI infrastructure construction, excluding chips such as accelerators and memory, and the power sector including power plants and cables.