#South Korea #Seoul #NVIDIA #LG Electronics #AI Cooling

LG Electronics Achieves NVIDIA Certification for Cooling Technology

In a significant advancement for AI cooling solutions, LG Electronics has proudly announced that it has secured official NVIDIA certification for its state-of-the-art 600kW Coolant Distribution Unit (CDU). This prestigious validation confirms the reliability and efficiency of LG’s thermal management technology according to NVIDIA's stringent AI infrastructure standards. The certification is crucial for LG's ambition to play a leading role in the AI data center cooling market, known for its rigorous demands in performance and reliability.

Meeting Rigorous Standards

LG has successfully met over 100 thorough evaluation criteria set forth by NVIDIA, which are essential for ensuring standardized cooling performance and failover capabilities within high-density AI environments. As one of an exclusive group of companies certified by NVIDIA worldwide, this accomplishment signifies LG's commitment to providing innovative and efficient cooling solutions tailored for sophisticated AI technologies.

The certification enhances LG's standing as an officially validated partner in NVIDIA’s AI Factory infrastructure. This designation offers reassurance to global hyperscalers and data center operators, enhancing supplier confidence and minimizing procurement risks during large-scale infrastructure deployments.

Advanced Technologies for AI Data Centers

The 600kW CDU from LG is specifically engineered to handle the demanding thermal requirements of high-performance AI GPUs. It features a Direct-to-Chip (DTC) liquid cooling technology that directly delivers coolant to critical components such as GPUs and CPUs. This design creates a highly efficient hybrid cooling system that complements existing air-cooling infrastructures, optimizing performance and operational efficiency.

Offering precision temperature control of ±0.25 degrees Celsius, the CDU aids in achieving thermal stability, which is vital for the reliable operation of AI servers. Additionally, it incorporates virtual sensor technology for real-time monitoring and leak detection. When paired with LG’s Data Center Cooling Control Manager (DCCM), the CDU facilitates predictive control and centralized monitoring, streamlining operations within data center management systems via standardized communication protocols.

Comprehensive Validation Process

To guarantee the utmost reliability, LG employs stringent evaluation processes at its dedicated Chip-to-Chiller validation site located in Pyeongtaek. Here, the company tests system-wide responses to various conditions, including temperature changes and stability under low-load scenarios, ensuring each component—from AI chip loads to CDUs and chillers—meets NVIDIA’s strict performance requirements.

Expanding Future Prospects

The certification is just the beginning for LG's strategic initiative aimed at expanding its AI infrastructure enterprise. Plans are in place to obtain NVIDIA certifications for an entire range of large-capacity CDU models, including the 1MW, 2.5MW, and 4MW units. With over 60 years of experience in the HVAC field, LG is positioned to deliver high-density data center solutions that guarantee precise temperature control and boost overall reliability.

LG is actively pursuing global projects, including significant contracts for hyperscaler facilities across North America and the AI-ready Sinar Mas Data Center in Jakarta, Indonesia. By leveraging its technological expertise, LG aims to enhance its global presence and foster strategic partnerships in the evolving AI data center landscape.

As James Lee, president of LG ES Company, remarked, "With the increasing scale of AI data centers, the demand for reliable liquid cooling solutions has never been more critical. Our focus on integrated technologies combined with our extensive cooling capabilities is key to positioning LG as an essential player in these next-generation infrastructures."

Note: Performance metrics such as temperature control precision may differ based on server loads and operational conditions.