[Interview] NHN Cloud: "All Four Dokpamo Companies Supported Here with Korea’s Best Performance"

Iljun Lee, Director of NHN Cloud

Four Dokpamo Companies to Utilize GPU Resources from This Week

"AIDC: Platforms and Applied Technologies Are Now a Priority"

"Starting this week, independent AI foundation model (IndeFAMO) companies participating in the National Representative AI selection project will use the 4th and 5th floors. As they require large-scale GPU resources, we plan to provide stable operational support."

Iljun Lee, Director and Head of Technology Business Division at NHN Cloud, said in an interview with The Asia Business Daily on August 7 that "while the second round of evaluations for the IndeFAMO project is underway this month, all four participating companies—LG AI Research, SK Telecom, Upstage, and Motif Technologies—will use NHN Cloud’s 'NHN Factory X Seoul' GPU resources." He explained, "That’s because the GPU cluster built by NHN Cloud offers the best performance in Korea."

Iljun Lee, Director and Head of Technology Business Division at NHN Cloud. Provided by NHN Cloud.

NHN Factory X Seoul was designed with energy efficiency in mind, featuring power and cooling infrastructure optimized for high-performance GPU-based AI computing environments. The Power Usage Effectiveness (PUE), the primary indicator of a data center’s energy efficiency, stands at 1.34. PUE is calculated by dividing total data center power consumption by IT equipment power consumption; the lower the number, the less energy is lost to non-IT infrastructure and the higher the efficiency.

Lee explained that the trend in AI data centers (AIDC) is undergoing change. "In the past, the physical size of a data center was important. Now, platforms and applied technologies have become even more critical," he said. "As AI models grow larger, not only single GPU performance but also parallel processing technology utilizing multiple GPUs simultaneously, and cluster management capabilities, are essential," he emphasized.

Citing Nvidia as an example, he noted, "As computational loads on Nvidia GPUs increase, so do the heat generation and power consumption. Nvidia has announced plans to release a 600kW product in the future, which is equivalent to the power consumed by an entire floor of a typical data center." Therefore, he stressed that acquiring advanced cooling technology for stable operation of high-performance GPUs is critical, and that continuous monitoring systems are essential to ensure stable service not only during heat waves like the current one but also in various emergency situations.

Iljun Lee, Director and Head of Technology Business at NHN Cloud. Courtesy of NHN Cloud.

To ensure stable operation of AI GPU clusters and efficient energy management, NHN Factory X Seoul is constructing a power infrastructure with a total supply of 20MW. The external power adopts a dual power supply system to ensure supply reliability, while uninterruptible power supplies (UPS) and emergency generators have been installed to maintain service continuity even in the event of failures. Additionally, telecommunications lines have been fully dualized, so that even if one line is disrupted, services can continue via another line.

The location is also counted among NHN Factory X Seoul’s strengths. Lee explained, "Located in the city center, it offers excellent accessibility, which means not only internal personnel but also external engineers for foreign equipment can offer prompt support." Lee added, "Physical latency is also comparatively low." He further stated, "Since GPU-based AI services require ongoing high-bandwidth traffic for large-scale computation and data processing, a robust security system to counter external attacks and abnormal traffic is also crucial." He explained, "We have firewalls, DDoS mitigation equipment, and IPS (Intrusion Prevention System) in place to counter such attacks."

With corporate interest in AIDC construction surging recently due to major mega-projects, Lee emphasized that reliable power supply, shorter licensing periods, and accurate demand forecasting are crucial for project success. "When running a DC business, licensing issues are a common hurdle and must be resolved as a priority," he said, adding, "Since AIDC requires substantial capital outlay for GPUs, demand becomes a critical factor for cloud service providers. If the government would present plans for resource leasing in advance, it would make it much easier for companies to invest."

This content was produced with the assistance of AI translation services.