NHN Cloud has completed building a large-scale artificial intelligence (AI) infrastructure based on more than 7,600 high-performance graphics processing units (GPUs) and begun supplying services to the public sector as well as industry-academia-research.
NHN Cloud said on the 13th that it has started operating an AI-dedicated data center (Yangpyeong Data Center) located in Yangpyeong-dong, Yeongdeungpo-gu, Seoul.
The project was promoted through last year's "AI computing resource utilization infrastructure enhancement project" by the Ministry of Science and ICT and the National IT Industry Promotion Agency (NIPA). Built under the national mid- to long-term AI infrastructure expansion strategy, the facility's infrastructure is managed as a national asset, and resources are allocated and used in line with government policy.
The Yangpyeong Data Center houses infrastructure built on a total of 7,656 NVIDIA B200 GPUs. Of these, 6,120 will be used for national AI projects promoted by the Ministry of Science and ICT. With the center's launch, NHN Cloud has given priority supply of some of these GPU resources to industry, academia, and research institutes since Apr. 1.
The center uses a liquid-cooling method that circulates coolant to reduce heat. Through this, about 4,000 GPUs were grouped into a single cluster to create a large-scale high-performance computing environment, and the company said energy use can be cut by 15%–20% compared with conventional air cooling.
NHN Cloud reflected its experience improving the power usage effectiveness (PUE) by applying a high-density design of 8 kilowatts (kW) per rack and an outside-air cooling method when building the Pangyo Data Center "NCC1" in this design. It also said it minimized errors by analyzing in advance the power load and heat variables that can occur in large-scale GPU environments using operational data from the Gwangju National AI Data Center it has run for the past three years and reflecting the results in the design.
An NHN Cloud official said, "With the operation of the Yangpyeong Data Center, we have established a practical foundation to meet large-scale AI computing demand," and added, "We plan to serve as a core infrastructure leading the spread of domestic AI services and the growth of the ecosystem by advancing customized GPU environments tailored to the characteristics of each industry."