NHN Cloud said on the 3rd it will accelerate its push into Korea's AI infrastructure market with its artificial intelligence (AI) full‑stack brand "NHN FactoryX (FactoryX)." Based on an ultra‑large graphics processing unit (GPU) cluster and its own platform, it plans to support corporations' artificial intelligence transformation (AX) and expand public and private AI projects to help strengthen national AI competitiveness.
As the global AI industry enters a mature stage, the ability not only to secure high‑performance computing infrastructure such as GPUs but also to operate it efficiently and apply it to real services is becoming more important. In May, NHN Cloud held a press briefing and unveiled FactoryX, which provides integrated AI infrastructure, platforms, and services.
FactoryX is a brand that combines "Factory," meaning a plant that produces large‑scale AI, with NHN Cloud's experience (eXperience) and "X," which signifies the customer's AI transformation journey. To connect corporations' proof‑of‑concept (PoC)‑stage AI projects to actual business, it offers an integrated execution environment composed of three layers—infrastructure, platform, and service.
In the infrastructure institutional sector, it operates an AI cluster with a total capacity of 27.4 exaflops (EF), based on the Gwangju National AI Data Center and the AI‑dedicated data center "FactoryX Seoul." It runs Nvidia H100 GPUs alongside a domestically made NPU and has built 7,656 of the latest B200 GPUs, implementing Korea's first exascale‑class AI infrastructure.
In particular, it commercialized a single GPU cluster of 4,080 units, the largest in Korea, and applied a 100% liquid‑cooling system, cutting the annual GPU failure rate to about one‑third of that of conventional air‑cooling and improving mean time between failures (MTBF) by about 2.6 times, the company said.
In the platform domain, it maximizes GPU utilization and operational efficiency through its in‑house‑developed "GPU Live" and "AI EasyMaker." GPU Live automatically separates training and inference workloads and provides dynamic resource allocation and energy‑efficiency‑based scheduling to minimize idle resources.
In the services domain, it plans to unveil "Project X," slated for release in the second half of this year. Project X is a solution that helps even non‑developers build AI agents optimized for corporate work using only natural language, with a focus on implementing workflow automation by linking with in‑house systems and data.
NHN Cloud is also continuing to deliver results in building AI infrastructure for the public and private sectors. Last year, it was selected as the largest builder in the government's "AI Computing Resource Utilization Foundation Reinforcement Project" and carried out projects such as building Krafton's GPU cluster. The clusters built through the project—4,080 and 2,040 units based on B200 GPUs—were ranked 20th and 40th, respectively, in the 67th TOP500 list of global supercomputers announced in June. By Korea's supercomputer standards, those rankings correspond to No. 1 and No. 4, respectively.
An NHN Cloud official said, "We launched the AI‑specialized brand 'FactoryX' based on the capabilities and know‑how accumulated by operating the country's first large‑scale AI infrastructure facility, the 'National AI Data Center,' and succeeded in building clusters comparable to global supercomputers," adding, "As a corporation leading Korea's AI infrastructure market, we will responsibly and proactively support the government's public AI projects and private corporations to enhance national AI competitiveness."