NHN Cloud announced on the 19th that it has been selected as the GPU cluster provider for Krafton. The company has signed a contract with Krafton to provide infrastructure based on 'GPU-as-a-Service' (GPUaaS).

This project follows Krafton's declaration of 'AI First' as its core management strategy last October. It is aimed at responding to the rapidly increasing demand for AI computation. The GPU cluster serves as a key foundation for executing mid-to-long-term AI strategies, such as AI platform integration and the advancement of data automation. NHN Cloud will support Krafton's AI strategy by building a large-scale cluster.

The infrastructure will be equipped with approximately 1,000 of NVIDIA's latest 'Blackwell Ultra' GPUs. A multi-cluster structure will be used to build a large-scale computing environment. Additionally, an XDR-800G grade high-speed Infiniband Network has been applied to ensure low-latency transmission of large-capacity data.

A dynamic management architecture is applied to increase resource utilization efficiency. Even when AI development, training, and inference are conducted simultaneously, resources can be flexibly adjusted. This is intended to minimize idle resources and maximize overall cluster efficiency, allowing for elastic responses ranging from small-scale development to large-scale Large Language Model (LLM) training.

A development environment through customized GPUaaS is also provided. In line with the characteristics of large-scale training, a resource management solution based on Kubernetes and Slurm, which is suitable for High-Performance Computing (HPC) environments, has been applied. Through this, the company plans to support stable operations even in environments where various tasks are performed simultaneously.

This infrastructure will be built at the Pangyo NCC. The facility is equipped with power and cooling systems capable of operating high-density computing. Construction is scheduled to be completed and fully operational by this July.

A Krafton representative stated, "In the process of pursuing our 'AI First' strategy, the GPU cluster is the core foundation for company-wide AI operations," adding, "We expect to strengthen the scalability and efficiency of our AI research and services through NHN Cloud's excellent GPUaaS service."

An NHN Cloud official stated, "This project is a case that proves NHN Cloud's technical prowess and operational capabilities to build and operate large-scale GPU clusters in the form of GPUaaS," adding, "Starting with the launch in July, we will focus on providing stable clusters and technical support to practically support Krafton's AI First strategy."