Naver Cloud Details AI Data Center GPU Operations

SEOUL, South Korea — Naver Cloud held a technical briefing at its Gak Sejong data center on Tuesday, the company announced.

Chief Information Officer Sang-jun Lee stated that AI infrastructure competitiveness depends on GPU resource utilization.

Naver Cloud operates GPU clusters and has developed cooling systems, power distribution, and network infrastructure for AI workloads, according to the company.

Naver Cloud deployed Nvidia SuperPod infrastructure in 2019 and has applied that experience to GPU cluster operations at Gak Sejong, the company said.

The data center uses direct outside air, indirect outside air, and chilled water for cooling.

The system adjusts cooling methods based on seasonal requirements.

The facility also includes liquid-immersion cooling infrastructure in containers.

Naver Cloud stated that it has filed patent applications for water-cooled server cooling.

The facility separates power, cooling, and server operations.

Uninterruptible power supply systems and distribution equipment are configured to meet the power requirements of GPU servers.

Servers are tested for performance, power efficiency, and operational characteristics before deployment.

The infrastructure includes monitoring and automated recovery functions for GPU resources. Resource management, deployment, and recovery processes are automated.

The AI platform supports model development, training, inference, and serving.

Naver's HyperCLOVA language model operates on this platform, which also manages GPU allocation, model administration, and scheduling.

Naver Cloud offers GPU-as-a-Service (GPUaaS) to corporate clients.

Lee stated that the company plans to expand AI infrastructure services to domestic enterprises through GPUaaS.