NAVER, a leading AI Cloud provider, announced a major expansion of its sovereign artificial intelligence infrastructure. Utilizing the NVIDIA DSX platform, NAVER plans to design, build, and scale full-stack, end-to-end AI factories. The rollout begins with a 55-megawatt expansion at its GAK Sejong data center, with long-term blueprints to scale operations to a gigawatt capacity to serve enterprises, governments, and local industries.
In light of the global increase in production-level demand for AI, the rise of AI factories has become necessary infrastructure that supports training, post-training, and inference processes. Through the partnership, a performance-oriented platform is developed specifically designed to lower the cost of tokens.
Advancing the Agentic Era and Physical AI
The partnership leverages NVIDIA’s comprehensive accelerated computing stack to transition AI workloads from experimental environments to full production deployment.
“Useful AI has arrived, and demand for AI factories is extraordinary,” said Jensen Huang, founder and CEO of NVIDIA. “NAVER is building AI factory infrastructure that will serve its companies, developers and industries. With NVIDIA DSX, we can help Korea scale sovereign intelligence infrastructure for the agentic era from AI agents to AI factories and physical AI.”
“NAVER is building sovereign AI infrastructure that can serve Korea’s industries and global customers with trusted, high-performance AI,” said Haejin Lee, founder and chairman of NAVER. “By building on the NVIDIA DSX platform, we can help customers move from AI experimentation to production-scale AI factories that power models, agents and real-world services.”
Hyperscale Cloud Expansion at GAK Sejong
The integration of the NVIDIA DSX platform will directly scale NAVER’s next-generation GAK Sejong hyperscale data center in Sejong, South Korea. Engineered specifically for high-density accelerated computing, GAK Sejong features advanced system automation, sustainability protocols, and robust disaster-recovery frameworks to manage heavy AI workloads efficiently.
Beyond domestic infrastructure, NAVER is expanding its AI cloud footprint into Europe and the Middle East. By combining hyperscale capabilities with the secure framework of NVIDIA DSX, the company offers public and private sector clients an alternative for data-sovereign digital services that fully comply with regional privacy regulations.
Also Read: Anyscale Integrates Natively with Microsoft Azure to Deliver Sovereign AI and Optimize Enterprise Compute Costs
Collaborative Model Development and Open-Source Innovation
The infrastructure expansion deepens a continuous collaborative effort between the two companies focused on regional model development:
- HyperCLOVA X Optimization: NAVER is refining its native HyperCLOVA X large language models by fine-tuning the NVIDIA Nemotron™ 3 Ultra open model with proprietary data assets to enhance cultural and linguistic fluency for enterprise customers.
- Global Model Coalitions: NAVER is the first South Korean entity to join the NVIDIA Nemotron Coalition, contributing technical expertise to open-source model development across pretraining, post-training, and reinforcement learning.
- AI Agent Blueprints: The company plans to launch a dedicated AI Agent Platform in South Korea during the second half of the year, built upon NVIDIA NemoClaw™ blueprints.
- Spatial Computing: NAVER is building the “Seoul World Model” a sophisticated digital twin utilizing its urban street-view data and spatial mapping software, integrated with NVIDIA Cosmos™ world foundation models.
Maximizing Efficiency through the NVIDIA DSX Platform
The NVIDIA DSX platform provides an optimized, co-designed architecture across hardware, networking, operating software, and data center facilities. NAVER will pair its operational experience in managing large-scale GPU clusters with NVIDIA DSX MaxLPS™ software, which maximizes token throughput per megawatt to drive down computational costs. Additionally, the platform is governed by NVIDIA DSX OS™, providing a modular operating layer for multi-tenant management, lifecycle automation, and system resiliency to maximize infrastructure margins.
