Look, let's be honest: NVIDIA and Naver are teaming up to push the boundaries of 'sovereign AI infrastructure' using the NVIDIA DSX platform. Naver is starting with an initial 55MW capacity, with ambitions to scale to gigawatt levels, creating a full-stack, end-to-end AI platform designed to support enterprises, various industries, and government bodies.

Jensen Huang, the founder and CEO of NVIDIA, weighed in: 'As we enter the age of practical AI, the hunger for AI factories is going through the roof. Naver is laying the groundwork for an infrastructure that empowers businesses, developers, and the entire industrial ecosystem. With NVIDIA DSX, we’re helping Korea expand its sovereign intelligence—from AI agents to AI factories and physical AI—in this new agentic era.'

Naver Chairman Lee Hae-jin added, 'We are building sovereign AI infrastructure that brings high-performance, trustworthy AI to both our domestic industries and global partners. By leaning on the NVIDIA DSX platform, we’re helping our clients move past mere experimentation and into the realm of productive AI factories where they can actually deploy models, agents, and services.'

As 'practical AI' moves from buzzword to reality in operational environments, these AI factories are becoming the backbone for training, fine-tuning, and inference tasks. Built on NVIDIA’s accelerated computing, Naver’s AI factory will serve as the engine for domestic firms, manufacturers, and government agencies to cook up their own intelligent solutions. This cloud setup is being integrated into Naver’s 'Gak Sejong' data centre. Located in Sejong City, Gak Sejong was purpose-built for high-density NVIDIA computing, featuring heavy automation, sustainability metrics, and robust disaster recovery to keep things running efficiently.

Naver isn't just looking at its own backyard; it’s aiming to lead global digital innovation from Gak Sejong. By expanding its infrastructure into Europe and the Middle East, Naver intends to meet the demand for sovereign AI, providing a reliable alternative for governments and enterprises that need to stay within local regulatory and data sovereignty lines. Using NVIDIA’s full-stack AI platform, they are accelerating their own data centre business while powering these sovereign models and agentic services.

This new cloud infrastructure is the fruit of a partnership focused on refining domestic sovereign and physical AI models. Naver is taking the NVIDIA Nemotron 3 Ultra open model and fine-tuning it with its own proprietary data and expertise to enhance its 'HyperCLOVA X' model. The goal? To offer international clients a model that’s not just powerful, but culturally nuanced. This Nemotron-based HyperCLOVA X is also set to underpin sovereign AI initiatives across Europe and the Middle East.

Naver holds the title of the first Korean company to join the Nemotron Coalition, a move aimed at supercharging global innovation. They’ve also confirmed plans to launch an 'AI Agent Platform' in Korea later this year, based on the NVIDIA NemoClaw blueprint. Furthermore, they are developing a 'Seoul World Model,' which marries the NVIDIA Cosmos world foundation model with Naver’s own street-view data and spatial modelling tech.

The NVIDIA DSX platform is effectively an end-to-end co-design stack for these AI factories, covering everything from chips and systems to software and infrastructure, all aimed at slashing token costs and getting operations off the ground faster. It’s a marriage of Naver’s expertise in massive GPU clusters and hyperscale data centres with NVIDIA’s tech. The DSX MaxLPS software aims to squeeze the most out of every megawatt to keep token costs at rock bottom, while the DSX OS acts as the operational layer—handling lifecycle management, runtime consistency, and automated health checks to help cloud providers actually turn a decent margin.