NAVER and NVIDIA Partner to Build Korea's First Gigawatt-Scale AI Factory

South Korea has been watching the AI infrastructure arms race from an uncomfortable position: technologically sophisticated, home to some of the world’s most advanced semiconductor manufacturers, but still dependent on American hyperscalers for the AI compute capacity that modern frontier models require.

A deal announced on June 7 is designed to change that.

NAVER — South Korea’s dominant internet company, operator of the country’s most widely used search engine, and the developer of the HyperCLOVA X large language model — has entered into a strategic partnership with NVIDIA to build what will become Korea’s first gigawatt-scale AI factory infrastructure. The vehicle for the deal is NVIDIA’s new DSX platform, the company’s most comprehensive attempt yet to offer a complete, end-to-end blueprint for sovereign AI deployments.

“Useful AI has arrived, and demand for AI factories is extraordinary,” Jensen Huang said in remarks tied to the announcement. “NAVER is building for the agentic era — and gigawatt-scale infrastructure is how you serve it.”

What the DSX Platform Actually Is

The NVIDIA DSX platform is not a product in the conventional sense. It is a full-stack architecture — spanning chips, systems, networking, software, data center facilities design, and operational tools — that NVIDIA is offering as a complete blueprint to national and regional AI infrastructure builders who want to build at scale without having to assemble the pieces themselves.

At the software layer, DSX MaxLPS maximizes token throughput per megawatt, delivering the lowest token cost at scale. DSX OS handles lifecycle management, runtime operations, health automation, resiliency, and multi-tenant AI factory management — the kind of operational complexity that becomes unmanageable at gigawatt scale without purpose-built tooling.

For NVIDIA, DSX is a strategic response to a competitive threat: Google’s custom TPU silicon and Amazon’s Trainium chips are both designed to reduce hyperscaler dependence on NVIDIA GPUs. By packaging its hardware advantage into a complete platform that also handles software and operations, NVIDIA is trying to make its stack the default choice for sovereign AI deployments — national infrastructure projects where governments and large regional players are building from scratch rather than migrating from existing hyperscaler relationships.

NAVER’s Infrastructure Roadmap

The partnership unfolds in stages. NAVER’s GAK Sejong hyperscale data center — described as a next-generation facility with advanced automation and sustainability features, located in Sejong City, South Korea — serves as the anchor site.

Deployment begins at 55 megawatts in the first half of 2027, scaling to 100 megawatts later that year and 200 megawatts in 2028. The long-term stated target is 1 gigawatt — a scale that would make GAK Sejong one of the largest AI-dedicated data center complexes in Asia.

For context: 1 gigawatt of AI compute is roughly equivalent to what the largest current hyperscaler AI clusters consume globally. Training a frontier model like GPT-5.5 or Claude Fable 5 requires enormous sustained compute bursts. Inference at scale for hundreds of millions of users requires sustained capacity. A gigawatt-scale AI factory is not research infrastructure — it is production infrastructure at a scale that enables genuine frontier model development and commercial AI services.

NAVER’s technical roadmap for the partnership includes several distinct workstreams: development of next-generation HyperCLOVA X models fine-tuned on NVIDIA’s Nemotron 3 Ultra; creation of a “Seoul World Model” using proprietary urban data and NVIDIA Cosmos technology (a generative world model framework designed for physical AI and robotics applications); and an AI Agent Platform launch targeted for the second half of 2026, built using NVIDIA NemoClaw blueprints.

NAVER is also joining the NVIDIA Nemotron Coalition — a consortium of regional AI developers building models atop NVIDIA’s foundational Nemotron architecture — which gives it access to shared research on model pretraining, fine-tuning techniques, and inference optimization.

The Sovereign AI Context

The geopolitical backdrop for this deal is as important as the technical details.

South Korea has declared sovereign AI — the capability to develop, train, and serve advanced AI models on domestically operated infrastructure — a national strategic priority. The country’s government has committed to ensuring that sensitive government data, defense applications, and critical AI workloads are not processed on foreign-owned cloud infrastructure, particularly following the data sovereignty debates that followed several high-profile incidents involving U.S. cloud providers and Korean government data.

NAVER’s position is unique: it is simultaneously Korea’s most important domestic AI champion and a potential sovereign AI infrastructure provider for other governments in Asia and beyond. The company already offers secure, regulatory-compliant AI services to government and enterprise customers in Europe and the Middle East — markets where buyers want frontier AI capabilities but cannot or will not use American hyperscalers for regulatory or national security reasons.

Haejin Lee, NAVER’s EVP of Cloud Business, framed the partnership in exactly these terms: “By building on the NVIDIA DSX platform, we can help customers move from AI experimentation to production-scale AI factories — wherever they are, under whatever regulatory constraints they operate.”

What It Means for the Broader Ecosystem

The NAVER-NVIDIA deal is the largest of several NVIDIA sovereign AI partnerships announced around the same time, in a week that also saw NVIDIA deepen alliances with SK hynix (on high-bandwidth memory), SK Telecom (on telecom AI), LG, and Hyundai — all Korean conglomerates making concentrated bets on AI infrastructure.

The pattern reflects a structural reality of the current AI moment: the countries and companies best positioned to benefit from the AI transition are not necessarily those with the most data scientists or the best model research — they are those with the manufacturing, infrastructure, and energy resources to build at scale. South Korea, with its semiconductor manufacturing capabilities, robust energy infrastructure, and government willingness to commit capital to strategic technology priorities, fits that profile well.

For NVIDIA, the deal extends the DSX platform’s reach to one of the most sophisticated technology markets in Asia, and adds a credible regional anchor tenant that can demonstrate what gigawatt-scale AI production actually looks like in practice. For NAVER, it is a declaration that the company intends to compete not just as a Korean internet company, but as a serious global AI infrastructure player.

The AGK Sejong data center’s first megawatts go live in less than a year. What runs on them — and for whom — will be the real test of whether Korea’s sovereign AI moment has arrived.