Naver, Brookfield and Nvidia will invest a combined $10 billion (about 14.63 trillion won) to expand South Korea's sovereign artificial intelligence (AI) infrastructure. Based on this, the companies plan to expand the initial scale of the AI factory being built at the Sejong data center "Gak Sejong" from 55 megawatts (MW) to 200 MW, and over the long term to build gigawatt (GW)-class AI infrastructure.
According to Naver on the 25th, the plan was announced to coincide with President Lee Jae-myung's attendance at the AI summit in San Francisco. The newly built AI factory will be based on Nvidia's "DSX AI Factory" platform for building AI factories, and will be established as a 200 MW-class facility capable of housing approximately 100,000 graphics processing units (GPUs).
The infrastructure will be built at the hyperscale data center "Gak Sejong" in Sejong City, and will provide commercial-level computing resources so that domestic and international AI innovators can develop next-generation AI models, AI agents, and AI-based services.
This project is an expansion of the "Gak Sejong" DSX construction plan announced last month. Starting from this, Naver presented a roadmap to build GW-class sovereign AI infrastructure over the long term, targeting domestic industry, public institutions, and global AI cloud customers.
The investment structure was also disclosed. Brookfield, as the exclusive capital partner, will raise up to $9 billion, and Nvidia will strategically invest $1 billion. Naver will contribute the remaining amount of the $10 billion project, oversee project operations and infrastructure construction, and drive the AI factory business.
Lee Hae-jin, Naver founder and board chairman, said, "Through Nvidia's strategic investment and the infrastructure supply agreement with Brookfield, the AI factory business has entered a full-fledged execution phase." He added, "Based on cooperation with global partners, we will build a sovereign AI ecosystem and contribute to strengthening Korea's AI competitiveness."
As an Nvidia cloud partner, Naver has experience operating hyperscale infrastructure based on a full-stack AI platform. This 200 MW-class AI factory will apply Nvidia's next-generation AI platforms, Vera Rubin and Blackwell. The built infrastructure will be operated as a dedicated resource pool for AI startups and companies, providing computing resources, AI software, and technical support together.
The two companies will also expand cooperation in the AI model field. Naver is advancing HyperCLOVA X based on Nvidia's Nemotron 3 Ultra open model using its own data and training technology, and, as the first Korean company to join the Nvidia Nemotron Coalition, is participating in open model development including pre-training, post-training, and reinforcement learning.
Naver plans to introduce an AI agent platform based on the Nvidia agent toolkit, including the Nvidia NemoClaw blueprint, in Korea in the second half of this year. In addition, using its own street view and spatial modeling data, Naver is developing a "Seoul World Model" based on the Nvidia Cosmos world foundation model.
