NAVER, Brookfield, NVIDIA Join Forces for Korea AI

South Korea just made a big move towards seriously competing with the rest of the world for AI infrastructure. The nation’s largest internet enterprise, NAVER, has announced a major growth of its project on sovereign AI, again with two large scale companions: investment giant Brookfield and chip maker NVIDIA.

These figures are not to be taken lightly and were announced during South Korea’s leadership of the AI Summit in San Francisco by President Jae Myung Lee.

From 55MW to 200MW — A Serious Scale-Up

NAVER, Brookfield and NVIDIA announced that the sovereign AI factory infrastructure in Korea will be expanded, and that the initial deployment of NVIDIA’s AI factory (NVIDIA DSX) will be expanded from 55 megawatts to 200 megawatts. This is almost a quadruple in capacity and it’s not light work, by any means. 

The 200 megawatt expansion will be powered by the NVIDIA DSX and will be located in NAVER’s GAK Sejong hyperscale data center in Sejong, South Korea. When it goes live, the infrastructure will provide both Korea and U.S.-based AI developers with access to production scale compute to develop next generation models, agents, and AI-powered services.

Who’s Paying for It?

That’s when things get interesting from a business perspective. The project has an overall value of $10 billion, of which Brookfield will provide as much as $9 billion, as the exclusive capital partner, while NVIDIA will provide $1 billion and NAVER will cover the balance. 

This is a somewhat unusual funding model, as a chipmaker, an internet company and an asset management firm are all getting together to fund a single sovereign AI initiative. It marks the extent to which the three are taking the nation’s AI aspirations seriously.

The expansion is part of NAVER’s previous news on its June announcement to extend the GAK Sejong data center by NVIDIA DSX, and eventually, it aims to become a gigawatt-scale sovereign AI infrastructure for not only Korean enterprises and government organizations, but for international customers of AI cloud systems as well. 

The partnership with NVIDIA and Brookfield was, in Haejin Lee’s words, the company’s AI ambitions transitioning from idea to reality, as part of NAVER’s AI Factory vision to create a self-sustaining AI ecosystem and drive the nation’s AI competitiveness.

What’s Actually Inside This AI Factory

This is where it becomes relevant for all the hardware enthusiasts in the audience. The 200mw AI factory will support NVIDIA’s Vera Rubin and Blackwell platforms. It’s not only about the number of racks, but the number of the latest and most capable GPU architectures rolled out in volume. It’s a programme designed to offer a dedicated pool of compute, software and support to emerging AI businesses to develop and deploy competitive AI models and applications. 

The software side of the relationship is also included here. NAVER is also developing its HyperCLOVA X models using NVIDIA’s Nemotron 3 Ultra open models and integrating its own data and training techniques. In addition, NAVER has been the first Korean firm to join the NVIDIA Nemotron Coalition to participate in open model development work, including pretraining, post training, and reinforcement learning.

But there’s more to come on the roadmap. NAVER is set to introduce an AI agent platform in Korea in the second half of 2018, to be based on NVIDIA’s Agent Toolkit software, featuring NemoClaw blueprints. It comes alongside the Seoul World Model that is made out of its own urban street view and spatial modeling data, which uses NVIDIA’s Cosmos world foundation models. The latter is also very interesting, as it hints at future plans for physical/spatial AI that can eventually be used in robotics or even smart cities.

A Bit About NAVER

If you are not familiar with the company, NAVER was established in 1999 and has become South Korea’s largest internet company, run by the country’s top search engine and businesses across commerce, fintech, cloud, AI and robotics. It sold approximately KRW 12.04 trillion or USD 8.18 billion with continuous growth in Japan, North Ameica and Europe.

Why This Matters

This offer is part of a larger worldwide trend: nations and firms competing to create “sovereign AI” that doesn’t rely on cloud firms from other countries, giving them autonomy over their own AI systems, data and economic destiny. Korea has an edge over the other AI infrastructure bases that are emerging globally, as it has entered this game with NVIDIA’s most recent silicon and with serious institutional capital behind it.

But for gadget and tech fans, the real story is not just numbers, it’s what is constructed on top of this compute. In the near future, these features might be rolled out to consumer products, apps, and services such as Agent platforms, world models, and next gen HyperCLOVA X versions.

TheTweaks Verdict

There’s no big launch of a new toy, but it’s one of those things that quietly defines the future of gadgets and AI services a year or two in the future. A $10 billion AI factory equipped with NVIDIA chips and featuring real, institutional capital is a good indicator that Korea isn’t only buying into AI, it’s establishing a base on which it can develop its own. Watch HyperCLOVA X’s next iteration and this “AI agent platform” that’s set to launch later this year that’s probably where this investment in infrastructure will first be seen for the everyday user.