NAVER, NVIDIA and Brookfield plan $10 billion Korean AI factory expansion to 200MW

NAVER, NVIDIA and Brookfield Asset Management announced on Saturday a $10 billion expansion of Korea's national AI factory buildout, growing the planned deployment of NVIDIA DSX infrastructure at NAVER's GAK Sejong hyperscale data center from 55 megawatts to 200 megawatts. Brookfield would fund up to $9 billion of the project as exclusive capital partner, and NVIDIA is reported to be taking an equity stake of roughly $1 billion in NAVER. The financing is not closed: coverage of the deal describes Brookfield's commitment as a nonbinding term sheet, and NVIDIA's investment as conditional on NAVER securing the Brookfield capital.

A 145 megawatt jump at GAK Sejong

The expansion, announced through a release on PR Newswire, would nearly quadruple the DSX footprint planned for the Sejong site. Under the structure the companies describe, Brookfield provides capital, NVIDIA contributes technology and investment, and NAVER funds the remainder itself. The release says the expanded facility is meant to give Korea and US based AI teams access to production-scale compute for "building next-generation models, agents and AI-powered services." NAVER also plans to deploy an AI agent platform in Korea in the second half of the year, powered by NVIDIA Agent Toolkit software, according to the same announcement.

Financing announced, not closed

The deal structure matters as much as the headline number. Brookfield's role as exclusive capital partner covers up to $9 billion, but that commitment is reported as a nonbinding term sheet rather than a signed financing agreement. NVIDIA's roughly $1 billion equity investment in NAVER is likewise reported to be conditional on NAVER locking in the Brookfield money. Until those pieces resolve, the $10 billion figure describes an intention, not a completed transaction, and GaaS News is reporting it as such.

Seoul Economic Daily reported that NAVER founder Lee Hae-jin framed the partnership as a way to accelerate construction of ultra-large AI factories while "scaling up Naver's data center operation know-how" through the two partners' global networks. The same report says NAVER's longer-term ambition runs to a full gigawatt of AI infrastructure, positioned as a "sovereign AI" ecosystem that does not depend on any single nation or corporation.

Compute earmarked for agents

For the agentic AI market, the notable detail is that agents appear by name in the release language, sitting alongside next-generation models as a first-class workload the compute is being built for. Sovereign agent infrastructure is becoming a pattern rather than a one-off: Alibaba has been retooling its cloud around agent workloads, as GaaS News covered in our report on the agent-native cloud push, and demand signals from the open-model world point the same direction, as in Hugging Face's compute demand figures. National buildouts on the NAVER model would give domestic agent platforms a home that does not route through a foreign hyperscaler.

Part of a larger Korea-US weekend

The announcement did not land in isolation. It arrived amid a wave of Korea-US technology agreements unveiled on July 25 around the San Francisco AI Summit, which Korean press reports, including coverage cited by TechTimes and the Korea JoongAng Daily, have put at roughly $950 billion in combined value. Seoul Economic Daily's related reporting pointed to presidential office statements on large chip cooperation deals announced the same day.

The near-term test is whether the term sheet converts into binding commitments. NAVER has published no closing timeline, and none of the three companies has said when the conditions on NVIDIA's stake would be satisfied. What is concrete is the direction: a fourfold capacity expansion at one site, with agents written into the purpose statement, and one of the largest infrastructure investors in the world negotiating to carry most of the cost.