1 hour ago
7
South Korea just entered the AI infrastructure arms race with a facility that makes most data centers look like home offices. NAVER, the country’s dominant search and tech platform, is scaling its GAK Sejong AI data center from an initial 55 megawatts to a full 200 megawatts by 2028, backed by roughly $10 billion in combined investment from NVIDIA and Brookfield Asset Management.
The money and the muscle behind it
Brookfield, the Canadian infrastructure and asset management giant, has signed on as the exclusive capital partner with a commitment of up to $9 billion. NVIDIA is kicking in a targeted $1 billion investment directly into NAVER. NAVER itself covers whatever remains.
The project rolls out in phases. The first 55 MW deployment is scheduled to begin operations in the first half of 2027. A second expansion to 100 MW is planned for later that same year. The full 200 MW capacity hits by 2028.
The partnership builds on an earlier agreement established in June 2026 between NAVER and NVIDIA that laid the groundwork for that initial 55 MW phase.
On the hardware side, the facility will run NVIDIA’s latest AI platforms, including the Vera Rubin and Blackwell architectures.
Sovereign AI, explained
The explicit goal is building what NAVER and its partners call “sovereign AI infrastructure” — computing power that a country owns and controls domestically, rather than relying on hyperscalers headquartered in the US.
The announcement was timed to coincide with South Korean President Jae Myung Lee’s visit to an AI summit in San Francisco.
The facility is explicitly designed to serve both Korean enterprises and US clients. The longer-term vision involves building out to gigawatt-scale sovereign AI infrastructure.
NAVER is also developing its HyperCLOVA X models in collaboration with NVIDIA’s Nemotron Coalition.
Why crypto and digital asset investors should pay attention
Projects focused on decentralized compute, AI model training, and GPU marketplaces — from Render to Akash to io.net — are all betting that the world will need more AI compute than centralized providers can deliver. A $10 billion centralized facility scaling to 200 MW validates the thesis that AI compute demand is genuinely exploding, even as it simultaneously represents the centralized competition these decentralized networks are trying to disrupt.
There’s also the energy dimension. A 200 MW facility consumes serious power, and the economics of that power consumption directly affect the viability of both AI training and cryptocurrency mining. As AI facilities compete with miners for energy resources, particularly in regions with constrained grids, the cost dynamics for proof-of-work networks could shift.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·