The global race for sovereign AI compute just added another datapoint as NAVER pulls in capital from both NVIDIA and Brookfield to rapidly scale its Korean data center footprint. The move reflects a broader trend where nations are treating AI infrastructure as strategic assets on par with energy grids or transportation networks, with South Korea joining Japan, France and Saudi Arabia in securing dedicated compute capacity backed by the world’s dominant AI chip supplier.
NAVER’s existing GAK Sejong facility is the centerpiece. The original 55-megawatt NVIDIA DSX deployment announced in June is now being expanded to 200 megawatts, with a stated roadmap stretching further to a full gigawatt. Brookfield is providing the bulk of the financing through a nonbinding term sheet worth up to $9B, while NVIDIA takes a $1B equity position in NAVER itself. The arrangement builds on Brookfield’s existing partnership with NVIDIA under the Brookfield Global AI Infrastructure Program.
On the technology side, the Sejong cluster will feature NVIDIA Vera Rubin and Blackwell processors integrated through the DSX software stack, which handles resource orchestration and power optimization across the multi-tenant AI factory. NAVER is also deepening its software partnership with NVIDIA by migrating HyperCLOVA X large language models onto the open-weight Nemotron architecture and joining the Nemotron Coalition — NVIDIA’s organized framework for collaborating on model pretraining and alignment work. A commercial agent platform using NemoClaw blueprints is due in Korea before the end of 2026.
