South Korea bets big on Nvidia for sovereign AI infrastructure
260,000+ GPUs, gigawatt clouds, and a memory partnership with SK hynix.
South Korea's AI infrastructure push, unveiled in June 2026, positions the country as a test case for sovereign AI compute dependent on a single vendor—Nvidia. Major players like SK hynix, SK Telecom, Naver, LG, Hyundai, and Doosan are aligning their plans with Nvidia's platform. The government-backed National AI Computing Center and private AI factories will deploy over 260,000 Nvidia GPUs, including 50,000+ for public infrastructure. SK hynix signed a multiyear technology partnership to co-develop memory for Nvidia's Vera Rubin supercomputers, Vera CPUs, RTX Spark PCs, and Jetson Thor robotic platforms. SK Telecom aims to build a gigawatt-scale AI cloud using Nvidia DSX, with its first factory in 2027.
Naver is expanding its GAK Sejong data center to gigawatt scale and became the first Korean company in the Nemotron Coalition to fine-tune HyperCLOVA X models on Nvidia Nemotron 3 Ultra. LG, Hyundai, and Doosan extend the ecosystem into robotics, autonomous driving, smart factories, and physical AI. Despite this locally operated compute, the architecture, hardware, and memory roadmaps remain heavily tied to Nvidia. The real challenge lies in execution: obtaining power, land, cooling, and permits for gigawatt-scale facilities—constraints already affecting AI data center buildouts globally.
- Over 260,000 Nvidia GPUs planned across South Korea's government and private AI factories, including 50,000+ for national infrastructure.
- SK hynix and Nvidia are co-developing memory for Vera Rubin, Vera CPUs, RTX Spark PCs, and Jetson Thor robotic platforms.
- SK Telecom and Naver plan gigawatt-scale AI clouds using Nvidia DSX, with first factory expected in 2027.
Why It Matters
Sovereign AI infrastructure still relies on a single vendor, raising supply chain and independence questions for professionals.