SK Group and NVIDIA on July 24, 2026 unveiled a $500-billion-plus strategic partnership to build gigawatt-scale AI factory infrastructure in South Korea, pairing NVIDIA's Vera Rubin accelerated computing platform with SK Telecom-owned data centers and long-term supply of SK hynix HBM4 memory. The two sides signed letters of intent at NVIDIA's AI Summit in San Francisco.
SK Telecom to build a 2-gigawatt NVIDIA DSX AI factory
Under the deal, SK Telecom will construct a 2-gigawatt AI cloud in Korea using NVIDIA's DSX full-stack AI factory architecture, with the first facility slated to come online in 2027. The buildout will run on NVIDIA Vera Rubin accelerators tightly coupled to SK hynix HBM4 stacks — the highest-bandwidth memory currently on NVIDIA's roadmap. NVIDIA says DSX is designed to deliver the lowest token cost at maximum energy efficiency for sovereign, agentic and physical AI workloads.
SK hynix locks in a multi-year memory supply
Separately, SK hynix will co-develop and optimize next-generation AI memory with NVIDIA under a long-term agreement, giving Jensen Huang's company assured access to HBM4 and successor products at a moment when supply is the single biggest constraint on data-center scaling. "In the AI era, competitiveness depends not just on how effectively AI is utilized, but on how much intelligence we can produce," SK Group Chairman Chey Tae-won said in the release.
Korea's play to become a global AI hub
The announcement is the largest of several partnerships NVIDIA unveiled during South Korean President Jae Myung Lee's San Francisco AI Summit trip. It sits alongside a newly launched Samsung humanoid robotics division, an NVIDIA-ABB physical AI push, and last month's Databricks $188B strategic round, positioning Korea as a top-three global consumer of accelerated compute alongside the US and Saudi Arabia. NVIDIA-KAIST also stood up a joint agentic AI research lab in Seoul earlier the same week.
Reporting based on coverage from GlobeNewswire, NVIDIA, RTT News and Yahoo Finance.