SK Telecom plans to build a 2-gigawatt AI data center using Nvidia’s Vera Rubin chips and SK Hynix’s (SKHY) HBM4 memory. The first facility is expected to come online in 2027. Nvidia and SK Hynix will also work together on high-bandwidth memory for AI training, AI agents, and robotics.
The memory deal is an important part of the announcement. High-bandwidth memory has been a key supply constraint for advanced AI systems, and a long-term partnership with SK Hynix could help Nvidia secure enough HBM4 capacity as Vera Rubin production expands.
Meanwhile, NVDA shares dropped nearly 1% on Friday, closing at $206.84. Nevertheless, the stock is up about 11% year-to-date.
Nvidia Deepens Its South Korea Ties
The broader Korea plan also includes Nvidia’s work with NAVER and Brookfield. The companies plan to expand NAVER’s AI data center in Sejong to 200 megawatts, with about 100,000 GPUs based on the Vera Rubin platform. Hyundai Motor Group (HYMLF) also outlined plans to use 50,000 Nvidia Blackwell GPUs for robotics, autonomous vehicles, and other physical AI applications.
