Nvidia and South Korea’s SK Group have announced a long-term artificial intelligence partnership valued at over $500 billion, covering large-scale AI data centers and next-generation memory. SK Hynix will jointly develop AI memory including HBM4 with Nvidia and provide stable, long-term supply to meet the demands of Vera Rubin systems as well as AI training, agents, and physical AI applications. As the core of this initiative, SK Telecom plans to construct Vera Rubin DSX AI factories in South Korea in phases, reaching up to 2 gigawatts, with initial facilities expected to come online by 2027. Such scale could deploy hundreds of thousands of GPUs, indicating that ultra-large-scale AI infrastructure construction is expanding beyond US hyperscalers to overseas governments and major corporate groups. Meanwhile, Samsung Electronics signed a memorandum of understanding with Broadcom, planning more than $200 billion worth of collaboration over the next five years in areas such as HBM, sub-2-nanometer foundry services, and advanced packaging. Additionally, according to announcements from Nvidia and Naver, Nvidia intends to invest $1 billion in Naver and, together with Brookfield and others, expand a Korean sovereign AI factory capacity to 200 megawatts; the total project investment plan amounts to approximately $10 billion. These collaborations highlight how High Bandwidth Memory has become a critical resource within the supply chain for advanced AI chips. SK Hynix is currently one of the primary manufacturers of HBM and listed on NASDAQ earlier this month to secure financing for subsequent infrastructure expansion.