Nvidia strikes AI cloud, memory deals in Korea - Light Reading

NVIDIA CEO Jensen Huang has announced a series of AI partnerships with Korean firms in Seoul, following earlier deals in Taipei. These agreements underscore South Korea's growing role in NVIDIA's glob... Read original →

NVIDIA’s Korea push signals a strategic deep integration of AI compute and memory ecosystems. The SK Hynix HBM4 co-development mitigates supply constraints from U.S.-led export controls while redefining bandwidth alignment between GPUs and next-gen DRAM via 3nm/EUV co-optimization. This pressures Samsung to accelerate GAA transistor adoption and CPO packaging, and may compel TSMC to expand CoWoS capacity to retain key clients. Geopolitically, Seoul leverages NVIDIA to position itself as a neutral AI infrastructure hub—but risks steep compliance costs and mainland China market exposure if U.S. AI chip restrictions tighten. Over the next 18 months, sovereign AI clouds by Naver and LG could catalyze regional data-localization norms, forcing NVIDIA to loosen software-stack control and eroding its full-stack dominance.