Dream factory: how Nvidia’s South Korean AI deals stack up - South China Morning Post

NVIDIA CEO Jensen Huang's recent visit to South Korea underscores the company's strategic push in the Asian AI landscape. During his three-day trip, NVIDIA signed multiple agreements with key South Ko... Read original →

Jensen Huang’s Korea tour signals NVIDIA’s intent to anchor its Asian AI infrastructure strategy in Seoul. Technically, co-developed HBM with SK Hynix will alleviate bandwidth bottlenecks, accelerating GPU-memory co-design; DSX deployments will force Korean software stacks into CUDA dependency, locking in ecosystem control. On compliance, tighter U.S. export controls mean deep integration with NVIDIA exposes Korean firms—especially in autonomous driving and robotics—to secondary sanctions on global shipments. Competitively, Intel and Samsung will likely fast-track OneAPI and Exynos AI integration to counter CUDA hegemony. Over the next 12–24 months, a ‘gigawatt bubble’ may form: multiple AI factories are planned, but power and cooling constraints will reveal who can truly scale. Winners will be those merging physical AI use cases—like Hyundai’s smart factories—with efficient compute orchestration.