Industry Analysis
Chey Tae-won’s warning signals a structural shift: HBM has evolved from a peripheral component to the core bottleneck in AI infrastructure. Technically, co-packaging 3nm logic with HBM3E/4 is accelerating demand for TSV and hybrid bonding, forcing SK Hynix and TSMC into tighter integration to secure CoWoS capacity. Geopolitically, while U.S.-Korea export controls tighten, SK’s packaging facilities in Wuxi and Chongqing may serve as supply chain buffers—though power, water, and cleanroom constraints could inflate Capex by 15–20%. Competitively, Samsung will likely fast-track HBM4 to reclaim NVIDIA share, while Micron pushes LPDDR5X-HBM hybrids via AI PCs. Over the next 24 months, persistent shortages will drive the 'memory-as-platform' paradigm; SK’s 'AI factory' vision aims to leap from component supplier to intelligence integrator—success with NVIDIA’s Vera Rubin could redefine value capture across the AI stack.
This page displays AI-generated summaries and metadata for research purposes. Original content belongs to the respective publishers.