This page was machine-translated and may differ from the original. View original
At the 'Dell Technologies Forum 2026', SK Hynix Vice President Joo Young-pyo emphasized that as AI proliferation advances, memory bandwidth and efficiency emerge as critical bottlenecks, and SK Hynix must address this through HBM, 3D packaging, PIM, CXL, and memory pooling. As generative and inference-type AI increase, TB-level memory capacity, high-speed bandwidth, and sophisticated data processing have become essential, and a transition from GPU-centric architecture to memory-centric architecture is underway. HBM demand is surging rapidly, with memory efficiency, computational optimization, and system integration identified as key competitiveness factors for AI.
*The full article is paid content for e4ds+ members. Click the ad below to read it free of charge.
To request a correction, reply or follow-up report on this article, see how to file a request. Previously published statements are collected in corrections & replies.














