NVIDIA's GPU Demand Spurs SK Hynix to Fast-Track HBM4 Chip Production
SK Group's Chairman, Chey Tae-won, highlighted a significant supply challenge faced by NVIDIA in response to the growing market demand for its GPUs. Speaking at the group's AI summit in Seoul, Chey disclosed that NVIDIA's CEO, Jensen Huang, expressed the pressing need for SK Hynix to bring forward their production schedule of the high-bandwidth memory chips, HBM4, by six months. This request underscores the heightened expectations for NVIDIA's next-generation GPUs, which are crucial components for AI and other computational tasks.
The urgency for these components was discussed in detail during recent meetings between SK Hynix and NVIDIA. Originally, SK Hynix planned to commence the supply of HBM4 chips, featuring a 12-layer DRAM stack, in the latter half of 2025, with a more advanced 16-layer version anticipated in 2026. However, the market's voracious appetite has necessitated a re-evaluation of these timelines, with proactive efforts underway to meet these demands ahead of schedule.
SK Hynix's adaptability and collaboration with NVIDIA emphasize the strategic importance of their partnership, particularly as NVIDIA commands over 80% of the AI chip market share. This collaboration highlights a mutual commitment to overcoming current supply constraints and addressing the competitive demands of advancing AI technologies.
Further underlining their forward-looking approach, SK Hynix has already initiated the mass production of 12-layer HBM3E products as of September and is on track to release 16-layer HBM3E versions by early 2025. As the anticipation for HBM5 chips and other advancements continues to build, both companies are evidently steering their innovations to realign with the accelerating pace of technological adoption globally.