Nvidia(US:NVDA) Hua Er Jie Jian Wen·2025-11-20 12:58Core Insights - SK Hynix is significantly expanding its advanced memory chip production capacity, betting on market opportunities arising from the shift in AI applications from training to inference [1][3] Group 1: Production Capacity Expansion - SK Hynix plans to increase its monthly production capacity of sixth-generation 10nm DRAM (1c DRAM) from approximately 20,000 300mm wafers to between 160,000 and 190,000 wafers, representing an increase of 8 to 9 times, which will account for over one-third of its total DRAM capacity [1][2] - The company aims to enhance its production efficiency by focusing on 1c DRAM, which has a yield rate exceeding 80% and is primarily used for manufacturing the latest general-purpose DRAM products like DDR5, LPDDR, and GDDR7 [2] Group 2: Strategic Shift in AI Applications - The strategic adjustment reflects a shift in focus from high-bandwidth memory (HBM) to general-purpose DRAM, driven by the increasing demand for cost-effective memory solutions in AI inference applications [1][3] - Major tech companies, including NVIDIA, Google, OpenAI, and Amazon Web Services, are developing custom AI accelerators that integrate large amounts of general-purpose DRAM, indicating a trend towards more energy-efficient and economical memory solutions [4] Group 3: Pricing and Profitability - SK Hynix successfully negotiated a price increase of over 50% for HBM4 to above $500 per unit, which is supported by technological advancements that enhance data transmission capabilities [6][7] - The company is expected to achieve a significant increase in operating profit, with projections indicating that operating profit could exceed 70 trillion KRW, marking a historical high, driven by both HBM and general DRAM sales [8]