全球存储市场 - 2026 年展望:通缩延续,AI 推理需求上升叠加供应受限-Global Memory Market-2026 outlook Disinflation continues with AI inference pick-up and supply constraints
Micron TechnologyMicron Technology(US:MU)2025-12-16 03:30

Summary of J.P. Morgan's Global Memory Market Conference Call Industry Overview - The report focuses on the Global Memory Market, particularly the semiconductors sector, with an emphasis on DRAM and NAND memory products [1][6]. Key Insights Supply and Demand Dynamics - Concerns regarding new capacity in 2027 potentially leading to DRAM oversupply are addressed. However, it is expected that DRAM bit supply growth will lag behind demand growth over the next two years due to higher capacity allocation to High Bandwidth Memory (HBM) and structural demand from AI inference applications, which consume three times more memory than training [3][6]. - The memory market is projected to experience a stronger and longer up-cycle, with diverging pricing trends between B2B (business-to-business) and B2C (business-to-consumer) segments. B2B pricing is expected to remain steady due to AI inference, while B2C pricing may decline due to customer resistance [3][6]. - The 2027 Memory Total Addressable Market (TAM) is forecasted to be approximately US$420 billion, with a potential market cap upside for top memory makers to reach nearly US$1.5 trillion [3][6]. Capital Expenditure and Capacity - Current capital expenditure (Capex) initiatives by memory suppliers are not yet sufficient to close the supply-demand gap. The expected growth in memory wafer fabrication equipment (WFE) is projected to outpace gross Capex spending growth [3][6]. - The report anticipates that DRAM and NAND capital intensity will remain below the average of the last five years, with DRAM at sub-30% and NAND at sub-20% [3][6]. AI and HBM Demand - The debate between GPU and ASIC technologies is expected to drive HBM demand, particularly with advancements in AI applications. The introduction of next-generation TPUs by companies like Google is likely to further tighten HBM supply-demand dynamics [3][6]. - AI inference is also projected to positively impact enterprise SSD TAM, with expectations of reaching mid-400EB by 2026 [3][6]. Investment Recommendations - J.P. Morgan recommends focusing on memory stocks, anticipating earnings per share (EPS) upgrades driven by ASP hikes. The near-term pecking order for large-cap Asian memory stocks is Samsung Electronics (SEC) and SK Hynix (SKH) [4][6]. - The report suggests maintaining an overweight (OW) rating on Micron Technology (MU) due to its rising exposure to AI, while being neutral on Nanya Technology (NYT) [4][6]. Market Trends and Projections - The memory market cap is nearing US$1 trillion, with expectations of continued upward trends in memory ASP due to CSP-driven demand [6][62]. - The report raises memory TAM forecasts by 37% to 44% for FY26-27, driven by tightness in conventional DRAM and HBM supply-demand [62][63]. - DRAM revenue is projected to grow significantly, with ASP expected to rise by 57% in FY26, followed by a modest 1% growth in FY27 [66]. Conclusion - The memory market is poised for a significant up-cycle, driven by AI demand and supply constraints. Investors are encouraged to focus on the longevity of this cycle and the potential for substantial market cap growth among leading memory manufacturers [6][62].

全球存储市场 - 2026 年展望:通缩延续,AI 推理需求上升叠加供应受限-Global Memory Market-2026 outlook Disinflation continues with AI inference pick-up and supply constraints - Reportify