智通财经网·2026-01-16 02:17Group 1 - The core viewpoint of the article is that semiconductor revenue from AI applications is expected to grow significantly, from $220 billion in 2025 to over $550 billion by 2028, according to RBC Capital Markets [1] - Current market supply is tight, with enterprise order delivery cycles extended to 18 months, which clarifies the industry's outlook [1] - Infrastructure bottlenecks may delay some projects, but this could smooth out the spending cycle in the AI sector rather than being a negative factor [1] Group 2 - RBC Capital Markets has initiated coverage on several semiconductor companies, giving them an "outperform" rating, including Nvidia, Micron Technology, Marvell Technology, Arm, Astera Labs, ASML, Applied Materials, Lam Research, and Lattice Semiconductor [1] - For companies like Broadcom, AMD, Intel, KLA, SanDisk, Qualcomm, Skyworks, and Silicon Labs, RBC has assigned a "market perform" rating [2] Group 3 - High Bandwidth Memory (HBM) demand is expected to be a core growth driver, potentially reducing the cyclical volatility of the memory market [3] - AI workloads are shifting towards reinforcement learning and distributed inference, which require high memory performance [3] - The upcoming HBM4 iteration is anticipated to be beneficial, with average prices expected to increase by 30-50% [3] - The demand for high-capacity server memory (DIMM) and solid-state drives (NAND eSSD) is also driven by the explosion of generative AI [3] - Despite high memory prices potentially impacting demand in the PC and smartphone markets, the memory industry is expected to remain in a supply-demand imbalance until 2027 [3] - Capital expenditures in the wafer fabrication equipment (WFE) sector are projected to maintain strong growth over the next two years [3] - Technological trends such as backside power delivery, advanced packaging, and three-dimensional structures are expected to drive at least high single-digit growth in the wafer fabrication equipment market over the next two years [3]