东方证券:AI带动功率IC等需求成长 有望持续推动成熟制程需求提升
智通财经网·2026-01-16 01:54

Core Viewpoint - The report from Dongfang Securities highlights that while investors are focused on AI computing power chips driving advanced process foundry demand, there is insufficient attention on the demand for mature process foundries driven by AI. The firm believes that AI will continue to boost the demand for power-related ICs, thereby enhancing the demand for mature process foundries [1]. Group 1: Demand Insights - AI is expected to drive growth in power IC demand, which will continue to enhance the demand for mature processes. The demand increase is attributed to AI server power ICs and the localization trend in mainland China, leading to higher demand for local foundries' BCD/PMIC [3]. - The utilization rate of eight-inch capacity at some foundries has significantly increased since mid-2025, driven by the demand from AI applications [3]. Group 2: Supply Dynamics - TSMC is gradually reducing its eight-inch capacity starting in 2025, with plans for some facilities to cease operations by 2027. Samsung is also initiating eight-inch production cuts in 2025 [4]. - TrendForce forecasts a 0.3% year-on-year reduction in global eight-inch capacity in 2025, with the reduction expected to expand to 2.4% in 2026. This reduction, coupled with rising demand, is projected to increase the average utilization rate of eight-inch capacity to 85-90% in 2026, up from 75-80% in 2025 [4]. Group 3: Domestic Foundry Benefits - The trend of localization in IC manufacturing in mainland China is ongoing, with companies like SMIC reporting growth in market share for various products during the domestic replacement process. This trend is expected to provide continued order growth for domestic foundries [5]. Group 4: Investment Recommendations - The report suggests that the price increase in mature processes and the demand growth driven by AI present investment opportunities. Recommended investment targets include wafer manufacturing companies such as SMIC, Hua Hong Semiconductor, and others, as well as semiconductor equipment companies [6].