SCS(SH:601555) 智通财经网·2025-12-31 03:41Group 1: Cloud Computing Power - The AI computing power industry chain is expected to see significant performance growth by 2026, with domestic GPUs entering a period of performance realization, emphasizing the importance of industry chain collaboration [1][2] - Domestic computing chip leaders are anticipated to enter a performance realization phase, benefiting from capacity release due to advanced process expansion [2] - AIASIC service providers are expected to play a crucial role in the supply chain as domestic computing enters a super node era, testing both GPU manufacturers' capabilities and the localization level of Switch chips [2] Group 2: Edge Computing Power - The hybrid architecture of edge and cloud computing is expected to empower AI applications, with edge AI benefiting from innovations and the release of 3D DRAM [3] - The first half of 2026 may see consumer upstream companies facing pressure from rising storage prices, but the second half is expected to present opportunities from new wearable AIoT products [3] - 2026 is projected to be a pivotal year for NPU deployment, with companies like Rockchip and Amlogic being relevant players [3] Group 3: 3D DRAM - 2026 is anticipated to be a year of significant demand for 3D DRAM, driven by the deployment of AI hardware across various fields [4] - The demand for high bandwidth and low-cost 3D DRAM is expected to rise, particularly in robotics, AIoT, and automotive sectors [4] - The introduction of multiple NPU tape-outs will provide diverse application scenarios for 3D DRAM, with mobile and cloud inference also gradually adopting this technology [4] Group 4: AI Models and Ecosystem - By 2026, cloud models are expected to enhance their complex planning capabilities through data quality and post-training optimization, while edge models will leverage cloud capabilities through distillation [5] - The ecosystem will see terminal manufacturers controlling the OS and system-level entry points, while super apps will create closed loops for application agents [5] Group 5: AI Terminals - 2026 marks the beginning of a new era for AI terminal innovations, with major companies like Meta, Apple, Google, and OpenAI launching new products [6] - New terminal forms, such as smart glasses and AI pins, are expected to emerge, driven by model iterations and application scenario development [6] - Key components such as SoC, batteries, and communication technologies will be crucial for the development of these new terminals [6] Group 6: Longxin Chain - Longxin's focus on 3D technology is expected to drive continuous capacity expansion, benefiting its supply chain partners [7] - The expansion of Longxin's production capacity will create opportunities for equipment and testing companies [7] Group 7: Wafer Foundry - The advanced logic foundry capacity is expected to double, maintaining a favorable outlook for the foundry industry [8] - The supply of advanced processes, especially at 7nm and below, is currently insufficient, with significant expansion anticipated in 2026 [9] Group 8: PCB Industry - The demand for high-speed signal integrity and low dielectric performance in AI servers is driving a comprehensive upgrade cycle in PCB materials [10] - M9 CCL is becoming a key substrate for AI servers and high-speed communication systems, expected to significantly increase the value of the PCB industry [10] Group 9: Optical-Copper Interconnection - The growth of commercial GPUs in 2026 will lead to large-scale deployment of CSPASIC, with a surge in demand for optical and copper interconnections [11] - The dual resonance of copper and optical technologies is expected to drive both volume and price increases in interconnection demand [11] Group 10: Server Power Supply - The rising power density in AI data centers is making HVDC power supply architecture a core focus, with significant upgrades across the supply chain [12] - The technology upgrades in server power supplies are expected to enhance the value of PCB components significantly [12] Group 11: Recommended Companies - Companies to watch in cloud computing power include Cambricon, Haiguang Information, and Zhaoxin [13] - In edge computing power, relevant companies include Amlogic, Rockchip, and Hengxuan Technology [13] - For storage, companies like GigaDevice and Longsys are noteworthy [13] - In AI terminals, companies such as Luxshare Precision and Sunny Optical are recommended [13] - For wafer foundry, companies like SMIC and Hua Hong Semiconductor are relevant [13] - In the PCB industry, companies such as Shennan Circuits and Unimicron Technology are highlighted [13] - For optical-copper interconnection, companies like Long光华芯 and Huafeng Technology are suggested [13] - In HVDC, companies like Eurotech and Megmeet are recommended [13]