半导体ETF（159813）业绩比较基准为国证半导体芯片指数收益率，管理人为鹏华基金管理有限公司， 基金经理为罗英宇，成立（2020-04-17）以来回报为86.00%，近一个月回报为3.62%。 声明：市场有风险，投资需谨慎。本文基于第三方数据库自动发布，不代表新浪财经观点，任何在本文 出现的信息均只作为参考，不构成个人投资建议。如有出入请以实际公告为准。如有疑问，请联系 biz@staff.sina.com.cn。 2月24日，半导体ETF（159813）开盘涨1.53%，报1.260元。半导体ETF（159813）重仓股方面，中芯国 际开盘涨0.94%，海光信息涨3.00%，寒武纪涨1.39%，北方华创涨1.17%，兆易创新涨2.85%，中微公司 涨1.26%，澜起科技涨3.27%，豪威集团涨1.02%，拓荆科技涨0.92%，长电科技涨1.67%。 来源：新浪基金∞工作室 ...

来源：新浪基金∞工作室 科创芯片ETF汇添富（588750）业绩比较基准为上证科创板芯片指数收益率，管理人为汇添富基金管理 股份有限公司，基金经理为孙浩，成立（2024-12-18）以来回报为78.38%，近一个月回报为5.60%。 声明：市场有风险，投资需谨慎。本文基于第三方数据库自动发布，不代表新浪财经观点，任何在本文 出现的信息均只作为参考，不构成个人投资建议。如有出入请以实际公告为准。如有疑问，请联系 biz@staff.sina.com.cn。 2月24日，科创芯片ETF汇添富（588750）开盘涨0.95%，报1.800元。科创芯片ETF汇添富（588750）重 仓股方面，中芯国际开盘涨0.94%，海光信息涨3.00%，寒武纪涨1.39%，澜起科技涨3.27%，中微公司 涨1.26%，拓荆科技涨0.92%，芯原股份涨3.45%，华虹公司涨0.16%，沪硅产业涨1.12%，东芯股份涨 0.91%。 ...

来源：新浪基金∞工作室 2月24日，芯片ETF天弘（159310）开盘涨1.12%，报2.342元。芯片ETF天弘（159310）重仓股方面，寒 武纪开盘涨1.39%，中芯国际涨0.94%，海光信息涨3.00%，北方华创涨1.17%，兆易创新涨2.85%，澜起 科技涨3.27%，中微公司涨1.26%，豪威集团涨1.02%，紫光国微涨1.02%，拓荆科技涨0.92%。 芯片ETF天弘（159310）业绩比较基准为中证芯片产业指数收益率，管理人为天弘基金管理有限公司， 基金经理为洪明华，成立（2024-04-18）以来回报为131.72%，近一个月回报为3.46%。 声明：市场有风险，投资需谨慎。本文基于第三方数据库自动发布，不代表新浪财经观点，任何在本文 出现的信息均只作为参考，不构成个人投资建议。如有出入请以实际公告为准。如有疑问，请联系 biz@staff.sina.com.cn。 ...

继续坚定看好国产AI&海外发散： 继续坚定看好国产AI&海外发散： 1.国产算力：H和#海光信息，H主要看#华正新材，禾盛新材，IDC：#金开新能 2.存储： 雅克科技、帝科股份、开普云，闪迪美光7709 3.应用：#看好优刻得、首都在线、青云科技、中控技术、合合信息、真爱美家、德才股份（联系岚琪） 4.上游： 宏和科技、东材科技、延江股份 5.液冷： 飞龙股份、华光新材、江南新材 6.OCS：君逸数码（OCS上游透镜阵 ...

来源：新浪基金∞工作室 芯片ETF（159995）业绩比较基准为国证半导体芯片指数收益率，管理人为华夏基金管理有限公司，基 金经理为赵宗庭，成立（2020-01-20）以来回报为93.54%，近一个月回报为3.60%。 声明：市场有风险，投资需谨慎。本文基于第三方数据库自动发布，不代表新浪财经观点，任何在本文 出现的信息均只作为参考，不构成个人投资建议。如有出入请以实际公告为准。如有疑问，请联系 biz@staff.sina.com.cn。 2月24日，芯片ETF（159995）开盘涨1.86%，报1.972元。芯片ETF（159995）重仓股方面，中芯国际开 盘涨0.94%，海光信息涨3.00%，寒武纪涨1.39%，北方华创涨1.17%，兆易创新涨2.85%，中微公司涨 1.26%，澜起科技涨3.27%，豪威集团涨1.02%，拓荆科技涨0.92%，长电科技涨1.67%。 ...

Investment Rating - The report maintains an "Overweight" rating for the computer sector [4]. Core Insights - The report highlights the concentrated release of domestic large models, including GLM-5, Doubao Model 2.0, Seedance 2.0, and MiniMax M2.5, alongside rapid iterations of overseas models like GPT-5.3-Codex-Spark and Gemini 3 Deep Think. It emphasizes the release of policies by the National Development and Reform Commission to accelerate the application of AI in the bidding and tendering sector [4][5]. - The domestic large models are advancing in foundational models and multimodal capabilities, with GLM-5 achieving a global ranking of fourth and first in open-source performance. The report notes significant improvements in efficiency and task execution capabilities in the latest models [4][5]. - The report outlines the competitive landscape among tech giants, with OpenAI and Google making substantial advancements in their respective models, enhancing performance in various benchmarks and applications [4][5]. Summary by Sections Domestic Large Models - The report discusses the release of several domestic large models, focusing on their capabilities in agent-based applications and multimodal understanding. GLM-5 has been recognized for its open-source performance, while Doubao Model 2.0 and Seedance 2.0 have been upgraded to meet practical application needs [4][5]. Overseas Large Models - The report details the rapid iteration of overseas models, particularly OpenAI's GPT-5.3-Codex-Spark, which is designed for real-time programming and achieves near-instantaneous response speeds. Google’s Gemini 3 Deep Think has also shown significant performance improvements across various challenging benchmarks [4][5]. Policy Developments - The report highlights the government's initiative to empower the entire bidding and tendering process with AI, aiming for comprehensive application coverage by the end of 2026 in select provinces and nationwide by 2027. This initiative is expected to enhance the regulatory framework and market development [4][5].

Core Viewpoint - The 2026 CCTV Spring Festival Gala serves as a cultural feast and a showcase of the global chip industry, highlighting the advancements in China's chip technology from "usable" to "highly usable" [1][29] Group 1: Robotics Showcase - Four humanoid robots, MagicBot Z1, Galbot G1, Unitree H1, and "Little Bumi," showcased at the gala, demonstrating the capabilities of eight chip companies that support these "tech performers" [3][29] - MagicBot Z1 features 100% domestically produced core chips, with Allwinner Technology providing the main SoC chip for overall control and motion planning [4][5] - Galbot G1 relies on NVIDIA's AGX Orin 64GB chip, achieving 275 TOPS of computing power for AI interactions [6][7] - Unitree H1, the most commercially successful model, utilizes multiple chips including Intel's Core i7 for high-performance motion control and NVIDIA's Jetson Orin NX for edge AI inference [8][9][10][11] - "Little Bumi" targets the consumer market with a price under 10,000 yuan, using Rockchip's RK3588 SoC for approximately 6 TOPS of AI computing power [12][13][14] Group 2: Broadcasting Technology - The gala achieved 8K ultra-high-definition wireless broadcasting, supported by four chip companies that built the computational foundation [15][29] - Huawei's Hi3536DV400 chip enabled real-time compression of 8K video, while the Kunpeng 920S chip supported over 1 billion concurrent live requests [16][17] - AMD's EPYC 9004 series chips provided rendering power for real-time graphics during the event [18] Group 3: Communication Infrastructure - The seamless connection between the main venue and four sub-venues relied on three chip companies for communication technology [19][29] - Huawei's Tian Gang 02 base station chip facilitated low-latency transmission of 8K signals, marking the large-scale application of 5G-A technology [19] - ZTE and China Star Network contributed to the 5G-A network and satellite communication, ensuring stable signal transmission in remote areas [20][21] Group 4: Stage Control and Perception - The precise operation of stage lighting, robotic arms, and AR effects depended on three chip companies providing MCU, power chips, and perception chips [22][29] - GigaDevice's GD32H7 series MCU enabled rapid light switching and robotic movement control [23] - STMicroelectronics' IGBT power chips ensured stable operation of high-power stage equipment [24] - Orbbec provided depth perception chips for the Unitree H1 robot, enhancing spatial awareness and obstacle avoidance [25] Group 5: End-User Experience - The viewing experience for audiences was enhanced by three chip companies ensuring high-quality visuals and security [26][29] - Hisense's X9 image quality chip optimized 8K visuals for the gala [27] - Unisoc's financial-grade security chips safeguarded interactive data during the event [28] - BDU's BD960 positioning chip ensured synchronization of devices across different locations [29]

Core Viewpoint - The release of Qwen 3.5 has sparked significant interest in the global AI industry, showcasing advancements in model efficiency and cost-effectiveness [1] Group 1: Model Performance and Features - Qwen 3.5 has a total of 397 billion parameters, with only 17 billion activated, outperforming the previous trillion-parameter model Qwen 3-Max [1] - The deployment memory usage has decreased by 60%, while the maximum inference throughput has increased by 19 times [1] - The innovative underlying model architecture allows Qwen 3-Plus to offer competitive performance at a low cost of 0.8 yuan per million tokens on Alibaba Cloud's Bailian API, achieving high performance comparable to Gemini 3 at less than 5% of the price [1] Group 2: Industry Adoption and Compatibility - Major international hardware manufacturers such as NVIDIA, AMD, and Apple have adapted their development frameworks and chips for Qwen 3.5 [1] - Domestic GPU and platform providers, including Huawei Ascend, Moore Threads, and others, have also announced Day 0 compatibility with the new model [1] - Various platforms, both domestic and international, including National Supercomputing Internet Platform, Shuguang Cloud, and others, have quickly integrated the Qwen 3.5 model, offering API services or experience interfaces [1]

Core Insights - Despite macroeconomic challenges, Lenovo Venture Capital has maintained a clear strategic focus, achieving significant investment milestones in 2025 [3] - The company invested in over 40 new enterprises and facilitated nearly 100 portfolio companies in securing new rounds of financing, with total investment transactions exceeding 60, 80% of which were directed towards sectors closely related to the "silicon-based ecosystem" [3] - The total number of portfolio companies has surpassed 300, with 25 successfully going public, including leading firms like CATL, Cambricon, and Haiguang Information, each valued over 500 billion [3] Industry Outlook - The industry faces a gap between technological ideals and commercial realities, particularly in fields like embodied intelligence, where establishing scalable business models is a key challenge [3] - High-quality vertical data is expected to become a core barrier for industrial AI, with China's complete industrial chain and rich scenarios providing unique advantages for incubating breakthrough technologies [3] Future Investment Strategy - Looking ahead to 2026, investments are expected to become more rational, with "silicon-based intelligence" remaining a long-term focus [4] - The company plans to deepen its investment in three main areas: the evolution of AI towards "system intelligence," focusing on intelligent agents and AI-native hardware; the commercialization of machine intelligence, leveraging the benefits of autonomous driving technology; and forward-looking investments in disruptive computing architectures such as integrated storage and computing and RISC-V [4]

Core Viewpoint - The article discusses the current funding logic in the market, highlighting the significant movements of public funds in the technology sector, particularly in the semiconductor and biopharmaceutical industries, as they adjust their portfolios based on performance and valuation metrics [3][8]. Group 1: Public Fund Movements - Public funds have shown a notable shift in their holdings, particularly in the STAR Market, with the STAR 50 Index rising by 12.1% this year [3][4]. - The total market capitalization of STAR Market companies reached 10.4 trillion yuan, with the technology sector dominating, accounting for 62.1% of the total market cap [5][6]. - The semiconductor industry remains the core focus for fund allocation, with 12 companies in the sector having a market capitalization exceeding 10 billion yuan [9][11]. Group 2: Semiconductor Sector Insights - The market's pricing anchor for the semiconductor sector has shifted from "valuation expansion" to "performance realization," emphasizing the importance of actual earnings [13][14]. - Key drivers for future growth in the semiconductor sector include strong order backlogs, profit growth through acquisitions and expansions, and sustained price increases in advanced processes [13][14]. - Public funds have significantly increased their holdings in semiconductor materials, chip design, and equipment, with companies like ShenGong Co. seeing an 11% increase in fund holdings [15][21]. Group 3: Biopharmaceutical Sector Insights - The biopharmaceutical sector is a critical area for public funds, with major holdings in companies like BeiGene and United Imaging Healthcare, although the sector has faced a reduction in holdings for several key companies [24][26]. - The article notes that innovative drug companies are currently under pressure, with significant reductions in holdings observed in companies like BaiLi TianHeng and RongChang Biopharma [26][28]. - Despite the challenges, companies with strong earnings potential and innovative drug pipelines are still attracting interest from public funds, indicating a selective investment approach [35][40].