行 业 华福证券 电子 2026 年 02 月 02 日 行 业 投 资 策 略 截止到 2025Q3，三星 25Q1-Q3 全球智能手机出货量达到 6060 万 台，市场份额达 19%，保持全球第一；苹果 25Q3 出货量为 5650 万台， 同比上升 4%，市场份额位列第二；其次为小米/传音/vivo。据环球网 援引IDC 报告显示 2025年全球智能手机出货量预计将达到 12.4 亿部， 2024-2029 年 CAGR=1.5%，呈现温和增长。展望 2026 年，我们持续看 好手机结构化升级带来的创新需求，包括潜望式镜头成为安卓/苹果高 端手机标配带来渗透率持续提升。 研 究 电子行业 2026 年度策略报告：AI 与自主可控共 振 投资要点： 预计 2026 年手机行业温和复苏，看好结构创新带来增量需求 证 券 研 随着国产算力需求的不断增加，先进制程的产能需求也不断提升。 中芯国际、华虹公司、长鑫存储等企业的先进制程制造产能预计不断 增加。成熟制程工艺各个制造商各有千秋，特色工艺投资机会显现。 芯联集成、晶合集成、赛微电子各自特色领域优势显著。内生增长以 及外延并购提升半导体设备覆盖率。半导体制造 ...

Core Insights - In 2025, China's primary market witnessed a unique "queueing up" phenomenon, with numerous hard tech companies attracting multiple investors in single rounds of financing, indicating a shift in investment strategies towards collective participation rather than individual dominance [1][13] Group 1: Investment Trends - A total of 13 projects in 2025 had over 15 investors in single financing rounds, covering strategic sectors such as AI chips, commercial aerospace, robotics, semiconductors, and biomedicine [1][2] - Notable companies include Mu Xi Co., Ltd. (C round, nearly 50 investors), Qingwei Intelligent (C round, 24 investors), and the Hainan Commercial Aerospace Innovation Center (angel round, nearly 30 investors) [2][3] Group 2: Mu Xi Co., Ltd. Case Study - Mu Xi Co., Ltd. completed a record financing round in February 2025, raising over 7.2 billion yuan with nearly 50 investment institutions participating, marking it as the most participated equity transaction in the primary market for the year [3][4] - The investor lineup included state-owned institutions, market-oriented VC/PE firms, and industrial capital, showcasing a diverse range of market participants [4][5] - The motivation behind the investment was driven by the urgency for domestic AI computing power alternatives due to restrictions on high-end chips from Nvidia, positioning Mu Xi as a key player in the domestic GPU market [5][6] Group 3: Hainan Commercial Aerospace Innovation Center - The Hainan Commercial Aerospace Innovation Center was established in July 2025 with nearly 30 industry chain companies participating, marking a precedent for competitors in the aerospace sector to collaborate [7][9] - The motivations for participation included leveraging Hainan's favorable launch conditions and tax policies, establishing technology standard alliances, and securing future launch window resources [8][9] Group 4: Qingwei Intelligent Financing - Qingwei Intelligent completed a 2 billion yuan C round financing in December 2025, with 24 institutions participating, setting a record for the largest single financing in the domestic AI chip sector in recent years [10][11] - The investment was characterized by a leading role from state-owned platforms and collaboration with market-oriented institutions, emphasizing the strategic importance of the company’s reconfigurable computing architecture [11][12] Group 5: Market Dynamics and Challenges - The phenomenon of multiple investors in single financing rounds reflects structural changes in the primary market, driven by a scarcity of high-quality projects and a surplus of capital [13][14] - The need for risk-sharing mechanisms in hard tech projects, which require significant investment and have long development cycles, has led to a trend of group investments to mitigate individual risks [15][16] - The influence of state-owned capital and the trend of pre-IPO speculation, as seen in Mu Xi's dramatic stock performance, highlight the potential for valuation bubbles and exit pressures in the primary market [17]

Group 1 - Shanghai Jiyue Xingchen Intelligent Technology Co., Ltd. launched the new generation open-source Agent base model Step3.5Flash, designed for real-time Agent workflow scenarios, balancing inference speed, intelligence level, and cost [1] - Step3.5Flash achieves a maximum inference speed of 350 tokens per second for single request code tasks, providing a "faster, stronger, and more stable" option for Agent base models [1] - The model utilizes a sparse MoE architecture, activating approximately 11 billion parameters per token out of a total of 196 billion parameters, significantly enhancing inference efficiency while maintaining model capability [1] Group 2 - Nearly 10 chip and infrastructure manufacturers, including Huawei Ascend and Alibaba Pingtouge, have completed adaptations for Step3.5Flash, enhancing model adaptability and computing efficiency through collaborative innovation [1] - The establishment of the "MoXin Ecological Innovation Alliance" in July 2025 aims to break down technical barriers between chips, models, and platforms, optimizing performance and accelerating the application of large models across various industries [2] - The industry recognizes that deep collaboration between models and computing power will be a crucial path for the large-scale application of inference models [2]

Group 1 - The core point of the news is that Muxi Integrated Circuit (Shanghai) Co., Ltd. experienced a stock decline of 5.02%, with a current share price of 530.08 yuan and a total market capitalization of 2120.85 billion yuan [1] - Muxi's main business involves the research, design, and sales of full-stack GPU products for artificial intelligence training and inference, general computing, and graphics rendering, with revenue composition primarily from training and inference GPU boards at 97.55% [1] - The company was established on September 14, 2020, and is located in the Pudong New Area of Shanghai, with its listing date set for December 17, 2025 [1] Group 2 - Huafu Fund has a significant holding in Muxi, with its Huafu Digital Economy Mixed A Fund (018358) holding 3.44% of its net value in Muxi shares, ranking as the ninth largest holding [2] - The fund has a total scale of 35.2582 million yuan and has achieved a year-to-date return of 16.77%, ranking 398 out of 9000 in its category [2] - The fund manager, Huang Xinglin, has been in position for 2 years and 229 days, with the best fund return during his tenure being 90.93% [3]

Core Viewpoint - The release of the new open-source Agent base model Step3.5Flash by Jumpsky is aimed at real-time Agent workflow scenarios, utilizing a sparse MoE architecture with a total parameter count of 196 billion, designed to balance inference speed and cost [1] Group 1: Model Specifications - Step3.5Flash achieves an inference speed of up to 350 tokens per second for single request code tasks [1] - The model activates approximately 11 billion parameters per token [1] Group 2: Industry Collaboration - Multiple chip manufacturers, including Huawei Ascend, Muxi Co., Biran Technology, Suiruan Technology, TianShu Zhixin, and Alibaba Pingtouge, have completed adaptations for the model [1] - Jumpsky previously initiated the "MoCore Ecological Innovation Alliance" in July 2025 with several chip and infrastructure companies to enhance computing efficiency and promote the application of large models [1] Group 3: Strategic Implications - The release of Step3.5Flash is viewed as a further practice in the collaboration between model and computing power [1]

Core Viewpoint - The release of the new open-source Agent base model Step3.5Flash by Jumpsky is aimed at real-time Agent workflow scenarios, utilizing a sparse MoE architecture with a total parameter count of 196 billion, balancing inference speed and cost [1] Group 1: Model Specifications - Step3.5Flash achieves an inference speed of up to 350 tokens per second for single request code tasks [1] - The model activates approximately 11 billion parameters per token [1] Group 2: Industry Collaboration - Multiple chip manufacturers, including Huawei Ascend, Muxi Co., Biran Technology, and Suiruan Technology, have completed adaptations for the Step3.5Flash model [1] - Jumpsky previously initiated the "MoCore Ecological Innovation Alliance" in July 2025 with several chip and infrastructure manufacturers to enhance computing efficiency and promote the application of large models [1] - The model release is seen as a further practice in the direction of model and computing power collaboration [1]

Core Insights - The release of the new open-source Agent base model Step 3.5 Flash aims to enhance real-time Agent workflow scenarios by balancing inference speed, intelligence level, and cost [1] - Step 3.5 Flash achieves a maximum inference speed of 350 tokens per second for single-request code tasks, positioning itself as a preferred choice for Agent applications [1] Group 1: Model Features - Step 3.5 Flash utilizes a sparse MoE architecture, activating approximately 11 billion parameters per token out of a total of 196 billion parameters, significantly improving inference efficiency while maintaining model capability [1] - The model is designed to provide a more efficient and affordable foundational model option for Agent applications [1] Group 2: Industry Collaboration - In July 2025, the company initiated the "MoCore Ecological Innovation Alliance" with nearly 10 chip and infrastructure manufacturers to eliminate technical barriers between chips, models, and platforms [2] - The alliance aims to enhance computing power utilization efficiency through joint optimization, accelerating the application of large models across various industry scenarios [2] - The industry recognizes that deep collaboration between models and computing power will be a crucial pathway for the large-scale application of inference models [2]

Core Insights - The core focus of the news is the launch of the new open-source Agent base model, Step 3.5 Flash, by Jiyue Xingchen, which is designed for real-time Agent workflow scenarios, balancing inference speed, intelligence level, and cost efficiency [1][3] Model Performance - Step 3.5 Flash achieves a maximum inference speed of 350 tokens per second for single request code tasks [1][3] - The model utilizes a sparse MoE architecture, activating approximately 11 billion parameters per token out of a total of 196 billion parameters, significantly enhancing inference efficiency while maintaining model capability [1][3] Industry Collaboration - Jiyue Xingchen has collaborated with several chip manufacturers, including Huawei Ascend, Muxi Co., Birun Technology, Suiyuan Technology, Tianshu Zhixin, and Alibaba Pingtouge, to adapt Step 3.5 Flash, enhancing model adaptability and computational efficiency [1][3] - The collaboration aims to lower inference costs and reduce the overall barriers for enterprises and developers in applying large models, accelerating the practical application of large models in various scenarios [1][3] Ecosystem Development - In July 2025, Jiyue Xingchen initiated the "MoCore Ecological Innovation Alliance" with nearly 10 chip and infrastructure manufacturers to bridge the technical barriers between chips, models, and platforms [2][4] - The alliance aims to optimize and enhance computational efficiency, facilitating the large-scale application of models across various industry scenarios [2][4] - The industry recognizes that deep collaboration between models and computational resources will be a crucial path for promoting the large-scale application of inference models [2][4]

阶跃星辰发布新一代开源Agent基座模型Step 3.5 Flash。该模型面向实时Agent工作流场景，最高推理速 度可达每秒350个token。据悉，Step 3.5 Flash采用稀疏MoE架构，每个token仅激活约110亿个参数（总 计960亿参数）。包括华为昇腾、 沐曦股份、壁仞科技、燧原科技、天数智芯、阿里平头哥在内的多家 芯片厂商，已完成适配。 ...

Group 1 - In 2025, the new "National Nine Articles" and the "1+6" reform of the Sci-Tech Innovation Board drove a moderate development of the Shanghai and Shenzhen new stock market, with a total of 90 new stocks listed, an increase of 13 from 2024 [1][7] - The fundraising scale significantly increased, with total funds raised reaching 124.24 billion yuan, a year-on-year growth of 98.25%, and large IPOs in the second half contributed 72% of the total fundraising [1][18] - The average fundraising amount per new stock was 13.80 million yuan, up 69.62% year-on-year, with the Sci-Tech Innovation Board leading at an average of 20.03 million yuan [1][24] Group 2 - In the primary market, the offline subscription yield remained stable compared to 2024, but accounts with a scale of 5 billion yuan and above saw a significant increase in yield [2][31] - The average first-day closing price increase was 227.90%, maintaining a high level, with no stocks experiencing a loss on the first day [2][41] - The Sci-Tech Innovation Board had the highest average first-day closing price increase, while industries such as non-ferrous metals and social services performed notably well [2][50] Group 3 - In the secondary market, timing remained a key factor for improving investment returns and success rates, with new stocks listed during market sentiment lows or upward cycles performing better [2][8] - The average investment return in the secondary market was 14.53%, a slight decrease from 2024, but the investment success rate increased to 33.33% [2][8] - The ChiNext Board outperformed in both investment returns and success rates, with industries like electronics and social services leading [2][8]