Core Viewpoint - The recent supply disruption of CMP slurry, specifically DST slurry produced by AGC in Taiwan, has accelerated the pace of domestic alternatives in the semiconductor industry, highlighting the urgency for local manufacturers to innovate and adapt [1][2]. Group 1: Importance of CMP Slurry - CMP slurry is essential for the chemical mechanical polishing process in semiconductor manufacturing, akin to smoothing and leveling surfaces in construction [2]. - The slurry consists of ultra-fine abrasives, chemical agents, and water, and is crucial for achieving the required surface smoothness in logic and memory chip production [2]. Group 2: Market Dynamics - The CMP slurry market is predominantly controlled by American and Japanese companies, with Cabot Microelectronics holding 33% of the global market share, while Japanese firms account for an additional 23% [3]. - Domestic manufacturers currently hold a market share of 25-30%, which is even lower in high-end processes (e.g., below 28nm) [3]. - The global CMP slurry market was valued at $1.89 billion in 2021 and is projected to exceed $2.6 billion by 2026, with a compound annual growth rate (CAGR) of approximately 6% [3]. - In China, the CMP slurry market is expected to grow from approximately 2.3 billion yuan in 2023 to 4.6 billion yuan by 2028, reflecting a CAGR of 15% [3]. Group 3: Demand Surge Factors - The demand for CMP slurry is increasing due to the growing complexity of high-end processes, with the number of CMP steps for 14nm logic chips doubling, requiring more types of slurry [3]. - The rise of 3D NAND memory chips, with increasing stacking layers, necessitates more intricate polishing processes, leading to a significant increase in CMP slurry usage [3]. Group 4: Domestic Players - Anji Technology has been a leader in CMP slurry since 2006, achieving a market share of over 5% globally and increasing its share in China from 20.9% in 2020 to 30.8% in 2021 [4]. - Sanhe New Materials focuses on diamond polishing liquids for hard materials and is gradually increasing its market share by supplying to companies like SMIC [4]. - Dinglong Co. is a key player in the domestic polishing pad market, offering a comprehensive solution that includes polishing liquids, cleaning liquids, and polishing pads, enhancing its competitive edge [5].

Core Viewpoint - Synopsys, Inc. announced the immediate lifting of U.S. export restrictions to China, allowing the company to resume supply and customer support for recently restricted products in the Chinese market [1]. Group 1 - The U.S. Department of Commerce's Bureau of Industry and Security notified Synopsys on July 2 that the export control measures imposed on May 29, 2025, have been revoked [1]. - The lifting of these restrictions is expected to positively impact Synopsys' operations and market presence in China [1].

Core Viewpoint - The U.S. government has lifted export restrictions on chip design software for China, allowing companies like Siemens to restore full access to their technology for Chinese customers [1][3]. Group 1: Export Restrictions and Compliance - The U.S. Commerce Department's Bureau of Industry and Security had previously instructed leading EDA providers to halt shipments to Chinese customers, which included semiconductor design software and related chemicals [4]. - Siemens AG has confirmed that it received a notice from the U.S. government regarding the lifting of these export restrictions, enabling the company to fully resume its operations in China [3][1]. - Cadence and Synopsys, other major EDA software manufacturers, have also been affected by these restrictions, with Synopsys suspending its financial forecasts and operations in China until further clarity is provided [2].

Core Viewpoint - In the first half of 2025, 7 semiconductor companies that are still in the red managed to go public, indicating a shift in the capital market's attitude towards unprofitable firms, particularly in the semiconductor sector [1]. Group 1: Changes in IPO Regulations - The traditional A-share IPO process required companies to meet profit thresholds, but the introduction of the Sci-Tech Innovation Board in 2019 allowed unprofitable companies with strong technology to list [3]. - On February 17, 2023, the China Securities Regulatory Commission (CSRC) approved a third set of financial standards for the ChiNext board, allowing unprofitable companies with a market value of at least 5 billion yuan and revenue of at least 300 million yuan to go public [3]. - The first unprofitable company to be accepted for listing on the ChiNext was Dapu Microelectronics on June 27, 2025, marking the implementation of the new standards [3]. Group 2: Overview of the 7 Unprofitable Companies - The 7 companies, despite their losses, are engaged in critical sectors within the semiconductor industry [4]. - **Mole Thread and Muxi Co., Ltd.** are focused on the GPU market, facing high costs in architecture innovation to compete with dominant players like NVIDIA [5]. - **Dapu Microelectronics** specializes in storage control chips, essential for the smart storage systems, and aims to secure funding through its IPO to support ongoing high R&D costs [6]. - **Shiyatech** is developing silicon-based OLED display chips for AR/VR devices, requiring significant investment in R&D and production capabilities [7]. - **Zhaoxin Integrated** is tackling the CPU market, aiming to create a complete domestic computing platform despite facing significant challenges in ecosystem adaptation and performance optimization [8]. - **Shanghai Super Silicon** focuses on producing high-purity silicon wafers, a foundational element in chip manufacturing, requiring substantial upfront investment [10]. - **Angrui Micro** is dedicated to RF front-end chips critical for mobile signal quality, needing to innovate to compete in the 5G market [11]. Group 3: Policy Support for Semiconductor Industry - The ability of these 7 companies to go public reflects government support for the semiconductor industry, which is crucial for achieving self-sufficiency in high-end chips and key materials amid global competition [12].

Core Viewpoint - Navitas Semiconductor has announced a strategic partnership with Powerchip Semiconductor Manufacturing Corporation (PSMC) to produce and develop 200mm silicon-based Gallium Nitride (GaN) technology after TSMC plans to terminate its GaN foundry services by 2027 [1][2]. Group 1: Partnership and Production - Navitas will utilize PSMC's 200mm wafer fabrication facility located in Zhunan Science Park, Taiwan, which has been operational since 2019, to support various GaN mass production processes [2]. - The collaboration aims to produce Navitas' GaN product portfolio with voltage ratings from 100V to 650V, addressing the growing demand for GaN in 48V infrastructure, including large-scale AI data centers and electric vehicles [2]. Group 2: Technological Advancements - PSMC's advanced 180nm CMOS process technology will enhance performance, power efficiency, integration, and cost-effectiveness of the GaN devices [2]. - The first devices are expected to complete certification by Q4 2025, with the 100V series anticipated to enter production in H1 2026, while the 650V devices will transition from TSMC to PSMC within the next 12-24 months [2]. Group 3: Market Applications and Collaborations - Navitas has recently made announcements regarding its GaN and SiC technology support for AI data centers, electric vehicles, and solar markets, including collaborations with NVIDIA and Enphase [3]. - The CEO of Navitas expressed confidence in the partnership with PSMC to advance the mass production of GaN-on-silicon technology, emphasizing ongoing innovation in product performance and cost efficiency [3].

Core Viewpoint - China is expected to become the largest semiconductor production center globally, aiming for a 30% share of the global semiconductor foundry capacity, driven by significant domestic investments and government support for self-sufficiency in chip production [1][3]. Group 1: Current Market Position - Taiwan currently leads the global semiconductor foundry market with a 23% capacity share, followed by China at 21%, South Korea at 19%, Japan at 13%, the United States at 10%, and Europe at 8% [1]. - The monthly semiconductor production in China is projected to reach 8.85 million wafers in 2024, reflecting a 15% year-on-year increase, with expectations to rise to 10.1 million wafers per month by 2025 [5]. Group 2: Investment and Capacity Expansion - China is actively constructing 18 new wafer fabs, including a collaboration between Huahong Semiconductor and a pure foundry in Shanghai to build a 12-inch wafer fab, which commenced operations in Q1 of this year [5]. - From 2025 to 2030, China is set to add 21 new wafer fabs, significantly outpacing other regions such as the United States (11), Europe (7), South Korea (6), and Taiwan (4) [6]. Group 3: U.S. Market Dynamics - The U.S. accounts for approximately 57% of global silicon wafer demand but only holds about 10% of global semiconductor capacity, necessitating substantial imports from Taiwan, South Korea, and China to meet domestic needs [5]. - Major companies like TSMC, Intel, Samsung, Micron, GlobalFoundries, and Texas Instruments are investing in new wafer fabs in the U.S., which will enhance domestic production capabilities [7]. Group 4: Technological Advancements and Challenges - Despite U.S. sanctions aimed at curbing China's chip industry, China's semiconductor sector has shown resilience and is accelerating in areas such as advanced process chip design and domestic production of semiconductor equipment [8]. - The Chinese government is investing heavily to overcome technological barriers, particularly in critical areas like lithography equipment and electronic design automation (EDA) software [7][8].

自美国实施出口限制以来，中国一直在各个科技领域努力实现自给自足，无论是人工智能还是 GPU 计 算。中国已成功研发出替代品，但我们不会过多讨论其竞争力如何。同样，中国公司砺算也成功研发出 了国内首款采用 6 纳米工艺的 GPU，虽然这看起来是个乐观的项目，但 Geekbench 测试显示，该 GPU 无法与英伟达/AMD 的现代产品相抗衡。不仅如此，其目前的性能仅相当于十三年前的 GeForce GTX 660 Ti。 根据目前我们所掌握的情况，砺算（Lisuan）上个月宣布他们一直在研发一款 6 纳米 GPU，该产品已进 入运营阶段。该公司声称这款 GPU 的性能应可与英伟达（NVIDIA）的 RTX 4060 GPU 相媲美。然而， 这款 GPU 在 Geekbench 上被发现，有趣的是，其 OpenCL 得分在 Geekbench 数据库中属于表现最差的 之一，仅获得 15524 分。其性能相当于十年前 AMD 和英伟达的产品，虽然这听起来对该公司不利，但 目前并非如此。 Geekbench 测试结果显示，该 GPU 拥有 32 个计算单元和 256MB 的显存，这显然不是目前应有的水 平，因为采用 ...

台湾地区 证交所 昨日 公告上市公司 2024年非担任主管职务全时员工薪资资讯，包括员工人数、 薪资总额、薪资平均数及中位数。 联发科非主管以平均薪资431万元（约合人民币105.38万元）夺冠，是唯一非主管平均薪资超过 400万元 新台币 （约合人民币 97.8万元）的上市公司。 （注：本文货币单位位新台币） 转自：EETOP 推荐阅读： 第八届中国 IC 独角兽榜单发布 中国半导体行业高质量发展创新成果榜单发布 加入"中国IC独角兽联盟"，请点击进入 根据证交所统计， 2024 年非主管员工 平均薪资 前十名，依序为： 至于 2024 年上市公司非主管员工 平均薪资中位数 前十名，依序为： 联发科 431 万元（约合人民币 105.38 万元） 瑞昱 391.5 万元（约合人民币 95.72 万元） 达发 357.1 万元（约合人民币 87.31 万元） 爱山林 354.1 万元（约合人民币 86.58 万元） 联咏 352.1 万元（约合人民币 85.99 万元） 瑞鼎 341.3 万元（约合人民币 83.45 万元） 台积电 339.1 万元（约合人民币 82.91 万元） 创意 316 万元（约合人 ...

Core Viewpoint - The acceptance of IPO applications for Shanghai-based Mu Xi Integrated Circuit Co., Ltd. and Beijing-based Mo Er Thread Intelligent Technology Co., Ltd. by the Shanghai Stock Exchange signifies recognition of breakthroughs in domestic GPU technology and the potential for growth in the semiconductor industry [1][10]. Group 1: Company Background - Mu Xi was established on September 14, 2020, to address the growing demand for high-performance computing and graphics processing capabilities, particularly in AI and cloud computing [2]. - Mo Er Thread was founded in October 2020, aiming to meet the urgent need for domestic GPU development and to break the foreign technology monopoly in this field [2]. Group 2: Founding Teams - Mu Xi's core team includes members from leading chip design companies like NVIDIA, AMD, and ARM, with founder Chen Weiliang having extensive experience at AMD [3]. - Mo Er Thread's team comprises top GPU talents, primarily from NVIDIA, with founder Zhang Jianzhong having over 15 years of experience in the GPU industry [3]. Group 3: Product Directions - Mu Xi focuses on developing high-performance GPU chips and computing platforms for AI, general computing, and graphics rendering, with applications in data centers and cloud computing [4]. - Mo Er Thread targets next-generation GPUs for metacomputing applications, emphasizing integration across various computing domains, including AI and 3D graphics [6]. Group 4: Competitive Landscape - Both companies face intense competition from global giants like NVIDIA and AMD, which dominate the GPU market due to their technological advantages and established ecosystems [7]. - Domestically, they also compete with other Chinese GPU manufacturers such as Cambricon and Jingjia Micro, which are increasing their R&D investments [7]. Group 5: Fundraising Plans - Mu Xi aims to raise CNY 3.904 billion for projects related to high-performance GPU development and industrialization [8]. - Mo Er Thread plans to raise CNY 8 billion to support various chip development projects and to meet significant funding needs for business expansion [8]. Group 6: Financial Performance - Mo Er Thread's projected revenue for 2024 is CNY 438 million, with cumulative R&D investment of CNY 3.81 billion from 2022 to 2024, resulting in a net loss of CNY 1.49 billion [9]. - Mu Xi anticipates a revenue of CNY 740 million in 2024, with a net loss of CNY 1.4 billion, reflecting the capital-intensive nature of GPU development [9].

Core Viewpoint - The article discusses the recent financing of Sunrise, a domestic GPU company that has completed nearly 1 billion yuan in funding, highlighting its potential in the AI chip market and its strong product lineup [1][4][12]. Company Overview - Sunrise was spun off from SenseTime at the end of 2024 and focuses on high-performance GPU development, positioning itself as a chip company that understands AI better [3][4]. - The company has successfully completed two rounds of financing within a short period, indicating strong market confidence in its future prospects [12]. Leadership and Team - The co-CEOs of Sunrise are Wang Zhan, a veteran from Baidu, and Wang Yong, a former AMD executive, both bringing significant industry experience to the company [6][8]. - The core technical team consists of only 150 members, primarily from renowned companies like AMD, Baidu, and SenseTime, showcasing a strong talent pool [8][12]. Product Line - Sunrise's product lineup includes three GPUs: S1, S2, and S3, each targeting different application scenarios [9][10]. - The S1 chip is designed for cloud-edge visual inference and has already achieved mass production with over 20,000 units shipped [11]. - The S2 chip is a GPGPU for large model inference, compatible with the CUDA ecosystem, and has performance metrics that match NVIDIA's A100, also achieving mass production [11]. - The S3 chip is under development and is expected to reduce inference costs significantly by 2026, aiming to make domestic large model inference more cost-effective [11][12]. Market Position and Challenges - Sunrise's rapid financing and product achievements position it as a potential "new hope" in the domestic GPU market, with strong backing from notable investors [12]. - The domestic GPU market is competitive, with several companies like Muxi and Biran Technology also entering the field, but challenges remain in high-end manufacturing processes and software ecosystem development [13].