Core Viewpoint - The announcement of the stock split for the Sci-Tech Innovation Artificial Intelligence ETF (589520) aims to lower the trading threshold for investors, facilitating broader participation in the domestic AI industry investment opportunities [1] Fund Split Details - The ETF completed a 1:2 stock split, reducing the trading threshold from approximately 120 yuan to about 60 yuan per unit [1] - Before the split, the total fund shares were 465 million, with a net value of 1.1681 yuan; after the split, the total shares increased to 930 million, and the net value decreased to 0.5841 yuan [4] Performance Analysis - Since the low point on April 8, the ETF has increased by 51.7%, outperforming the Sci-Tech 50 index (37.26%) and the Sci-Tech Composite Index (48.29%) [4][9] - The semiconductor sector is highlighted as a key area for growth, with expectations for significant market capacity expansion by 2025 due to domestic demand [5] Market Outlook - Analysts suggest that the AI sector may lead the current bull market, with its pervasive impact across various industries [5] - The ETF is positioned to benefit from the ongoing domestic semiconductor market growth, especially in light of U.S. restrictions on chip exports [5] Investment Highlights - The ETF is seen as a low-threshold investment option with a high degree of elasticity, allowing for efficient capital deployment during market upswings [7] - The top ten holdings account for over 60% of the ETF's weight, with the semiconductor sector representing nearly half of this concentration [8]

人民财讯9月8日电，企查查APP显示，近日，威兆集成电路（舟山）有限公司成立，法定代表人为张小 青，注册资本8000万元，经营范围包括集成电路芯片及产品制造、集成电路芯片及产品销售、集成电路 设计等。企查查股权穿透显示，该公司由深圳市威兆半导体股份有限公司全资持股。 ...

Core Insights - Semtech Corporation has introduced two new FiberEdge transimpedance amplifiers (TIAs), the GN1834D and GN1818, aimed at enhancing power efficiency in AI data centers and supporting the transition to 1.6T optical interconnects [1][2][4] Product Overview - The GN1834D is designed for the emerging 1.6T optical interconnect market, enabling 8x200G optical modules and providing advanced performance with innovative 2.5D mounting technology [4][6] - The GN1818 focuses on optimizing power efficiency for existing 800G infrastructure, achieving up to 20% power reduction compared to alternatives while maintaining high performance [5][6] Market Context - The demand for high-speed optical interconnects is driven by the exponential growth of AI workloads, with the high-speed datacom transceiver market projected to grow from approximately $9 billion in 2024 to over $17 billion by 2026 [3] - As operators transition to 1.6T optics, Semtech's comprehensive 200G TIA portfolio provides the necessary performance and design flexibility for next-generation optical modules [3][6] Strategic Advantages - Semtech's broad portfolio allows for customer standardization across diverse connectivity requirements, enhancing supply chain efficiency and enabling module manufacturers to optimize designs for specific applications [6][8] - The company's approach to power efficiency and performance positions it as a leader in the emerging 1.6T market, catering to the strict power budgets demanded by AI workloads [2][3][6]

J P M O R G A N Asia Pacific Equity Research 03 September 2025 This material is neither intended to be distributed to Mainland China investors nor to provide securities investment consultancy services within the territory of Mainland China. This material or any portion hereof may not be reprinted, sold or redistributed without the written consent of J.P. Morgan. Silicon Carbide Sector Technology Jimmy Huang AC (886-2) 2725-9865 jimmy.huang@jpmorgan.com J.P. Morgan Securities (Taiwan) Limited Gokul Hariharan ...

Summary of VeriSilicon Conference Call Company Overview - **Company**: VeriSilicon (688521.SS) - **Industry**: Semiconductor, specifically focusing on AI ASIC solutions and semiconductor IP products Key Points Order Momentum and Growth - Strong order momentum is evident, with orders on hand reaching **Rmb3 billion**, reflecting a **33% YoY** and **23% QoQ** increase, driven by AI ASIC design services and production management orders [4] - Over **80%** of these orders are expected to convert to revenue within one year, supporting sequential growth [4] - The company has **100+ design projects** on hand, indicating growing demand from AI computing projects [4] AI ASIC Projects - Management is optimistic about the growth of AI/ASIC solutions, particularly for cloud and edge devices, including AI/AR glasses, AI smartphones, AI Pads, and AI toys [4] - The trend towards localization of AI chips is expected to benefit the company [4] - Private placement funding is being utilized to accelerate new product development in ASIC IP and GPU IP [4] Semiconductor IP Coverage - VeriSilicon has a comprehensive semiconductor IP coverage, including GPU IP, NPU IP, VPU IP, DSP IP, Display IP, and over **1,600** analog/mixed signal IP and RF IP [4] - In-house IP products are designed to optimize chip size, bandwidth, and performance, saving clients time and costs [4] Financial Projections - The 12-month price target for VeriSilicon is set at **Rmb193**, representing a **26.1% upside** from the current price of **Rmb153** [7] - Revenue projections for the next few years are as follows: - **2024**: Rmb2,321.9 million - **2025E**: Rmb3,025.6 million - **2026E**: Rmb4,951.9 million - **2027E**: Rmb6,526.3 million [7] Risks - Potential risks include: 1. Slower-than-expected technology development [6] 2. Higher-than-expected costs for talent acquisition and retention [6] 3. Weaker-than-expected customer spending on IP and new chipset projects [6] Conclusion - Overall, management remains positive about the rising demand for AI solutions and the company's ability to capitalize on this trend through its AI ASIC turnkey solutions and extensive semiconductor IP offerings [4]

Summary of Key Points from the Conference Call on China's AI Chip Market Industry Overview - The focus of the conference call was on the **AI chip market in China**, specifically addressing the potential growth in production capacity by 2026 and its implications for foreign vendors like **NVIDIA** [1][2]. Core Insights - **Demand Growth**: The expert predicts that domestic AI chip shipments in China will accelerate significantly in **2026**, with estimates of demand reaching **2.9-3.6 million units** [3]. - **Market Share Dynamics**: - In **2024**, NVIDIA is expected to hold approximately **70%** of the market share, with Huawei at **20%** and other players like AMD and T-Head making up the remainder [3]. - By **2025**, NVIDIA's share is projected to drop to **50%**, while Huawei's share is expected to increase to **30%** [3]. - For **2026**, NVIDIA's market share is anticipated to further decline to around **30%**, with Huawei expected to capture **50%** [3]. Production Capacity Insights - **SMIC's Capacity Expansion**: The expert estimates that **SMIC** will increase its sub-7nm capacity from **15kwpm** to **25kwpm** by the end of **Q3 2025**, which could significantly impact AI chip production [4]. - **Yield Rates**: The yield rate is estimated at **20%**, with each **5kwpm** capacity generating approximately **900,000 AI chip dies** annually [4]. Challenges and Constraints - The expert highlighted that the growth in demand for AI chips in **2024-2025** is constrained by: 1. The **H20 ban**. 2. **US restrictions** on Chinese IC design companies collaborating with **TSMC** for high-performance AI chips [3]. - Despite these challenges, some Chinese IC design companies are still able to produce AI chips for inference purposes without high-bandwidth memory (HBM) at TSMC [3]. Additional Considerations - The expert expressed skepticism regarding claims in a **Financial Times** article that suggested China could triple its AI chip output by **2026**, labeling it as "too exaggerated" [2][3]. - The expert emphasized the need for breakthroughs in domestic semiconductor equipment to achieve further capacity increases beyond **25kwpm** [4]. Conclusion - The conference call provided a nuanced view of the **AI chip market in China**, highlighting both the potential for growth and the significant challenges posed by geopolitical factors and technological constraints. The insights suggest a competitive landscape where domestic players are poised to increase their market share at the expense of established foreign vendors like NVIDIA.

Core Insights - The artificial intelligence (AI) sector is experiencing significant growth, with major tech companies projected to spend over $300 billion on AI infrastructure this year [2] - Microsoft has a substantial backlog of $368 billion in contracted revenue, indicating strong demand for its services [4][7] - Microsoft Azure is growing rapidly, with a reported 39% year-over-year revenue growth and expectations for continued growth [10] Company Summaries - Microsoft is leading in AI infrastructure spending, committing $30 billion in capital expenditures for the current quarter, and is expected to continue investing heavily to meet demand [12] - Microsoft’s backlog includes long-term commitments, with only 35% expected to be recognized as revenue in the next 12 months, while the amount recognized beyond 12 months grew by 49% [9] - Azure is now a $75 billion business, significantly larger than Google Cloud, and is expected to maintain a growth rate of 37% in the next quarter [10] Industry Trends - Demand for cloud computing services, particularly AI services, is outpacing supply across the industry, with similar sentiments expressed by Amazon and Alphabet [11] - The growth in long-term commitments for cloud services is a trend seen across major players, with Google Cloud's backlog at $108 billion and Amazon Web Services at $195 billion [7] - The integration of AI into enterprise software, such as Microsoft 365, is enhancing productivity and driving higher commitments from commercial customers [13]

Core Viewpoint - Moer Technology is seeking to raise approximately 8 billion yuan through its IPO on the Sci-Tech Innovation Board, focusing on the development of next-generation AI and graphics chips [1] Company Overview - Moer Technology specializes in the research, design, and sales of GPUs and related products [1] - The company’s IPO application was accepted on June 30, 2025, and entered the inquiry phase on July 17, 2025 [1] Fundraising Purpose - The funds raised will be allocated to the development of a new generation of self-controlled AI training and inference integrated chips, a new generation of self-controlled graphics chips, a new generation of self-controlled AI SoC chips, and to supplement working capital [1] Regulatory Inquiry - The first round of inquiry raised questions regarding the company's product and market competition, core technology, lack of profitability, customer concentration, and distribution model [1]

确定性执行消除了动态执行的低效性和漏洞。预测性执行无需动态调度指令并回滚错误路径，而是确 保每条指令都在正确的时间以正确的资源发出。它不仅更高效，而且可预测、可扩展，并且本质上更 安全。 突破在于 Simplex 所谓的"时间资源矩阵"：这是一种新颖的专利调度机制，可以跨时间分配计算、 内存和控制资源。每条指令都有指定的时隙和访问窗口，确保零重叠并消除流水线停顿。可以将其想 象成火车时刻表——只不过火车是在同步计算结构中移动的标量、矢量和矩阵运算。 半个多世纪以来，计算的基础一直建立在单一架构之上：冯·诺依曼模型或哈佛模型。几乎所有现代 芯片——CPU、GPU，甚至许多专用加速器——都依赖于这种设计的某种变体。随着时间的推移，业 界不断提升复杂性和专业化程度，以满足新的需求。超长指令字 (VLIW) 架构、数据流芯片和 GPU 最初都是作为针对特定瓶颈的单点解决方案而引入的，但都未能提供全面的替代方案。直到现在。 Simplex Micro 开发出的技术可能是半个多世纪以来对传统范式最重大的突破——在单一确定性流水 线中实现统一的标量、矢量和矩阵计算。其核心是一个革命性的概念：预测执行。与猜测下一步会发 生什 ...

Core Viewpoint - Kopin Corporation faces a legal ruling in favor of BlueRadios, Inc., resulting in approximately $19.7 million in damages, while a permanent injunction was denied, indicating no immediate risk to customer orders [1][2]. Legal Outcome - The U.S. District Court for the District of Colorado awarded BlueRadios, Inc. $19.7 million in damages, which is a significant reduction compared to jury recommendations [1][2]. - The court denied BlueRadios' request for a permanent injunction, affirming that there is no risk to current or future customer orders for Kopin [2]. Company Response - Kopin's CEO, Michael Murray, expressed the belief that the monetary judgment is unsupported and inconsistent with governing law, indicating plans to appeal the decision [2]. Company Overview - Kopin Corporation specializes in application-specific optical systems and high-performance microdisplays for various sectors, including defense, training, enterprise, industrial, consumer, and medical products [1][4]. - The company's product portfolio includes microdisplays, display modules, eyepiece assemblies, and various types of displays such as AMLCD, FLCoS, µLED, and OLED [4][5]. Recent Developments - Kopin announced a follow-on contract worth approximately $9 million for a custom thermal imaging assembly for a major U.S. defense contractor, showcasing ongoing business activity despite the legal challenges [9]. - The company is also participating in the Lake Street Capital Markets 9th Annual Best Ideas Growth Conference, indicating engagement with investors and market stakeholders [8].