证券研究报告 2025年06月09日 海外行业报告：行业动态研究 快手可灵AI年化收入破1亿美元，谷歌新版Gemini 2.5 Pro 强势登顶 作者： 分析师 孔蓉 SAC执业证书编号：S1110521020002 分析师 李泽宇 SAC执业证书编号：S1110520110002 分析师 樊程安吉 SAC执业证书编号：S1110524080001 分析师 杨雨辰 SAC执业证书编号：S1110521110001 分析师 刘诗雨 SAC执业证书编号：S1110524120001 请务必阅读正文之后的信息披露和免责申明 1 全球AI周报 摘要 n 重点公司财报AI进展 n 全球AI动态： n 投资建议 Ø 海外AI： Ø 随着财报披露逐步收尾，我们看到AI主线在多个环节加速落地，显著对冲了市场对宏观与政策不确定性的阶段性悲观情绪。美股科技大厂对AI的资本投入与应用商业化均未放缓，AI主线逻辑持续强化。1）AI算力：Broadcom定制加速器持续 放量，AI基础设施需求强韧。Broadcom Q2营收达150亿美元，同比增长20%，其中AI相关半导体收入达44亿美元，同比+46%，大幅超预期。定制化AI加速器（A ...

The latest video from Boston Dynamics shares new insights about how its Atlas humanoid robot perceives the world around it. We dig into the details and explain why Boston Dynamics and many others in the robotics industry are aiming to move away from training robots one task at a time and towards training that makes robots more generally useful and versatile. In the latest demo video, Atlas is shown picking and placing car parts into their designated slots.An activity we've seen the robot do in previous vide ...

Group 1: Foreign Exchange and Gold Reserves - As of May 2025, China's foreign exchange reserves reached $32,853 billion, an increase of $36 billion from April, representing a growth rate of 0.11% [1] - By the end of May, China's gold reserves stood at 7,383 million ounces (approximately 2,296.37 tons), with a month-on-month increase of 60,000 ounces (about 1.86 tons), marking the seventh consecutive month of gold accumulation [1] Group 2: A-Share Market Analysis - After the "fill the pit" market phase, A-shares are expected to experience a strong upward trend driven by policies, external events, industry rotation, and market sentiment [2][3] - Historical data shows that after seven instances of rapid declines followed by rebounds since 2015, the average duration of subsequent market fluctuations lasted 86 trading days, with five out of seven instances leading to price increases [2] - Current positive policies and external events, including a recent call between the leaders of China and the U.S., have reduced concerns regarding U.S.-China tensions, contributing to a favorable market outlook [3] Group 3: Economic Indicators - Industrial profits for large-scale enterprises grew by 3% year-on-year in April, with a cumulative growth of 1.4% from January to April, indicating a continued positive trend [5] - The manufacturing PMI for May rose to 49.5%, reflecting an improvement in manufacturing activity, with production indices exceeding the critical threshold [5][6] - The market remains cautious about the impact of U.S. tariffs, but the overall economic indicators suggest a supportive environment for market stability [6] Group 4: Industry Configuration and Investment Recommendations - Short-term investment strategies should focus on technology, new consumption, and undervalued blue-chip stocks, with an emphasis on sectors like computing, communication, and innovative pharmaceuticals [4] - The technology and advanced manufacturing sectors are highlighted as key areas for investment, supported by ongoing policy initiatives and market trends [6] - Companies in the U.S. and Hong Kong, such as Circle and Coinbase, as well as A-share companies like Sifang Jingchuang and Zhongke Jincai, are identified as potential investment targets in the context of stablecoins and RWA [7][9] Group 5: NVIDIA's Robotics Ecosystem - NVIDIA is focused on developing platforms and tools for robotics, emphasizing the creation of a foundational development ecosystem [10] - The company has introduced a comprehensive solution for robotics that integrates training, optimization, and execution through its various computing platforms [11][12] - Investment recommendations include companies involved in automotive robotics, such as Top Group and Bertly, which are expected to benefit from NVIDIA's advancements in the robotics sector [13]

Investment Rating - The semiconductor production equipment industry is rated as Attractive [5][70]. Core Insights - The imminent launch of DeepSeek R2, which features 1.2 trillion parameters and significant cost efficiencies, is expected to positively impact the Japanese semiconductor production equipment (SPE) industry [3][7][11]. - The R2 model's capabilities include enhanced multilingual support, broader reinforcement learning, multi-modal functionalities, and improved inference-time scaling, which could democratize access to high-performance AI models [7][9][11]. - The development of efficient AI models like R2 is anticipated to increase demand for AI-related SPE, benefiting companies such as DISCO and Advantest [11]. Summary by Sections DeepSeek R2 Launch - DeepSeek's R2 model is reported to have 1.2 trillion parameters, a significant increase from R1's 671 billion parameters, and utilizes a hybrid Mixture-of-Experts architecture [3][7]. - The R2 model offers cost efficiencies with input costs at $0.07 per million tokens and output costs at $0.27 per million tokens, compared to R1's $0.15-0.16 and $2.19 respectively [3][7]. Industry Implications - The launch of R2 is expected to broaden the use of generative AI, leading to increased demand for AI-related SPE across the supply chain, including devices like dicers, grinders, and testers [11]. - The report reiterates an Overweight rating on DISCO and Advantest, which are positioned to benefit from the anticipated increase in demand for AI-related devices [11]. Company Ratings - DISCO (6146.T) is rated Overweight with a target P/E of 25.1x [12]. - Advantest (6857.T) is also rated Overweight, with a target P/E of 14.0x [15].

Investment Rating - The semiconductor production equipment industry is rated as Attractive [5] Core Insights - The imminent launch of DeepSeek R2, which features 1.2 trillion parameters and significant cost efficiencies, is expected to positively impact the Japanese semiconductor production equipment (SPE) industry [3][7] - The development of lightweight, high-performing AI models like DeepSeek R2 is anticipated to democratize access to generative AI, thereby expanding the market for AI-related SPE [11] Summary by Sections DeepSeek R2 Characteristics - DeepSeek R2 is reported to have 1.2 trillion parameters, with 78 billion active parameters and utilizes a hybrid Mixture-of-Experts architecture [3] - The input cost for R2 is $0.07 per million tokens, significantly lower than R1's $0.15-0.16, while the output cost is $0.27 compared to R1's $2.19 [3][7] - Enhanced multilingual capabilities and broader reinforcement learning are key upgrades in R2, allowing it to handle various data types including text, image, voice, and video [9][11] Market Implications - The anticipated launch of R2 is expected to boost demand for AI-related devices, including GPU and HBM, as well as custom chips and other AI devices [11] - The report reiterates an Overweight rating on DISCO and Advantest, which are expected to benefit from increased demand for AI-related devices [7][11] Company Ratings - Advantest (6857.T) is rated Overweight with a target price of ¥10,300 based on expected earnings peak [16] - DISCO (6146.T) is also rated Overweight with a target P/E of 25.1x based on earnings estimates [13]

Core Insights - CoreWeave, in collaboration with NVIDIA and IBM, achieved a significant milestone by delivering the largest-ever MLPerf® Training v5.0 submission using 2,496 NVIDIA Blackwell GPUs, marking a 34x larger scale than the next largest submission from a cloud provider [1][2] - The submission completed the training of the Llama 3.1 405B model in just 27.3 minutes, demonstrating over 2x faster training performance compared to other similar cluster submissions, showcasing the capabilities of the GB200 NVL72 architecture [2][3] - CoreWeave's cloud platform is designed to meet the demands of AI workloads, enabling faster model development cycles and optimized Total Cost of Ownership, effectively cutting training time in half for its customers [3] Company Overview - CoreWeave, known as the AI Hyperscaler™, provides a cloud platform that supports accelerated computing for enterprises and AI labs, with a growing presence of data centers across the US and Europe since its inception in 2017 [4] - The company has been recognized as one of the TIME100 most influential companies and featured in the Forbes Cloud 100 ranking in 2024, highlighting its impact and leadership in the cloud computing space [4]

Summary of Conference Call on Taiwan Technology: Semiconductors Industry Overview - The semiconductor industry is experiencing a shift towards more complex chip designs, particularly in the AI and High-Performance Computing (HPC) segments, which is driving demand for advanced chip testing [1][11][58]. - The probe card and socket markets are projected to grow steadily, with the probe card market reaching US$2.5 billion and the socket market reaching US$1.6 billion in 2024, representing approximately 17% and 11% of the total testing equipment market, respectively [52]. Key Companies MPI (6223.TWO) - MPI is ranked as the 4th largest probe card provider globally, with 57% of its revenue in 2024 derived from probe cards [10][52]. - The company is expected to achieve a revenue CAGR of 19% from 2024 to 2027, driven by increased demand in the AI/HPC segments [14][33]. - MPI's market share in the probe card market is projected to rise from 7% in 2024 to 11% by 2027 [13]. WinWay (6515.TW) - WinWay is the 2nd largest socket provider globally, generating 71% of its revenue from test and burn-in sockets [10][52]. - The company is expected to see a revenue CAGR of 23% from 2024 to 2027, benefiting from its expansion into the probe card market [14][37]. - WinWay's market share in the socket market is expected to increase from 8% in 2024 to 14% by 2027 [13]. Investment Opportunities - Both MPI and WinWay are initiated with a Buy rating and target prices of NT$1,000 and NT$1,500, respectively, implying potential upsides of approximately 41% and 54% [3][16]. - The investment thesis is supported by: 1. Increasing dollar content in the testing industry due to AI and HPC trends, leading to a doubling of average selling prices (ASPs) for probe cards and sockets [3][11][59]. 2. Local advantages for Taiwanese vendors, including proximity to leading foundries like TSMC and OSAT companies, which are expected to enhance market share [13][12]. 3. Expansion into new markets, particularly in high-margin segments like MEMS probe cards for MPI and probe cards for WinWay [14][58]. Financial Performance - MPI's earnings are forecasted to grow at a CAGR of 28% from 2024 to 2027, while WinWay's earnings are expected to grow at a CAGR of 37% during the same period [39][41]. - MPI's 2024 revenue is projected at NT$10.17 billion, with net income expected to reach NT$2.30 billion [10][33]. - WinWay's 2024 revenue is projected at NT$5.80 billion, with net income expected to reach NT$1.19 billion [18][37]. Risks - Potential downside risks include softer demand in AI and HPC markets, slower penetration into new total addressable markets (TAM), and intensifying competition [4]. Conclusion - The semiconductor testing industry, particularly for advanced chip testing, is poised for significant growth driven by increasing complexity in chip designs and the demand for faster time to market. MPI and WinWay are well-positioned to capitalize on these trends, making them attractive investment opportunities despite current market volatility.

Core Viewpoint - Cisco has signed a Memorandum of Understanding (MoU) to join the Stargate UAE consortium as a preferred technology partner, aiming to develop AI infrastructure in Abu Dhabi, including a data center with a target capacity of 1 GW, starting with an initial 200 MW by 2026 [2][6]. Group 1: Partnership and Collaboration - Cisco will collaborate with major technology firms including G42, OpenAI, Oracle, NVIDIA, and SoftBank Group to enhance AI innovation and infrastructure development [2][4]. - The partnership aims to provide advanced networking, security, and observability solutions to support the deployment of next-generation AI compute clusters [4][6]. Group 2: Strategic Importance - The initiative is part of a broader strategy to foster AI development in the Middle East, leveraging Cisco's expertise in networking and security [5][6]. - Cisco's involvement aligns with recent regional initiatives announced by the company, emphasizing the importance of secure, AI-powered digital infrastructure [5][6]. Group 3: Leadership and Vision - Cisco's Chair and CEO, Chuck Robbins, highlighted the transformative potential of AI in turning data into actionable insights, reinforcing the company's commitment to innovation and infrastructure development [5][6]. - The signing ceremony was attended by key figures, including UAE President His Highness Sheikh Mohammed bin Zayed, indicating the project's significance at a national level [3][6].

Core Insights - Primech AI Pte. Ltd. has successfully deployed its autonomous bathroom cleaning robot, HYTRON, at a major shopping destination in Singapore, marking a significant step in the company's mission to innovate facility services through robotics and AI technology [1][5] - The HYTRON robot utilizes NVIDIA's Jetson Orin Super module, which enhances its real-time AI processing, energy efficiency, and decision-making capabilities, making it suitable for high-traffic public environments [2][3] - The deployment of HYTRON is aimed at improving cleanliness and hygiene standards in commercial spaces, showcasing the company's commitment to addressing challenges in facilities management with advanced technology [3][5] Company Overview - Primech AI is a leading robotics company focused on innovation in technology, aiming to revolutionize the robotics industry with impactful solutions [6] - Primech Holdings Limited, the parent company, provides technology-driven facility services across Singapore, including specialized cleaning solutions and general maintenance, emphasizing sustainability and operational efficiency [7]

Investment Rating - The report indicates a modest sales beat expectation of approximately $500 million for NVIDIA's April quarter, with flat expectations for the July quarter at around $44 billion, compared to the Street's $46 billion [1] Core Insights - The CEO of NVIDIA, Jensen Huang, asserts that AI will become essential for every company, highlighting ChatGPT as a rapidly growing product that has transformed information retrieval [2] - Huang notes that AI diffusion was initially misunderstood, and recent changes in U.S. government attitudes towards AI restrictions have influenced NVIDIA's operations, particularly in the Middle East [3] - AI investments are projected to positively impact countries' GDP in the long term, especially in nations with excess energy that can export AI factories [4] - China is recognized as a significant but singular market for AI, with expectations that its AI market will reach $50 billion in a few years [5] - NVIDIA has been developing Physical AI for a decade, viewing it as a high-volume shipping opportunity, with the self-driving business currently valued at $5 billion [6] - NVLink Fusion technology is in demand from NVIDIA's customers, with the company advancing to NVLink Gen 5, which offers substantial improvements in bandwidth and latency [7][8] - NVIDIA emphasizes that its success will not compromise its partners and is actively diversifying its supply chain [9] - The company anticipates maintaining over 80% market share in the ASIC sector due to its rapid advancements [9]