• 受原料供应及春节假期影响，1 月氧化镧产量小幅下降，磁材企业复工时 间不一，导致稀土供给减少。同时，节前节后补货需求支撑市场价格，镨 钕价格已从年前的 39.1 万元/吨上涨至 44.1 万元/吨左右，显示出供需紧 张的市场态势。 • 2024 年稀土行业对价格预期总体乐观，预计氧化镨钕市场价有望达到 45 万元/吨甚至更高，保守估计全年最高可达 50 万元。供应方面，预计上半 年矿石指标将增加，废料回收和老挝矿进口也将有所增加，但缅甸矿进口 受限导致总体供应仍存在缺口。 • 缅甸矿进口受限对国内稀土供应形成压力。自 2024 年 11 月关闭以来，缅 甸矿仍未恢复正常通关，预计今年缅甸矿进口量将显著减少。若 3 月份能 恢复通关，预计进口量也将下降至 3 万吨左右，对国内稀土市场供给形成 压力。 • 废料回收将在未来几年逐步成为重要的补充来源之一，但其增量相对有限， 不足以完全弥补因缅甸等地原生矿减产带来的缺口。预计今年上半年废料 能提供 300-400 吨增量，但仍需依赖其他渠道来满足总需求。 • 独居石等其他资源开采量可能会有所增加，但由于处理资质和辐射安全问 题，其增量有限，难以完全填补因缅甸封关 ...

云资产重估：云及配套产业链梳理及投资机会展望 摘要 Q&A 近期云计算和 AI 领域有哪些显著的技术进步和投资趋势？ 近期在云计算和 AI 领域，我们观察到模型能力的大幅提升，同时推理成本显著 下降。例如，豆包提出的新实施架构使推理成本较原来的 MOE 下降了 23%。这 一趋势表明，AI 技术正在向端侧应用、推理等环节转移，推动 AI 普惠和泛化 时代的到来。与此同时，从 2024 年到 2025 年，基础资源方面的投入显著增加， 例如字节跳动计划投资 1,600 亿人民币，而阿里巴巴也大幅上调了 2025 年的资 • AI 技术正向端侧应用和推理环节转移，推动 AI 普及，字节跳动和阿里巴 巴大幅增加云计算基础资源投入，分别计划投资 1,600 亿人民币和同比增 幅超 30%的资本开支，反映市场对未来发展的信心。 • 云计算产业链部分环节供不应求，阿里巴巴云业务增长点已从电商转向 AI 模型，类似微软与 MIT 的结合，通过 AI 加速业务增长，DeepSeek 用户增 长迅速导致需求外溢，模型到业务结合再到应用价值量提升。 • 未来两三年内，云资源需求预计将指数级增长，不仅限于计算和存储，更 体现在生产 ...

Summary of Key Points from the Conference Call Industry and Company Involvement - The conference call primarily discusses the **AI market** and the collaboration between **Apple** and **Alibaba** to enhance their positions in this sector [1][25][26]. Core Insights and Arguments - The market exhibited unexpected performance, particularly with a significant rise after 1:30 PM, driven by a balance of bullish and bearish forces [2][4]. - The collaboration between Apple and Alibaba is expected to have a profound impact on the **AI industry**, particularly through the launch of **Apple Intelligence** [2][5][11]. - The **technology sector**, especially those focused on **AR (Augmented Reality)**, is currently a strong performer, with notable contributions from digital energy and semiconductor industries [2][12][13]. - The **real estate sector** saw a sudden surge, attributed to favorable news regarding major companies' debts and ongoing policy support [6][15]. - The **banking and dividend asset sectors** are becoming more active, influenced by policy support and increased capital inflow [7][19]. - The **monetary policy** has shifted to a moderately loose stance, which is expected to significantly impact economic growth and market dynamics by 2025 [8][19]. - Historical trends indicate that the market typically experiences a turnaround in the weeks leading up to and following the **Spring Festival**, with a strong upward trend expected [10][16]. Additional Important Insights - The **semiconductor industry** is projected to perform well in 2024, supported by the establishment of sub-funds under the **Big Fund Phase III**, which aims to boost domestic chip development [13]. - The **media sector** is also showing strong performance, with significant box office successes contributing to its growth [14]. - The **AR phone market** is rapidly evolving, with domestic manufacturers like **Vivo** and **Xiaomi** making substantial advancements, leading to a projected penetration rate of 22% by 2024 [27]. - The **human-shaped robot sector** is witnessing significant technological advancements, with companies like **H-One** and **Doriemus** making strides towards mass production, which will impact upstream material demand [28][29]. Conclusion - The collaboration between Apple and Alibaba is a pivotal development in the AI landscape, with broader implications for technology and real estate sectors. The market is poised for growth, driven by favorable policies, historical trends, and advancements in key industries such as semiconductors and AR technology.

Summary of Key Points from the AI Action Summit Industry and Company Involvement - The summit was co-hosted by French President Macron and Indian Prime Minister Modi, with notable attendees including US Vice President Harris, Chinese Vice Premier Zhang Guoqing, and executives from major tech companies like OpenAI and Google [2][3][4] Core Insights and Arguments - Macron called for reduced AI regulation and increased investment of €109 billion to accelerate AI development in Europe, likening it to the reconstruction of Notre-Dame [2][4] - A $30-50 billion data center investment plan was announced between France and the UAE, utilizing French nuclear energy [2][4] - The Global AI Competitiveness Index project was launched, led by France and Germany, with initial funding of $400 million aimed at providing high-quality data and open-source tools [2][5][6] - Li Fei-Fei emphasized the importance of perception algorithms, cognitive science, and computational breakthroughs in the evolution of AI, advocating for a human-centered approach [2][9] - The DeepSeek model gained attention for achieving OpenAI-like results at a lower cost, with OpenAI's CEO expressing interest in collaboration with China [2][11][12] - The summit highlighted a shift in the EU's regulatory stance, advocating for relaxed policies and increased AI investment [2][12] Additional Important Content - The summit addressed the challenges of AI governance, the uneven development of AI across nations, and the need for multilateral cooperation [3][4] - Discussions included the environmental impact of AI and the necessity for sustainable practices, with proposals for optimizing model structures to reduce energy consumption [3][15][18] - The summit revealed significant disparities in AI development, with the US leading in the number of AI companies (20,000) compared to China's 2,000, and a 41% increase in new AI companies in 2023 [2][23][26] - The report from the International Financial Forum (IFF) highlighted the concentration of AI companies in sectors like healthcare, marketing, and security, indicating a trend towards specialization [23][24] Future Directions - The IFF plans to expand the AI Competitiveness Index report to include research, innovation, and policy dimensions, with further sections to be released in 2025 [25][27] - The summit underscored the ongoing demand for high-end computing resources, despite the emergence of low-cost AI models, indicating a persistent gap in high-performance computing needs [26][28]

Summary of Key Points from the Conference Call Industry Overview - The conference call primarily discusses the **AI industry**, focusing on the valuation and growth of key players such as **NVIDIA**, **Microsoft**, and **Amazon**. The discussion highlights the significant valuation increases in the AI sector, particularly among leading companies. Core Insights and Arguments - **Valuation Trends**: Leading AI companies like NVIDIA, Microsoft, and Amazon have seen their valuations rise significantly, with Microsoft reaching historical highs. However, despite these increases, their valuations remain relatively low compared to historical averages, around **15% to 20%** of historical levels [2][3]. - **Beneficial Segments in AI Hardware**: The AI hardware value chain is primarily benefiting three segments: **AI chips**, **storage technology**, and **end devices**. AI chips, particularly, have the highest value increase and barriers to entry [2][4]. - **Computational Demand vs. Chip Performance**: Over the past six years, the demand for computational power for AI model training has surged nearly **200,000 times**, while the performance of computational chips has only improved by **260 times**. This indicates a significant gap that is expected to persist [5]. - **Capital Expenditure Growth**: In 2024, the capital expenditure of four major North American tech companies (Amazon, Google, Microsoft, and Meta) is projected to reach **$250 billion**, a **60%** increase year-over-year. This is expected to rise to **$320 billion** in 2025, indicating expanding downstream capital expenditures and increasing computational demand [6]. - **NVIDIA's Product Innovations**: NVIDIA has rapidly iterated its products over the past three years, significantly enhancing performance while reducing power consumption. For instance, the energy required for training the GPT-4 model has decreased from **1,000 GWh** to **3 GWh** [7]. - **End-Side AI Chip Development**: The development of end-side AI chips requires significant hardware upgrades and the ability to run smaller models locally. Key advancements include transitioning from **5nm** to **3nm** process technology and increasing memory capacity and bandwidth [11]. - **AI Terminal Devices**: AI technology is accelerating transformations in work and lifestyle, pushing industries towards smart transitions. AI terminals are becoming crucial for user interaction, with emerging devices like AI glasses expected to proliferate [2][19]. Additional Important Insights - **PC and Mobile Chip Developments**: The PC sector is witnessing significant advancements, with new chips like Apple's M4 and Qualcomm's Snapdragon 8,400 showing performance improvements of **25% to 30%**. This trend is expected to enhance the overall value of processors [12][14]. - **Emerging Devices and Market Trends**: New devices such as XR (VR/AR) are gaining traction, with the global AR/VR headset market expected to grow from **8 million units** in 2023 to nearly **30 million units** by 2027, reflecting a compound annual growth rate of over **30%** [27]. - **AI Glasses and Future Innovations**: The year 2025 is anticipated to be pivotal for AI glasses, with major companies like Xiaomi and Huawei planning to launch products that integrate AI capabilities, potentially leading to new applications and business models [31]. - **Long-term Focus Areas**: The development of AI engines and the transition of AI technology into mobile devices are critical areas to watch. Companies like Apple are investing in self-developed technologies that may lead to significant innovations in the future [32][33].

成本下降促使更多企业能够负担得起私有化部署大模型，从而推动应用软件支 出的增加。同时，这也促使软件公司的商业模式发生变化。传统的软件公司主 要依赖于产品销售，而随着大模型的普及，它们将逐渐转向订阅制 SaaS 模式。 这种转变在美国已经基本完成，但中国的软件公司进展相对较慢。然而，由于 大模型带来的便利和成本优势，中国的软件公司也将加速向 SaaS 模式转型。未 来，应用软件将不仅仅是提供软件服务，而是通过不断迭代更新的大模型来提 供服务。这种机器学习模型即服务（MLaaS）的商业模式将逐渐被企业用户接受， 因为它能够更频繁地更新并满足新的业务需求。 再谈 ERP：用友和金蝶 20250213 摘要 Q&A Deepseek 技术的推出对软件股市场产生了哪些具体影响？ Deepseek 技术的推出显著降低了企业部署大模型的成本，这对软件股市场产生 了积极影响。过去，企业购买大模型需要花费至少 500 万甚至上千万元，并且 还需自行配置算力。然而，Deepseek 技术使得整个部署成本迅速下降到 50 万 元左右，这其中包括算力。以 37B 蒸馏版为例，只需两台 4,090 服务器即可满 足需求，总成本约为 1 ...

ab 12 February 2025 Global Research China Semiconductors Initial implications of DeepSeek and China's AI progress Three major implications of DeepSeek for China's tech supply chain Mainly stimulated by DeepSeek and improved investor sentiment, the China semi stock index has risen 7.5% after Chinese New Year, outperforming the HS300/SOX indexes by 5.3ppt/2.8ppts. We think DeepSeek has shown real AI computing innovation cost efficiently, and we believe more AI propagation is positive for AI computing spending ...

Summary of BYD Company Limited Conference Call Company Overview - **Company**: BYD Company Limited - **Industry**: Automotive, specifically focusing on electric vehicles (EVs) and autonomous driving (AD) technologies Key Highlights from the Conference Call 1. **Product Launches and Innovations**: - BYD introduced its 'God's Eye system', a new navigation on autopilot (NOA) solution for city and highway driving, at a technology day event on February 10, 2025. This system will be available in 21 models, including those priced above Rmb100k and selected models below this price point [1][2][3] 2. **Pricing Strategy**: - The starting prices for new models featuring the highway NOA function are set between Rmb69.8k and Rmb79.8k, which is lower than the previously expected range of Rmb90k to Rmb100k. This pricing strategy is anticipated to positively surprise the market [2] - There will be no significant price increase for the NOA feature, contrary to investor expectations that costs would be passed on to consumers. This is attributed to scale effects and aggressive negotiations with suppliers, resulting in a cost reduction of approximately 15-20% [2] 3. **Market Penetration and Sales Forecast**: - BYD expects that around 80-85% of its domestic sales by the end of 2025 will come from models equipped with either highway or city-level NOA features. The company forecasts an AD penetration rate of 40-50% for its domestic sales throughout 2025 [3] 4. **Competitive Positioning**: - BYD's extensive distribution channels and diverse product portfolio across various price points are expected to enhance its market share and drive industry AD penetration, positioning it favorably against competitors like Huawei and XPeng [3] 5. **Global Expansion Plans**: - BYD aims to achieve global deliveries of 6.5 million units by 2026, with approximately 1.5 million units from overseas markets. This is projected to increase its market share in the global light vehicle market from 3% in 2023 to 7% by 2026 [8][13] - The company is set to complete four overseas production bases in Thailand, Indonesia, Brazil, and Hungary by 2026, which will support its global ambitions despite facing rising tariffs in the EU [8][13] 6. **Valuation and Price Target**: - Analysts have set a price target of HK$475.00 for BYD's stock, with expectations of consensus earnings upgrades following the results season in March 2025. The valuation is based on a sum-of-the-parts (SOTP) analysis, considering various segments such as EV battery and NEV manufacturing [5][7][9] Risks and Challenges - Potential risks include worse-than-expected sales performance and increased competition from both domestic and international mass-market brands, such as Volkswagen, Geely, and Great Wall Motor [12][17] Conclusion - BYD's recent product launches and strategic pricing are expected to enhance its competitive edge in the EV market. The company's ambitious global expansion plans and projected sales growth indicate a strong outlook, although it must navigate potential risks from competition and market dynamics.

V i e w p o i n t | 10 Feb 2025 15:38:42 ET │ 19 pages Taiwan Electronics & Semiconductors Taiwan technology sector monthly tracking – High-end applications support a good start of the year CITI'S TAKE Based on Taiwanese technology companies' reported Jan sales, we note that advanced applications such as AI lead better YoY momentum at the beginning of the year. Server players Jan sales were tracking ahead (-11% MoM/+81% YoY). Wiwynn and Accton both delivered better-than-seasonality monthly sales due to robu ...

10 February 2025 U.S. Semiconductors and Semiconductor Capital Equipment U.S. Semiconductors - Deconstructing 2024 Stacy A. Rasgon, Ph.D. +1 212 756 4403 stacy.rasgon@bernsteinsg.com Alrick Shaw +1 212 969 1458 alrick.shaw@bernsteinsg.com Arpad von Nemes +1 212 969 1518 arpad.vonnemes@bernsteinsg.com On the surface, 2024 was a solid upturn for the semiconductor industry, with total sales rising 19% YoY following an 8% decline in 2023. Growth was however dominated by memory (which rose 79% YoY following 2023 ...