Report Industry Investment Rating The provided content does not mention the report industry investment rating. Core Viewpoints of the Report - The inclusion of financial institutions such as banks has led to a record - high issuance of science and technology innovation bonds (Sci - tech bonds). The annual issuance of Sci - tech bonds may reach 2 trillion yuan [2][3]. - The first batch of Sci - tech bond ETFs was issued hotly, and attention should be paid to cross - market trading opportunities [3]. Summary According to the Directory 1. High Proportion of Financial Institutions, Annual Issuance of Sci - tech Bonds May Reach 2 Trillion - After the new policy, the issuance of Sci - tech bonds increased significantly, with a high proportion of financial institutions such as banks and non - banks. Since the launch of the bond "Sci - tech board" in May 2025, the issuance of Sci - tech bonds in Q2 2025 increased to over 70 billion yuan, and banks accounted for nearly 40% of the total issuance of about 59 billion yuan [2][10]. - Estimated from the historical growth trend of industrial Sci - tech bonds, the issuance in 2025 may reach around 1.7 trillion yuan. From the perspective of bank asset - liability management, there is still about 24 billion yuan of issuance space for bank Sci - tech bonds, and the total of the two is about 2 trillion yuan [3][12][16]. 2. Hot Issuance of Sci - tech Bond ETFs, Pay Attention to Cross - market Trading Opportunities 2.1 Hot Primary Subscription of Sci - tech Bond ETFs, High - rated and Long - duration Component Bonds - The first batch of 10 Sci - tech bond ETFs were all sold out on the first day of their launch, raising a total of about 30 billion yuan. The underlying index component bonds are high - rated bonds, all being publicly - offered bonds of AAA - rated entities listed on the exchange [17]. - The industry distribution of the component bonds of the Shanghai Stock Exchange AAA Sci - tech bond index is concentrated, with the construction industry having a high proportion and relatively long duration. The industry distribution of the component bonds of the Shenzhen Stock Exchange AAA Sci - tech bond index is relatively dispersed, and the duration is also relatively long [18][21]. 2.2 Similar Issuance Volumes of Sci - tech Bonds in the Inter - bank and Exchange Markets, the Exchange Market is Expected to Expand Further - The balance of Sci - tech bonds in the exchange market is currently 1.27 trillion yuan, which is relatively small compared to other mainstream bond varieties. The proportion of AAA - rated entities is high, and most of the bonds have an implied rating of AA+ or above [24]. - Since 2021, the issuance volumes of Sci - tech bonds in the exchange and inter - bank markets have been similar. After the launch of the bond Sci - tech board in May 2025, the number of inter - bank listed Sci - tech bonds was significantly higher. However, considering the issuance situation of industrial Sci - tech bonds, the scale of the exchange market is expected to increase [26]. 2.3 Valuation Differentiation between the Inter - bank and Exchange Markets, Pay Attention to Cross - market Trading Opportunities - Driven by factors such as the market's expectation of enhanced liquidity of index component bonds, ETF issuers' advance reserve of individual bonds, and institutional arbitrage motives, the trading volume of index component bonds has increased significantly, and the yields in the inter - bank and exchange markets have diverged [31]. - The yields of medium - and long - term component bonds are generally lower than those of inter - bank bonds of the same level and term. There may still be some downward space for component bonds in the construction industry with a 7 - 10 - year term and in the public utilities industry with a 3 - 5 - year term, as well as for short - term component bonds [4].

- The report does not contain any quantitative models or factors related to the requested analysis

证券研究报告：有色金属 | 公司点评报告 发布时间：2025-07-10 股票投资评级 资料来源：聚源，中邮证券研究所 公司基本情况 | 最新收盘价（元） | 11.49 | | --- | --- | | 总股本/流通股本（亿股）10.73 | / 7.52 | | 总市值/流通市值（亿元）123 | / 86 | | 52 周内最高/最低价 | 11.49 / 6.79 | | 资产负债率(%) | 50.8% | | 市盈率 | 17.41 | | 第一大股东 | 湖南水口山有色金属集 | | 团有限公司 | | 研究所 分析师:李帅华 SAC 登记编号:S1340522060001 Email:lishuaihua@cnpsec.com 研究助理:杨丰源 SAC 登记编号:S1340124050015 Email:yangfengyuan@cnpsec.com 株冶集团(600961) 25 年中报预增：冶炼端与矿山端向上共振 l 25Q2 归母净利润中枢为 3.28 亿元 公司发布 2025 年中报预增：经初步测算，公司预计 2025 年半年 度实现归属于母公司所有者的净利润 5.6 亿元到 6.5 ...

证券研究报告：固定收益报告 发布时间：2025-07-10 研究所 分析师:梁伟超 SAC 登记编号:S1340523070001 Email:liangweichao@cnpsec.com 近期研究报告 《策略选择"骑虎难下"？——流动 性周报 20250706》 - 2025.07.07 固收专题 下半年政府债供给怎么看？ ⚫ 国债发行节奏整体偏快，下半年供给压力趋于缓和 上半年普通国债发行 6.83 万亿，净融资 2.53 万亿，同比多增 3438 亿，发行期数减少、单期规模增加；特别国债发行 10550 亿，净 融资 8550 亿，同比多增 6050 亿，集中到期情况下，发行进度仍靠前。 我们估算下半年国债约有 7.78 万亿待发，预计实现净融资 3.19 万亿，下半年供给压力有望缓和。其中，普通国债下半年约有 6.94 万 亿待发，预计节奏将相对平稳。特别国债有 7450 亿待发，单期规模 或有所增加。综合来看，全部国债的发行高峰或集中于 7-9 月，单月 发行 1.3-1.5 万亿。净融资高峰有较大概率出现在 8、9、11 月。 ⚫ 化债主线切换至稳增长，下半年地方债供给料将加速 上半年地方新增一 ...

证券研究报告：金融工程报告 研究所 分析师:肖承志 SAC 登记编号:S1340524090001 Email:xiaochengzhi@cnpsec.com 近期研究报告 《上交 AI 智能体表现亮眼， AlphaEvolve 生成代码反超人类——AI 动态汇总 20250707》 - 2025.07.08 《低估值高盈利，基本面表现占优— —中邮因子周报 20250706》 - 2025.07.07 《ETF 流入金融与 TMT，连板高度与涨 停家数限制下活跃资金处观望态势— —行业轮动周报 20250706》 - 2025.07.07 《"量化新规"或将平稳落地，双均线 法再现买点——微盘股指数周报 20250706》 - 2025.07.07 《谷歌推出 Gemini Robotics On- Device 大模型，快手开源 keye-VL 多 模态模型——AI 动态汇总 20250630》 - 2025.07.02 《基于宏观经济状态划分的 BL 模型与 ETF 实践》 - 2025.07.01 《基于大模型外部评价体系框架介 绍》 - 2025.06.30 《beta 风格显著，高波占优——中邮 ...

Investment Rating - The investment rating for the company is "Buy" and is maintained [1] Core Views - The company offers a range of products that support various forms of AI devices, addressing customer demands in the AI product market. Key products include high-performance audio solutions, mature AI lighting products, comprehensive haptic feedback solutions, high-performance mixed-signal chips, power management, and signal chain IC products. Notably, the audio component of Xiaomi's latest AI smart glasses utilizes the company's high-performance DSP digital SmartK audio amplifier, which enhances sound quality significantly [4] - The company has launched a new piezoelectric micro-pump liquid cooling driver, which is expected to open new growth opportunities in industrial interconnection and consumer electronics. This product boasts a cooling efficiency that is over three times better than passive solutions and is anticipated to enter mass production in the fourth quarter of this year [5] Financial Projections - The projected revenues for the company are as follows: 3.54 billion yuan in 2025, 4.27 billion yuan in 2026, and 5.12 billion yuan in 2027. The net profit attributable to the parent company is expected to be 410 million yuan in 2025, 579 million yuan in 2026, and 748 million yuan in 2027, maintaining a "Buy" rating [6] - The company's revenue growth rates are projected at 15.88% for 2024, 20.77% for 2025, 20.53% for 2026, and 20.02% for 2027. The net profit growth rates are forecasted at 399.68% for 2024, 60.93% for 2025, 41.04% for 2026, and 29.34% for 2027 [8][10]

Investment Rating - The report assigns a "Buy" rating for the company, marking its first coverage [1]. Core Insights - The company is actively positioning itself in the hvlp5 high-frequency copper foil market, benefiting from the rapid development of AI servers, which demand high-performance CCL [4]. - The 3C consumer electronics market is gradually recovering, driving overall sales growth for the company's products, which include electromagnetic shielding materials and insulation materials [5]. - The company has made strategic acquisitions, including a 51% stake in Weisi Dual-Link Technology, to enhance its self-sufficiency in key raw materials and optimize supply chain management [6]. - A planned acquisition of 100% of Deyou New Materials aims to create an integrated solution covering all aspects of electronic components, enhancing the company's product offerings [7]. Financial Projections - Revenue is projected to reach 375 million yuan in 2025, 488 million yuan in 2026, and 635 million yuan in 2027, with corresponding net profits of 109 million yuan, 148 million yuan, and 202 million yuan [8][10]. - The company’s PE ratios for 2025, 2026, and 2027 are estimated at 87, 64, and 47, respectively, indicating a favorable valuation trend [8][10].

Quantitative Models and Construction Methods Model Name: ML-Master - **Model Construction Idea**: The ML-Master model is designed to simulate human expert cognitive strategies, addressing the three major bottlenecks in existing AI4AI systems: low exploration efficiency, limited reasoning ability, and module fragmentation[12] - **Model Construction Process**: - **Balanced Multi-Trajectory Exploration Module**: Utilizes a parallelized Monte Carlo tree search to model the AI development process as a dynamic decision tree, with each node representing a potential solution state. This module dynamically allocates computing resources based on the potential value of 75 Kaggle task branches, avoiding local optima and improving medium difficulty task medal rates to 20.2%, 2.2 times the baseline method[13] - **Controllable Reasoning Module**: Overcomes the static decision limitations of large language models by filtering key code fragments, performance metrics, and cross-node insights from historical explorations through an adaptive memory mechanism. This ensures the reasoning process is based on verifiable execution feedback rather than probabilistic guesses, improving high difficulty task performance by 30%, significantly surpassing Microsoft's system's 18.7%[13] - **Adaptive Memory Mechanism**: Integrates the exploration and reasoning modules, creating a closed-loop evolution system. The results of code execution collected during the exploration phase are embedded into the reasoning model's "think" phase after intelligent filtering, and the optimized solutions from the reasoning output guide subsequent exploration paths. This dual empowerment allows ML-Master to reach the Grandmaster level among the top 259 global Kaggle participants after 900 machine hours of training, with solution quality improving by 120% over multiple iterations[15] - **Model Evaluation**: The ML-Master model demonstrates significant advantages in exploration efficiency, reasoning ability, and module integration, making it a leading system in the AI4AI field[12][13][15] Model Backtesting Results - **ML-Master**: - **Average Medal Rate**: 29.3%[12] - **Effective Submission Rate**: 93.3%[19] - **Task Performance**: 44.9% of tasks outperform more than half of human participants, with 17.3% of tasks winning gold medals[19] Quantitative Factors and Construction Methods Factor Name: OpenEvolve - **Factor Construction Idea**: OpenEvolve is designed to autonomously evolve code, achieving significant performance improvements in GPU kernel optimization tasks[22] - **Factor Construction Process**: - **Algorithm Layer**: Through 25 generations of evolutionary iterations, OpenEvolve autonomously discovered three key optimization strategies. For example, the SIMD optimization for Apple Silicon demonstrated the system's precise grasp of hardware characteristics, perfectly matching the hardware's SIMD width when processing 128-dimensional attention heads[23] - **Technical Implementation**: Utilizes a multi-model collaborative evolutionary architecture. The main model, Gemini-2.5-Flash, is responsible for rapid exploration, while the auxiliary model, Gemini-2.5-Pro, performs deep optimization. The system divides the Metal kernel function source code into evolvable blocks, retaining the integration code with the MLX framework unchanged, and evolves five subpopulations in parallel using the island model, with each generation having a population size of 25 individuals[24] - **Performance Evaluation**: The evaluation phase adopts a high-robustness design, including Metal command buffer protection, memory access violation handling, and exponential backoff retry mechanisms, ensuring the system can boldly attempt aggressive optimizations without worrying about crashes[25] - **Factor Evaluation**: OpenEvolve redefines the boundary of human-machine collaboration, demonstrating the potential for AI to autonomously explore optimization paths that require deep professional knowledge[22][23][24] Factor Backtesting Results - **OpenEvolve**: - **Average Performance Improvement**: 12.5% in decoding speed, 14.4% in pre-filling speed, and 10.4% in overall throughput[25] - **Peak Performance Improvement**: 106% in decoding speed for repetitive pattern generation tasks[25] - **Accuracy and Error Rate**: Maintains 100% numerical accuracy and zero GPU errors[25]

Industry Investment Rating - The industry investment rating is "Outperform" [1] Core Insights - The recent policy environment remains stable, with "good houses" supporting the new housing market. The introduction of construction standards for "good houses" is accelerating the upgrade of residential quality by developers. New projects that comply with these standards are selling well, benefiting from high usable area ratios, technological systems, and green certifications. As product differentiation becomes more apparent, the price gap between second-hand and new houses is expected to widen, with second-hand house prices under downward pressure while compliant new houses will continue to support the market [1][2][3] Summary by Sections 1. Industry Fundamentals Tracking - New housing transaction area in 30 major cities last week was 202.1 million square meters, with a cumulative new housing transaction area of 4,798.22 million square meters this year, down 2% year-on-year. The average transaction area over the past four weeks in these cities was 224.03 million square meters, down 10.3% year-on-year but up 7.2% month-on-month [2][11] - The average transaction area for first-tier cities over the past four weeks was 62.48 million square meters, down 10.6% year-on-year but up 7.3% month-on-month. For second-tier cities, it was 116.32 million square meters, down 4.6% year-on-year and up 8.1% month-on-month. Third-tier cities saw an average of 45.23 million square meters, down 21.8% year-on-year but up 5% month-on-month [11] - The available residential area in 14 cities last week was 7,934.97 million square meters, down 10.29% year-on-year and down 0.52% month-on-month. The de-stocking cycle for these cities is 16.93 months, with first-tier cities at 11.95 months [16] 2. Market Review - Last week, the A-share Shenwan first-level real estate index rose by 0.29%, while the CSI 300 index increased by 1.54%, indicating that the real estate index underperformed the CSI 300 by 1.25 percentage points. In the Hong Kong market, the Hang Seng property services and management index fell by 0.9%, while the Hang Seng composite index decreased by 0.94%, with the property services and management index outperforming the composite index by 0.04 percentage points [4][31]

证券研究报告：有色金属 | 公司点评报告 发布时间：2025-07-08 股票投资评级 资料来源：聚源，中邮证券研究所 公司基本情况 | 最新收盘价（元） | 8.84 | | --- | --- | | 总股本/流通股本（亿股）46.52 | / 41.30 | | 总市值/流通市值（亿元）411 | / 365 | | 52 周内最高/最低价 | 9.62 / 6.17 | | 资产负债率(%) | 52.7% | | 市盈率 | 9.21 | | 第一大股东 | 石河子市锦隆能源产业 | | 链有限公司 | | 研究所 分析师:李帅华 SAC 登记编号:S1340522060001 Email:lishuaihua@cnpsec.com 研究助理:杨丰源 SAC 登记编号:S1340124050015 Email:yangfengyuan@cnpsec.com 天山铝业(002532) 电解铝扩产落地，降本增效可期 l 公司拟扩产 20 万吨电解铝项目 根据公司第六届董事会第十五次会议决议，公司计划利用石河子 厂区东侧预留场地，采用国内先进的电解铝节能技术，对公司 140 万 吨电解铝产能进行绿色低碳能 ...