Investment Rating - The report assigns a Neutral rating to iFlytek, indicating that the valuation is fairly priced [1][9]. Core Insights - iFlytek's management is optimistic about the potential of its self-developed foundation models and generative AI applications, particularly in the education and healthcare sectors [1][4]. - The company plans to continuously enhance its Spark X1 reasoning model, with upgrades expected in July 2025, to better meet client demands across various industries [2]. - iFlytek's education business is expected to benefit from generative AI, with a focus on personalized learning and interactive courses, which will strengthen its market position [3]. - The enterprise business outlook remains positive, with customized foundation models being developed for various industries, enhancing competitiveness [4]. Summary by Sections Spark X1 Upgrades - iFlytek upgraded its Spark X1 reasoning model in April 2025 and plans further upgrades in July 2025, aiming to provide tailored AI solutions for different sectors [2]. AI Education Updates - The management is positive about the education segment, leveraging generative AI to boost AI learning machine shipments and enhance product competitiveness through personalized learning paths [3]. Enterprise Business Outlook - iFlytek is focused on providing total solutions for enterprises, including customized foundation models and application software, to facilitate the adoption of generative AI across various industries [4].

Investment Rating - The report emphasizes a quantitative investment approach that focuses on predicting stock profit growth rankings rather than specific numerical values, aiming for investment returns [1]. Core Insights - The idealized testing indicates that accurately predicting changes in ROE and holding stocks ranked highly can yield excess returns, validating the feasibility of this method [5]. - The introduction of the acceleration concept, which refers to changes in growth rates, can optimize models, enhance prediction accuracy, and reduce risks [1][7]. - The secondary trend extrapolation model, which considers profit growth and acceleration, outperforms linear extrapolation and analyst consensus in terms of prediction success rate (72%) and false positive rate (13%) [8]. - The "Growth Trend Resonance Stock Selection Strategy," which combines the optimized profit prediction model, analyst expectations, valuation, and cash flow factors, has shown excellent performance since 2009, consistently achieving excess returns [9]. - Incorporating machine learning methods, particularly tree models like XGBoost and LightGBM, significantly improves prediction accuracy, achieving a success rate of 85% and reducing the false positive rate to 4.7% [10][18]. Summary by Sections Traditional Economic Investment Approach - Traditional economic investment relies heavily on fundamental research, focusing on deep analysis of individual stocks to understand their business models and future profitability trends [2]. Quantitative Perspective on Economic Investment - The quantitative approach emphasizes breadth over depth, predicting relative rankings of stocks rather than specific profit growth amounts [3]. Validating Quantitative Investment Strategies - Idealized testing can validate the effectiveness of quantitative investment strategies by demonstrating that accurately predicting future ROE changes leads to superior net value performance [5]. Optimizing Profit Prediction Models - The introduction of acceleration in profit prediction models enhances accuracy and reduces risks associated with performance changes [8]. Application of Machine Learning in Profit Prediction - Machine learning models, particularly tree models, are preferred for their ability to handle multiple dimensions of data and capture non-linear relationships, leading to improved prediction accuracy [12][18]. Stock Selection Strategy - The strategy based on the difference Boots prediction factor has shown superior performance across various indices, indicating its effectiveness in stock selection [19][20].

Investment Rating - The report maintains a "Sell" rating on both Sangfor and Venustech, indicating a cautious outlook on the cybersecurity sector [2][17]. Core Insights - The management of Qi An Xin sees generative AI as a significant opportunity in the cybersecurity market, which could lead to increased demand for upgraded cybersecurity solutions due to the intensification of cyber attacks [4][9]. - The integration of generative AI is expected to enhance the accuracy and efficiency of cybersecurity software, particularly in risk detection and analysis, which could attract more clients [4][7]. - There is a noted trend of clients shifting their software budgets from functional tools to generative AI software, which is anticipated to drive average revenue per user (ARPU) and attract new paying customers [2][4]. Summary by Sections Company Profile - Qi An Xin, founded in 2014, specializes in cybersecurity, providing software and services to government and enterprise clients. The company has advanced capabilities in endpoint security, cloud security, threat intelligence, and situation awareness [3]. Generative AI Opportunities - Management believes that generative AI will trigger an increase in cyber attacks, creating a potential for upgrading cybersecurity solutions. The company plans to launch the QAX-GPT Security Robot in 2024, which will utilize its self-developed QAX security foundation model to assist security professionals [4][7]. Integrated Cybersecurity Framework - Establishing an integrated cybersecurity framework is deemed more critical than deploying standalone security products. Management emphasizes that platform-based solutions with broad coverage and interoperability will enhance user experience [8]. 2025 Performance Outlook - The management expresses low visibility on near-term demand recovery in the cybersecurity market due to macroeconomic constraints affecting client budgets, particularly in government sectors. However, there is a positive outlook for demand from certain enterprise clients, such as financial institutions and telecom operators [9]. The company is focusing on leveraging generative AI to enhance product offerings and reduce operational costs while improving cash flow [9].

7 July 2025 | 9:24AM HKT China Software: Product Tracker: AI agents upgraded, multi-modal AI models to unlock use cases; Software projects bidding review We reviewed software product launches form 1Q25 till date (Exhibit 1), and see the momentum of AI-native applications and software with AI features new launches remaining strong in 2Q25, especially agentic AI, multi-modal AI model, and model deployment. Vendors across AI foundation models and applications are migrating from basic integration to LLMs to mor ...

6 July 2025 | 2:08PM HKT Desay SV (002920.SZ): Mgmt. visit: Smart driving continues to grow; full stack solutions to accelerate deployment We visited Desay SV in Huizhou recently. Management remains positive on the smart driving trend, domain controllers growth, full stack solutions, and ongoing costs control and operation efficiency. We are positive on Desay SV's product expansion from smart cockpit to smart driving and automotive software, supporting them to better ride on the growing smart driving trend ...

Investment Rating - The report initiates coverage on Huaqin Technology with a "Buy" rating and includes it on the APAC Conviction List [1]. Core Insights - Huaqin's target price is raised to Rmb94, with expectations for net income to grow at a 27% CAGR from 2024 to 2027, driven by an increase in AI server shipments [1]. - The contribution of AI servers to Huaqin's revenues is projected to rise from 14% in 2024 to 45% by 2027 [1]. - The company is diversifying from consumer electronics to data centers, capitalizing on stronger market demand and healthier competition [1]. Revenue and Growth Projections - Huaqin's blended revenues are expected to grow at a 29% CAGR from 2024 to 2028 [5]. - The data center business, which includes general servers, switches, and AI servers, is anticipated to see AI servers grow at a 73% CAGR from 2024 to 2028, followed by switches at 67% CAGR and general servers at 8% CAGR [6]. - Data center revenues are expected to increase from 21% in 2024 to 51% by 2028 [6]. Earnings Revisions - Revenue estimates for 2025-2028 have been revised upwards by 1%, 14%, 21%, and 29% respectively, primarily due to higher expectations for AI servers and switches [9]. - Despite a reduction in net income estimates by 12% to 0% for 2025-2028, the target price is raised by 14.9% due to a higher target P/E multiple [9][13]. - The new target P/E multiple is set at 18.8x, reflecting a positive outlook on earnings growth driven by the data center business [13]. Financial Metrics - The report outlines expected revenues of Rmb147.2 billion in 2025, Rmb208.4 billion in 2026, and Rmb253.2 billion in 2027 [21]. - The net income for 2025 is projected at Rmb3.89 billion, increasing to Rmb5.07 billion in 2026 and Rmb5.91 billion in 2027 [21]. - The gross margin is expected to decline from 9.4% in 2025 to 8.4% in 2028, while the operating margin is projected to decrease from 2.6% in 2025 to 2.5% in 2028 [10][18].

Investment Rating - The report does not explicitly provide an investment rating for Longcheer (603341.SS) but indicates a positive outlook on the company's growth opportunities in AI PC and AI/AR glasses [1][2]. Core Insights - Longcheer is optimistic about the increasing penetration of AI PCs, which are expected to have higher average selling prices (ASP) due to enhanced specifications for computing power and storage [3]. - The company is expanding its business into AI/AR glasses, leveraging its strengths in lightweight design and customized solutions, with projected annual shipments in China expected to grow at a 56% CAGR from 2024 to 2030, reaching 7 million units by 2030 [7]. - Longcheer is also focusing on overseas expansion, with production facilities in Vietnam and India, targeting markets in the US, Europe, Japan, and Korea [8]. Summary by Sections AI PC Market - Management is positive about the rising AI PC penetration rate, viewing Longcheer as a key beneficiary in the supply chain due to its R&D efficiency and automated production lines [3]. - The company has already penetrated local brands and aims to collaborate with global-tier brand clients [3]. AI/AR Glasses Market - Longcheer is actively providing PCBA and accessory services for AI/AR glasses, with a focus on enhancing user experience through lightweight solutions and proofing/simulation [7]. - The management anticipates significant growth in the AI/AR glasses market, driven by affordability and increased use cases [7]. Overseas Expansion - Longcheer has established production sites in China, Vietnam, and India, with plans to expand capacity in Vietnam to mitigate concentration risks amid trade tensions [8]. - The company is building a Manage Sales & Marketing (MSM) framework to enhance product development efficiency and service quality for its overseas markets [8].

Investment Rating - The report does not explicitly state an investment rating for the solar industry, but it suggests that polysilicon and module names could exhibit the highest potential upside within the sector due to inexpensive valuation and limited downside risks [4]. Core Insights - The solar glass manufacturers in China are beginning to cut production due to persistently weakening demand, with estimates suggesting a reasonable production cut of 10-20%, leading to an effective monthly production of around 45-50GW [2]. - The government is expected to push for capacity cuts across the solar supply chain, starting with the polysilicon segment, and discussions are ongoing regarding acquiring smaller players [3]. - There is a stronger commitment from the government to tackle overcapacity, with expectations of more policy support to phase out obsolete capacity and deter price competition, despite lingering fundamental pressures in the second half of 2025 [4]. Summary by Sections Production and Demand - Solar glass manufacturers are cutting production due to weak demand, with a potential cut of 10-20% rather than the 30% estimated by some media [2]. - The effective production capacity could be around 45-50GW monthly, with the possibility of resuming operations once prices rebound [2]. Government Policies - The government is anticipated to implement capacity cuts in the solar supply chain, particularly in the polysilicon segment, and is discussing measures to acquire smaller players [3]. - There is speculation about policies to curb excess capacity, with a belief that the government is determined to reduce involution competition [4]. Market Sentiment and Valuation - Despite expected fundamental pressures in the second half of 2025, market sentiment may improve in the long term due to better supply-demand dynamics [4]. - Polysilicon and module names are highlighted as having the highest potential upside within the sector, attributed to inexpensive valuations and limited downside risks [4].

Investment Rating - The report does not explicitly state an investment rating for the aluminum industry but provides insights into demand trends and expectations for growth in specific sectors [1][3]. Core Insights - Aluminum demand in the solar power industry is expected to increase by 5-8% year-over-year (YoY) in 2025, despite a recent decline in demand since mid-May [1][4]. - Overall aluminum demand in China is projected to grow by 3-4% YoY in 2025, with a weaker growth forecast for the second half of the year compared to the first half [3]. - The demand for aluminum building profiles is anticipated to decrease by 8-10% YoY in 2025, showing a slight improvement from a 10% YoY decrease in 2024 [5]. Summary by Sections Aluminum Production - The weekly semi-aluminum products output was approximately 610kt in early July 2025, reflecting a 7.5% YoY increase, while the aluminum profile output was around 190kt, down 6% YoY [2]. - The weekly aluminum foil sheet output was about 370kt, showing a slight decline of 1% YoY [2]. Demand Trends - The aluminum demand from the solar power industry is primarily for industrial profiles and foil sheets, with a noted decrease in demand since mid-May, which has slowed into July [1][4]. - The apparent consumption of aluminum is expected to remain weak in July but may rebound in September and October, with potential growth in the automotive and electricity sectors [3]. Inventory Levels - Current inventory levels for aluminum profile mills are around 520kt, which is a 20% decrease YoY, with raw materials and finished goods inventories down 18% and 21% YoY, respectively [6].

Investment Rating - The report does not explicitly provide an investment rating for the humanoid robot industry Core Insights - The humanoid robot transmission system is characterized by a mix of rotary and linear joints, with domestic companies favoring rotary modules while international firms like Tesla prefer linear actuators. The design trends indicate a shift towards hybrid systems for dexterous hands, driven by increasing demands for high degrees of freedom and precision [1][5][9] Summary by Sections Industry Overview - The humanoid robot industry is witnessing a convergence of technology standards, transmission schemes, and processing techniques, although a clear competitive landscape has yet to emerge. Continuous tracking is necessary to understand the evolving dynamics [1][2] Key Components and Cost Structure - Core components of humanoid robot transmission systems include lead screws, reducers, and bearings, which collectively account for over 30% of the hardware costs. The cost structure is still in flux as technology converges [2][3] Domestic vs. International Trends - Domestic companies focus on light-load industrial applications and consumer markets, utilizing mature rotary actuators. In contrast, international players like Tesla emphasize high-load industrial scenarios, favoring lead screw designs [6][8] Patent Activity - As of 2024, patent applications in China show a significant focus on perception systems and linear actuators, indicating substantial investment in these areas. Rotary actuators and dexterous hands have fewer related patents [12] Future Development Directions - The future of humanoid robots will revolve around cost reduction, structural design optimization, and material selection. Innovations in manufacturing processes, such as improved heat treatment for harmonic reducers, are critical for enhancing performance [13][15] Competitive Landscape - The competitive landscape for reducers is changing, with latecomers needing to establish customer loyalty and possess strong development capabilities to succeed. The lightweight material advancements, such as PEEK, are noteworthy for their potential to reduce energy consumption and improve efficiency [16][17] Application Scenarios - Different application scenarios dictate the choice of transmission schemes, with domestic firms focusing on simpler designs for home use, while international firms prioritize complex, high-load industrial applications [6][8] Investment Considerations - Investors should focus on new technologies, materials, and structural designs that could lead to non-linear cost changes and reshape the industry landscape. Companies that have converged on technology routes and possess competitive advantages are likely to outperform [21]