Core Insights - The report indicates that major overseas tech companies have reported better-than-expected earnings for Q3 2025, with significant revenue growth driven by cloud services and AI investments [1][2][3] - Capital expenditures (CapEx) for these companies continue to rise sharply, reflecting a focus on AI infrastructure and cloud capabilities, which has become a central concern for investors [2][58] Company Summaries Microsoft - Microsoft reported Q1 FY26 revenue of $77.67 billion, a year-on-year increase of 18%, slightly exceeding market expectations [11][12] - The intelligent cloud segment generated $30.9 billion in revenue, up 28% year-on-year, with Azure cloud services growing at 40% [12][15] - Capital expenditures reached $34.9 billion, a significant increase of 74.5% year-on-year, primarily for AI and infrastructure investments [14][15] Meta - Meta's Q3 2025 revenue was $51.24 billion, a 26.25% increase year-on-year, surpassing both guidance and market expectations [19][21] - The company faced a net profit decline of 82.73% due to a one-time tax asset impairment, but adjusted net profit was $18.6 billion [19][20] - Capital expenditures rose to $19.37 billion, primarily for servers and data centers, exceeding expectations [23][27] Google - Google reported Q3 2025 revenue of $102.35 billion, a 15.95% increase year-on-year, with net profit rising 32.99% [33][36] - The Google Cloud segment achieved $15.16 billion in revenue, growing 33.51% year-on-year, driven by strong demand for AI products [36][42] - Capital expenditures for Q3 were $23.95 billion, with an upward revision of the annual CapEx guidance to $91-93 billion [42] Amazon - Amazon's Q3 2025 revenue reached $180.17 billion, a 13% increase year-on-year, with net profit up 38% [43][46] - AWS revenue was $33.01 billion, marking a 20% year-on-year growth, the highest quarterly growth rate since 2023 [46][52] - Capital expenditures for Q3 were $34.2 billion, a 61% increase year-on-year, with expectations for continued growth in FY2026 [52][56] Industry Trends - The report highlights a trend of increasing capital expenditures across major tech firms, indicating a competitive "arms race" in AI and cloud infrastructure [58][61] - The demand for AI capabilities is driving significant revenue growth in cloud services, with all major players reporting double-digit growth in this segment [57][58] - Investors are closely monitoring the impact of rising CapEx on profit margins and return on investment (ROIC), as increased spending on AI infrastructure may pressure short-term profitability [61][62]

Macro and Strategy - The report discusses the ongoing expansion and diversification of public REITs in China, highlighting the inclusion of various asset types and industries, with a projected market size of 2.3 to 3.8 trillion yuan, indicating a potential 10-16 times expansion from current levels [7][8][10] - The average dividend yield of public REITs from 2022 to 2025 is 5.73%, surpassing the average yield of the CSI Dividend Index at 5.52%, showcasing their attractiveness as a stable income asset [8][9] - Public REITs are characterized by a dual return structure comprising dividend income and asset appreciation, with a notable annualized return of 23.66% over the past year [9][10] Industry and Company - The Chinese duty-free industry is entering a new cycle, with Hainan's duty-free sales showing signs of recovery, driven by policy support and improving consumer confidence, with sales growth of 3%, 13%, and 27% from September to November 2025 [17][18] - The report emphasizes the importance of policy optimization in the duty-free sector, with recent expansions in both offshore and onshore duty-free policies, enhancing consumer access and convenience [18][19] - The report identifies key players in the duty-free market, such as China Duty Free Group, which holds a 78% market share, and highlights the strategic importance of airport channels for future growth [20][21] Automotive Industry - The report outlines the advancements in smart driving technology, with companies like Tesla and Huawei leading the way in achieving Level 4 automation through end-to-end algorithms [24][25] - The penetration rate of smart driving is expected to see significant growth, with projections indicating an increase from 11.3% to 26.3% for highway NOA and from 6.1% to 10.9% for urban NOA by 2025 [25] - The global market for Robotaxi is projected to reach nearly 10 trillion yuan, with companies like Waymo and Apollo at the forefront of commercialization efforts [25][26] Non-Banking Industry - The report highlights the importance of the second pillar of the pension system in China, focusing on enterprise and occupational pensions, which are expected to grow at an annualized rate of 8%, outpacing nominal GDP growth [26][27] - The investment behavior of pension funds is shifting towards a "barbell" strategy, balancing stable cash flow assets with high-growth sectors, indicating a significant increase in equity allocations [27][28]

- The report discusses the methodology for calculating dividend points in stock indices, which is crucial for accurately estimating the premium or discount in stock index futures contracts. The calculation involves the following formula: **Dividend Points = (Sum of Dividend Amounts / Total Market Value) × Component Stock Weight × Index Closing Price** This formula considers only the component stocks with ex-dividend dates between the current date and the futures contract expiration date [42][45] - The weight of component stocks is dynamically adjusted to reflect daily changes in stock prices. The formula for calculating the weight is: **$W_{n,t} = \frac{w_{i0} \times (1 + r_{n})}{\sum_{i=1}^{N} w_{i0} \times (1 + r_{n})}$** Here, $w_{i0}$ is the weight of stock $n$ on the last disclosed date, and $r_{n}$ is the non-adjusted return of stock $n$ from the last disclosed date to the current date [46] - The estimation of dividend amounts is based on the product of net profit and dividend payout ratio. If the company has not disclosed its dividend amount, the net profit is predicted using historical profit distribution patterns, and the dividend payout ratio is estimated using historical averages. The formula is: **Dividend Amount = Net Profit × Dividend Payout Ratio** For companies with stable profit distribution, historical patterns are used, while for others, the previous year's profit is used as a proxy [48][51][52] - The ex-dividend date is predicted using a linear extrapolation method based on the stability of historical intervals between announcement dates and ex-dividend dates. If no historical data is available, default dates are assigned based on typical market behavior [52][57] - The accuracy of the dividend point estimation model is evaluated by comparing predicted dividend points with actual dividend points. For indices like the SSE 50 and CSI 300, the model achieves high accuracy with errors around 5 points, while for the CSI 500, the error is slightly larger, around 10 points [58][62]

证券研究报告 | 2025年12月03日 固收+系列报告之六 固收+的新选择：公募 REITs——扩围下的新机遇 政策红利持续释放，横向扩容，纵向深化。公募 REITs 自 2020 年启动以来， 国家发改委会同证监会持续推动公募 REITs 市场扩围扩容，实现从传统基建 向多元资产的破圈，初期将仓储物流、收费公路、市政设施、产业园区等纳 入发行范围；后续又逐步增加了清洁能源、数据中心、保障性租赁住房、水 利设施、文化旅游、消费基础设施等行业领域和资产类型。目前，发行范围 已涵盖了 12 大行业的 52 个资产类型，其中 10 个行业领域的 18 个资产类型 已经实现了首单发行上市。近期，国家发改委正积极推动基础设施 REITs 进 一步扩围至城市更新设施、商业办公设施等更多资产类型。从市值占比来看， 当前REITs 底层资产类型中，交通占比 26%，消费占比 19%，产业园占比 17% ， 三大板块合计占比达 62%，成为市场绝对主力。随着 REITs 扩围扩容，预计 市值规模为 2.3-3.8 万亿元，跟当前相比，REITs 市场规模还有 10-16 倍扩 容空间。 公募 REITs 是高分红的类固收权益 ...

证券研究报告 | 2025年12月03日 金融工程日报 沪指震荡下挫，AI 应用、锂电池题材领跌 市场表现：20251203 大部分指数处于下跌状态，规模指数中沪深 300 指数表 现较好，板块指数中北证 50 指数表现较好，风格指数中中证 500 价值指数 表现较好。交通运输、有色金属、煤炭、综合金融、家电行业表现较好，传 媒、计算机、房地产、商贸零售、综合行业表现较差。培育钻石、超硬材料、 锗镓锑墨、铝产业、铝空气电池等概念表现较好，WEB3.0、百度平台、小红 书平台、RCS、ChatGPT 等概念表现较差。 市场情绪：20251203 收盘时有 53 只股票涨停，有 16 只股票跌停。昨日涨停 股票今日收盘收益为 1.17%，昨日跌停股票今日收盘收益为-5.91%。今日封 板率 61%，较前日下降 7%，连板率 24%，较前日下降 0%。 市场资金流向：截至 20251202 两融余额为 24865 亿元，其中融资余额 24689 亿元，融券余额 176 亿元。两融余额占流通市值比重为 2.6%，两融交易占市 场成交额比重为 9.7%。 折溢价：20251202 当日 ETF 溢价较多的是 G60 创 ...

1. Report Industry Investment Rating - Not provided in the given content 2. Core Viewpoints - Policy dividends for public REITs are continuously released, with both horizontal expansion and vertical deepening. The market is expected to expand 10 - 16 times, reaching a scale of 2.3 - 3.8 trillion yuan [1][26]. - Public REITs are high - dividend, fixed - income - like equity assets. Their average annual dividend rate in the past four years was 5.73%, higher than the 5.52% of the CSI Dividend Index, and they have a certain allocation advantage compared to stocks and bonds [2][34]. - The returns of public REITs have both bond and equity attributes. The overall annualized return of the entire market's REITs in the past one - year, three - year, and since establishment are 23.66%, 3.24%, and 7.64% respectively. The longer the investment time, the higher the proportion of dividend income [3][39]. - Public REITs are a stable allocation choice in an alternating and volatile market. They are weakly or extremely weakly correlated with mainstream assets, can hedge against single - asset volatility risks, and fill the gap of medium - risk, stable - return assets between stocks and bonds [4]. 3. Summary by Relevant Catalogs Policy Evolution: From Pilot Breakthrough to Full - scale Expansion - REITs in China have transformed from private to public and from debt - like to equity - like. Policies from the central to local levels have promoted the implementation of public REITs, achieving market expansion, capacity increase, and deepening [12]. - Since 2008, central government departments have repeatedly issued documents to encourage REITs pilot projects. In 2020, the first batch of public REITs were launched, and in 2024, they entered the stage of normalized issuance [13]. - The scope of underlying assets for public REITs has expanded from traditional infrastructure to multiple asset types, covering 12 major industries and 52 asset types, with 18 asset types in 10 industries achieving the first - single issuance and listing [14]. Policy Dividends Continuously Released: Horizontal Expansion and Vertical Deepening - The underlying asset types of public REITs include property - based REITs and franchise - based REITs, with different investment returns and risk characteristics [15][17]. - As of November 21, 2025, 77 public REITs have been listed. In terms of market value, transportation, consumption, and industrial parks are the main forces, accounting for 62% in total [17]. Market Size Outlook: A Trillion - yuan Blue Ocean, Ready to Take Off - Globally, REITs have become an independent asset class. By referring to the US and Japanese markets, the scale of China's REITs market is estimated to be 2.3 - 3.8 trillion yuan, with 10 - 16 times expansion space compared to the current market value [22][26]. Investment Value: High - Dividend, Fixed - Income - like Equity Assets - Public REITs are both stock - like and bond - like. They are required to distribute over 90% of the annual distributable amount in cash, with a higher average annual dividend rate in the past four years than the CSI Dividend Index [33][34]. - The dividend rate spread between public REITs and the ten - year Treasury bond yield has been between 300 - 400BP in recent years, having a certain allocation advantage [34]. Return Decomposition: Dividend Income and Capital Gains - The investment returns of public REITs can be decomposed into dividend income and asset appreciation income. The overall average total return of listed public REITs reaches 17.21%, with significant category differentiation [36]. - The annualized returns of the entire market's REITs in the past one - year, three - year, and since establishment are 23.66%, 3.24%, and 7.64% respectively. The longer the investment time in the US, the higher the proportion of dividend income [39][40]. Asset Comparison: Medium Risk - Return, Low Correlation with Other Assets - Since 2025, the CSI REITs Index has been weakly or extremely weakly correlated with the CSI 300, ten - year Treasury bonds, gold, and CSI Dividend Stocks, with correlation coefficients of - 0.07, 0.14, 0.21, and 0.17 respectively [41]. - REITs can hedge against single - asset volatility risks and fill the gap of medium - risk, stable - return assets between stocks and bonds, meeting the needs of medium - and long - term funds [43]. Investment Methods: Primary Market Subscription and Secondary Market Trading - The investment in REITs can be made in the primary market (by subscribing at the issuance or expansion stage) or the secondary market (by trading on the exchange after listing). Currently, institutional investors are the main participants [50]. Primary Market: Centered on Dividends and Listing Premiums - Primary market subscribers include strategic investors, offline investors, and public investors, with different requirements and characteristics [53][56]. - The subscription price is determined by offline investors' inquiries. Strategic investors' original equity holders and their affiliates must subscribe at least 20% of the shares, and at least 70% of the remaining shares are allocated to offline investors [53]. - The short - term income of primary market subscribers comes from the difference between the subscription price and the secondary market trading price. In 2025, the new - listed public REITs had a significant first - day increase [58]. Secondary Market: Coexistence of Return Elasticity and Risks - The secondary market performance of public REITs has gone through six stages since 2021, affected by factors such as market rules, policies, and the fundamentals of underlying assets [62]. - After the recent adjustment, the allocation value of REITs has increased, and December 2025 is expected to be an important allocation window [66].

Core Insights - The report emphasizes the growing importance of AI in asset allocation, particularly in stock price prediction, highlighting the capabilities of AI models like TrendIQ in addressing the limitations of traditional machine learning approaches [3][4][10]. Group 1: AI in Stock Price Prediction - The introduction of AI large models has significantly improved the ability to predict stock prices by effectively collecting and analyzing unstructured information, which traditional models struggled with [3][4]. - TrendIQ is presented as a mature financial asset price prediction platform that offers both local and web-based deployment options, catering to different user needs [4][10]. - The report discusses the evolution of predictive models from LSTM to more advanced architectures like Transformers, which provide better handling of complex financial data and improve predictive accuracy [5][10]. Group 2: Model Mechanisms and Limitations - LSTM has been the preferred model for stock price prediction due to its ability to handle non-linear and time-series data, but it has limitations such as single modality and weak interpretability [6][7]. - The report outlines the integration of LSTM with other models like XGBoost and deep reinforcement learning to enhance predictive capabilities, addressing some of LSTM's shortcomings [6][10]. - The emergence of Transformer architecture is noted for its advantages in global context awareness and the ability to perform zero-shot and few-shot learning, which enhances its applicability in financial predictions [8][10]. Group 3: TrendIQ Implementation - The report details the implementation of TrendIQ, which includes a complete framework for data preparation, model training, and user interaction through a web application [12][20]. - The training process involves collecting historical stock data, preprocessing it, and training the LSTM model, ensuring that users can make predictions through a user-friendly interface [12][20]. - The app integrates various components, including real-time data fetching and prediction functionalities, allowing users to interactively engage with the predictive model [20][28]. Group 4: Future Directions - The report anticipates that future developments in AI stock prediction will focus on multi-modal integration, combining visual data from candlestick charts with textual analysis from financial news and numerical data from price sequences [39][40]. - The potential for real-time knowledge integration into predictive models is highlighted, suggesting that future AI models will be able to adapt to new information dynamically, improving their robustness and accuracy [40][41].

Investment Rating - The report maintains an "Outperform" rating for the non-bank financial and insurance sector [3]. Core Insights - The second pillar of the pension system, primarily composed of enterprise annuities and occupational annuities, is steadily progressing and is crucial for addressing the challenges of an aging population and enhancing national pension security [1][10]. - The growth rate of annuities is expected to maintain an annual compound growth rate of 8%, which is higher than the projected nominal GDP growth [1][2]. - The occupational annuity system has achieved full coverage due to reforms in public sector pension insurance, while enterprise annuities are expanding from state-owned enterprises to stable private enterprises [2][10]. - The investment strategy for annuity funds is shifting towards a "barbell" structure, balancing between high-dividend, low-volatility assets and investments in sectors with growth potential, such as technology and manufacturing [2][10]. Summary by Sections Current State of Pension System - The pension system in China has evolved into a multi-pillar structure, with the first pillar being the basic pension insurance, the second pillar consisting of enterprise and occupational annuities, and the third pillar being personal and commercial pension insurance [11][16]. - As of late 2024, the basic pension insurance has accumulated a surplus of 8.72 trillion yuan, with a year-on-year growth rate of 11.55% [13][15]. Enterprise Annuities - The enterprise annuity market has seen significant growth, with the investment operation scale reaching 3.64 trillion yuan by 2024, reflecting a year-on-year growth of 14.11% [13][15]. - The number of enterprises establishing annuities peaked in 2014 and has since seen a decline in growth rate, indicating a potential stagnation in new enterprise participation [31][33]. Occupational Annuities - Occupational annuities have rapidly developed due to their mandatory nature, with a projected investment operation scale of 3.11 trillion yuan by 2024, growing at a rate of 21.48% [13][15]. - The occupational annuity system is characterized by a centralized management model, which enhances operational efficiency and market influence [2][10]. Investment Strategies - Annuity funds are increasingly entering the market, with a focus on optimizing asset allocation to achieve long-term value growth amidst challenges such as declining interest rates and reduced supply of non-standard assets [2][10]. - The report anticipates that the equity allocation in enterprise annuities will rise from the current 10%-15% to 20%-25%, potentially adding around 500 billion yuan in equity investments [2][10].

Investment Rating - The report maintains an "Outperform" rating for the industry [1] Core Insights - The development of intelligent driving is an inevitable trend supported by national strategies and policies, leading to multi-dimensional improvements in society and industry [2] - Tesla and Huawei are leading the breakthrough in L4 autonomous driving with their end-to-end algorithms, significantly enhancing performance and capabilities [2] - The Robotaxi global market is projected to reach nearly 10 trillion, with ongoing commercialization efforts [2] Summary by Sections 01 Intelligent Driving Regulations: Gradual Policy Implementation - Domestic and international policies are progressively supporting the automation of driving applications, with various local governments exploring intelligent driving scenarios [6][7] 02 High-End Intelligent Driving: Tesla and Huawei's End-to-End Technology - Tesla's FSD V12 and Huawei's ADS 3.0 are leading advancements in L4 capabilities, with significant improvements in algorithm performance and urban coverage [2][20] 03 Intelligent Driving Equality: 2025 Penetration Rate Inflection Point - The penetration rates for highway NOA are expected to grow from 11.3% in 2024 to 39.0% in 2025, while urban NOA is projected to increase from 6.1% to 9.6% [41] - The high-end intelligent driving market is anticipated to reach 23,866 billion by 2025, doubling from 2024 due to increased penetration and market expansion [41] 04 Industry Chain and Component Manufacturer Analysis - BYD's "Tian Shen Zhi Yan" system is set to penetrate the mid-range market, with plans to offer intelligent driving features in vehicles priced below 100,000 [25][29] 05 Robotaxi: The Best Commercialization Scenario for Intelligent Driving - Companies like Waymo and Apollo are leading in the Robotaxi sector, with PONY AI achieving operational cost balance and WeRide aiming for a fleet of 100,000 by 2030 [2]

Core Insights - The report emphasizes the growing importance of AI in asset allocation, particularly in stock price prediction, highlighting the capabilities of AI models like TrendIQ in providing effective analysis and predictions [3][4][10] - It discusses the evolution of predictive models from traditional LSTM to more advanced architectures like Transformers, which offer improved performance in handling complex financial data [39][40] Group 1: AI in Stock Price Prediction - The introduction of AI large models has significantly enhanced the ability to predict stock prices by addressing the limitations of traditional machine learning models, particularly in processing unstructured data [3][4] - TrendIQ is presented as a mature platform that supports both local and web-based deployment, offering advantages in security, speed, and user-friendliness [4][12] Group 2: Model Evolution and Capabilities - The report outlines the transition from LSTM to Transformer architectures, noting that Transformers provide global context awareness and better handling of long-term dependencies, which are crucial for financial predictions [8][39] - It highlights the limitations of LSTM, such as its single modality and weaker interpretability, which can pose risks in a regulated financial environment [7][10] Group 3: TrendIQ Implementation - The implementation of TrendIQ involves a structured process including data preparation, model training, and user interaction through a web application, ensuring a seamless prediction experience [12][20] - The report details the specific Python scripts used in the TrendIQ framework, emphasizing the importance of each component in the overall predictive process [12][18][20] Group 4: Future Directions - Future advancements in AI stock prediction are expected to focus on multi-modal integration, combining visual data from candlestick charts with textual analysis from financial news, enhancing predictive accuracy [40][41] - The report suggests that real-time knowledge integration will further improve the robustness of AI models, allowing them to adapt to changing market conditions dynamically [40][41]