AI专题·从FOMO CapEx到ROI CapEx AI模型迭代聚焦工程能力，AI应用落地锚定高ROI场景 西南证券研究院 海外研究团队 2026年1月 核心观点 1 海外AI投入面临现金流压力，AI投资从FOMO CapEx转向ROI CapEx。2024-2025年，海外科技大厂资本开支高增，AI初创企业 投入力度加大，且未来开支预期进一步上调，当前行业普遍面临现金流压力，从而促使海外科技厂商寻求多种数据中心建设方式和融 资手段缓解压力，AI独角兽IPO进程也有望提速。2024年，AI大基建之初，部分海外云厂商表示"投资不足的风险远远大于投资过度 的风险"，AI投资伴随着FOMO情绪；2025年，海外云厂商均强调"云服务供不应求"、"根据需求信号扩展数据中心"、同时愈 发关注AI应用的商业化变现，AI投资逐步从FOMO CapEx向ROI CapEx转变。 数据中心面临电力容量限制，算力集群日益强调每瓦特Tokens产出效率。通常，海外云厂商在规划数据中心之初，需率先确定能够 提供给数据中心的电力容量，再针对数据中心内部的IT设备进行配置。因此，在电力容量限制下，海外云厂商均强调最大化每瓦特下 的Tok ...

家电行业2026年投资策略 基数承压，希冀仍存 西南证券研究院 家电研究团队 2026年1月 核 心 观 点 回顾2025年： 2025年初截至12月31日，申万家电指数指数上涨9.1%，在申万行业中涨幅排名第24。 从2025年上半年来看，国补政策几乎无缝延续且加大了补贴范围，线上市场的国补便捷性增强，此时申万 家电指数整体走势与沪深300较为一致。从2025年下半年来看，大盘跟随"科技+红利"持续走强，而家电指 数则由于内销国补退坡以及外销关税不确定性，开始跑输大盘。 从基本面来看，内销国补政策持续，但边际效果递减已经定局，国内需求的提前透支，以及地产及消费习 惯变化导致的需求减少，我们认为内销目前需要一个较长的恢复期；外销关税政策变化除带来了不确定性压制 PE外，出口型家电企业积极进行海外产能建设，产能爬坡期也压制了EPS，因此2025年下半年开始，家电行业 基本面整体偏弱。 从风格方面来看，家电行业中白电龙头主要表现出红利属性，但因为其分红能力依赖于盈利稳定性，若企 业面临市场竞争加剧、成本上升等问题，分红政策可能调整，因此白电在今年并未受到资金青睐；家电零部件 则更具备科技属性，在AI服务器、机器人 ...

ooo[Table_ReportInfo] 2026 年 01 月 12 日 证券研究报告•宏观深度报告 宏观深度 新经济时代的"动态革新" ——我国产业升级的赋能机制研究 摘要 西南证券研究院 、 [Table_Author] 分析师：叶凡 执业证号：S1250520060001 电话：010-57631106 邮箱：yefan@swsc.com.cn 分析师：刘彦宏 执业证号：S1250523030002 电话：010-55758502 邮箱：liuyanhong@swsc.com.cn 联系人：徐小然 邮箱：xuxr@swsc.com.cn 相关研究 请务必阅读正文后的重要声明部分 [Table_Summary] 理论框架："新经济"重视技术和知识产权。该概念最早出现在 2016 年中国 《政府工作报告》中，主张发展高技术产业和现代服务业等新兴产业。从理论 看，索洛模型认为经济处于稳态时，人均产出增长只能由外生技术进步所驱动。 而内生增长理论认为，由于知识、技术创新等要素具有正外部性，且存在规模 报酬递增现象，由此可实现内生的持续增长，打破了资本边际收益递减的假设。 实证检验：技术创新、数字经济与高质量 ...

交运行业2026年投资策略 航空盈利修复可期，航运绿色转型提速 西南证券研究院 交通运输研究团队 2026年1月 核心观点 1 航空：在美联储降息推动下，人民币兑美元汇率在2026年有望进一步提升，为航司带来汇兑收益，同时国 际原油价格在2026年继续下行，航司的燃油成本压力有望缓解。机队供给方面，发动机存在问题等原因导 致我国航司的运力扩张有限，需求方面，经济增长将驱动航空出行需求结构性增长，我们持续看好后续票价 及航司利润表现。重点推荐：南方航空、春秋航空、华夏航空。 公路：我国公路行业已进入成熟期，2024年我国高速公路总里程已超过美国居世界第一，随着公路建设投 资放缓和到期压力凸显，打破原有收费期限的新版收费公路管理条例或将出台。多种因素下，未来行业将呈 现改扩建、并购、业务多元化三大发展趋势。重点推荐：中原高速。 航运绿色甲醇：全球航运业向零排放能源转型势在必行，当前绿色甲醇整体技术成熟度较好，且减碳表现良 好成为市场一种主流选择。截至2025年11月，MI数据库追踪全球252个可再生甲醇项目，预计到2030年总 装机容量为4510万吨。到2030年，所有电制甲醇项目的总装机容量预计为2180万吨， ...

Domestic Developments - The People's Bank of China (PBOC) maintains a "moderately loose" monetary policy for 2026, focusing on precision and coordination to support economic growth and structural transformation[10] - The Ministry of Industry and Information Technology (MIIT) has launched an action plan for the integration of industrial internet and artificial intelligence, marking a new phase in industrial intelligence development[12] - A green consumption promotion plan was issued, aiming to stimulate domestic demand and support the transition to a circular economy[9] International Developments - The U.S. ISM Manufacturing PMI fell to 47.9 in December, marking the largest contraction since 2024, with inventory reduction being a major drag[18] - The Eurozone's harmonized CPI fell to 2% in December, indicating a return to target levels, while core inflation remains resilient[20] - The U.S. has initiated a global sale of Venezuelan oil, which may disrupt global energy trade and escalate geopolitical tensions[22] Market Trends - Brent crude oil prices increased by 0.94% week-on-week, while iron ore and copper prices rose by 1.88% and 3.60%, respectively[24] - Domestic real estate sales saw a significant decline of 62% week-on-week, indicating ongoing challenges in the sector[24] - The DXI index for storage DRAM prices rose by 7.45% week-on-week, reflecting positive trends in emerging industries[33]

Core Insights - The report emphasizes the importance of capitalizing on emotional consumption trends and reassessing cyclical value in the light of the 2026 investment strategy for the light industry sector [1][3]. 2025 Sector Review - In 2025, the light industry sector experienced relatively flat performance, with traditional cyclical and manufacturing companies facing valuation pressure. However, packaging and printing sectors benefited from price increases and cross-industry transformations, leading to better stock performance [4]. - The export sector showed some differentiation due to tariff policy disruptions, with companies that had balanced production capacity and strong demand performing better. The personal care sector saw excess returns in the first half of the year but faced valuation digestion in the second half due to intensified e-commerce competition [4][5]. - The report suggests a dual focus for stock selection in 2026: on one hand, to pay attention to undervalued cyclical assets for valuation recovery; on the other hand, to balance the valuation and growth potential of new consumption and export sectors [4]. Stock Selection Strategy - The report recommends four main lines for stock selection: 1. Gradually focus on undervalued cyclical stocks, particularly in the paper sector, which is expected to see price increases driven by seasonal demand and low channel inventory [4]. 2. Maintain a high allocation to export stocks with strong demand resilience and manufacturing capabilities, especially those less affected by tariffs [4]. 3. Invest in high-quality domestic personal care brands benefiting from product structure optimization and channel expansion [4]. 4. Explore new consumption trends in categories like AI glasses, new tobacco products, pet supplies, and trendy toys, which are expected to see significant growth [4]. Recommended Stocks - The report lists several recommended stocks, including: - Sun Paper Industry (002078.SZ) - Bohui Paper Industry (600966.SZ) - Weigao Medical (300888.SZ) - Baiya Co., Ltd. (003006.SZ) - Nobon Co., Ltd. (603238.SH) - Yiyi Co., Ltd. (001206.SZ) - Mengbaihe (603313.SH) - Gujia Home (603816.SH) [4]. 2025 Sector Performance Data - As of December 31, 2025, the SW light industry manufacturing sector had an overall increase of 20.1%, outperforming the Shanghai Composite Index by 1.7 percentage points. The packaging and printing sector performed particularly well with a 35.4% increase [12]. - The report highlights that the packaging sector benefited from price increases and cross-industry transformations, while the home and entertainment sectors also saw significant gains [12][14]. Export Sector Insights - The report notes that from November 2025, the U.S. reduced tariffs on Chinese imports to 20%, leading to a gradual recovery in orders. The fluctuations in tariff policies had previously caused delays in orders from U.S. buyers [76]. - The report indicates that the export sector is expected to see a return to competitive pricing against ASEAN countries following the tariff adjustments, which may accelerate industry consolidation [76][81]. Personal Care Sector Trends - The personal care sector is experiencing product structure upgrades and channel benefits, with brands focusing on high-demand segments such as oral care and women's hygiene products [31][50]. - The report forecasts that the market for women's hygiene products will reach 1079.6 billion yuan in 2025, with a compound annual growth rate (CAGR) of 3.0% from 2025 to 2029 [50][51]. Baby Care Market Dynamics - The baby care market is projected to grow at a CAGR of 3.1% from 2025 to 2029, with a focus on premiumization and specialized products to counteract declining birth rates [59][66]. - The report highlights that single-child consumption is increasing, which helps mitigate the impact of declining birth rates on the market [69].

Core Insights - The report indicates that the real estate market is in a bottoming phase, with a focus on differentiation among sectors and a core strategy for recovery [1][3] - New home sales are still in a contraction phase, with a year-on-year decline of 7.8% in sales area from January to November 2025, while the decline in new residential sales area is 8.1% [4][7] - The report anticipates that the market will continue to stabilize in 2026, driven by policies aimed at stopping the decline and promoting the construction of quality housing [4][30] Fundamental Analysis - New home sales remain in a contraction zone, with first-tier cities showing relative resilience. From January to November 2025, sales area in first-tier cities decreased by 7.5%, while second and third/fourth-tier cities saw declines of 16.3% and 10.2%, respectively [15][19] - The inventory level remains high, with the average de-stocking cycle for commercial housing at 10.4 months and 6.6 months for residential properties. First-tier cities experience relatively lighter de-stocking pressure [22][23] - The land market is characterized by "volume reduction and quality improvement," with residential land transactions down by 7.3% in area but with an increase in average floor price by 12.3% [40][44] Investment Themes - **Hong Kong Residential Market**: There is a recovery in residential transactions, with a 16.2% year-on-year increase in the number of sales contracts from January to November 2025. The private residential price index has risen by 3.4% since March [4][70] - **Commercial Sector**: Policies aimed at boosting consumption have led to a steady recovery in retail sales, with a 3.0% year-on-year increase from January to November 2025. Shopping center foot traffic has stabilized, showing a 14.1% increase in the first half of 2025 [4][5] - **Brokerage Sector**: The pressure to deplete new home inventory has led developers to rely more on brokerage channels, with the proportion of sales expenses attributed to distribution and agency commissions reaching 51.9% in the first half of 2024 [4][19] Market Outlook - The report forecasts that the overall market will continue to bottom out in 2026, with a projected year-on-year decline of 3% in sales area and sales amount [66][67] - New construction and investment are expected to decrease by 10% and 7%, respectively, in 2026, due to reduced land acquisition and weak sales [66][67]

Quantitative Models and Construction Methods 1. Model Name: GAN_GRU - **Model Construction Idea**: The GAN_GRU model combines Generative Adversarial Networks (GAN) for feature generation and Gated Recurrent Unit (GRU) for time-series feature encoding to construct a stock selection factor[4][13][14] - **Model Construction Process**: 1. **GAN Component**: - The generator (G) learns the real data distribution and generates realistic samples from random noise \( z \) (Gaussian or uniform distribution). The generator's loss function is: $$ L_{G} = -\mathbb{E}_{z\sim P_{z}(z)}[\log(D(G(z)))] $$ where \( D(G(z)) \) represents the discriminator's probability of classifying generated data as real[24][25][26] - The discriminator (D) distinguishes real data from generated data. Its loss function is: $$ L_{D} = -\mathbb{E}_{x\sim P_{data}(x)}[\log D(x)] - \mathbb{E}_{z\sim P_{z}(z)}[\log(1-D(G(z)))] $$ where \( D(x) \) is the probability of real data being classified as real, and \( D(G(z)) \) is the probability of generated data being classified as real[27][29][30] - GAN training alternates between optimizing \( G \) and \( D \) until convergence[30] 2. **GRU Component**: - Two GRU layers (GRU(128, 128)) are used to encode time-series features, followed by a Multi-Layer Perceptron (MLP) with layers (256, 64, 64) to predict returns. The final output \( pRet \) is used as the stock selection factor[22] 3. **Feature Input and Processing**: - Input features include 18 price-volume characteristics (e.g., closing price, turnover, etc.) sampled over the past 400 days, with a shape of \( 40 \times 18 \) (40 days of features)[18][19][37] - Features undergo outlier removal, standardization, and cross-sectional normalization[18] 4. **Training Details**: - Training-validation split: 80%-20% - Semi-annual rolling training (June 30 and December 31 each year) - Hyperparameters: batch size equals the number of stocks, Adam optimizer, learning rate \( 1e-4 \), IC loss function, early stopping (10 rounds), max training rounds (50)[18] 5. **Stock Selection**: - Stocks are filtered to exclude ST stocks and those listed for less than six months[18] - **Model Evaluation**: The GAN_GRU model effectively captures price-volume time-series features and demonstrates strong predictive power for stock returns[4][13][22] --- Model Backtesting Results 1. GAN_GRU Model - **IC Mean**: 0.1119*** (2019-2025)[4][41] - **ICIR (non-annualized)**: 0.89[42] - **Turnover Rate**: 0.83X[42] - **Recent IC**: 0.0331*** (December 2025)[4][41] - **1-Year IC Mean**: 0.0669***[4][41] - **Annualized Return**: 37.40%[42] - **Annualized Volatility**: 23.39%[42] - **IR**: 1.60[42] - **Maximum Drawdown**: 27.29%[42] - **Annualized Excess Return**: 22.42%[4][42] --- Quantitative Factors and Construction Methods 1. Factor Name: GAN_GRU Factor - **Factor Construction Idea**: The GAN_GRU factor is derived from the GAN_GRU model, leveraging GAN for price-volume feature generation and GRU for time-series encoding[4][13][14] - **Factor Construction Process**: - The GAN generator processes raw price-volume time-series features (\( Input\_Shape = 40 \times 18 \)) and outputs transformed features with the same shape (\( Input\_Shape = 40 \times 18 \))[37] - The GRU component encodes these features into a predictive factor for stock selection[22] - The factor undergoes industry and market capitalization neutralization and standardization[22] - **Factor Evaluation**: The GAN_GRU factor demonstrates robust performance across various industries and time periods, with significant IC values and excess returns[4][41] --- Factor Backtesting Results 1. GAN_GRU Factor - **IC Mean**: 0.1119*** (2019-2025)[4][41] - **ICIR (non-annualized)**: 0.89[42] - **Turnover Rate**: 0.83X[42] - **Recent IC**: 0.0331*** (December 2025)[4][41] - **1-Year IC Mean**: 0.0669***[4][41] - **Annualized Return**: 37.40%[42] - **Annualized Volatility**: 23.39%[42] - **IR**: 1.60[42] - **Maximum Drawdown**: 27.29%[42] - **Annualized Excess Return**: 22.42%[4][42] 2. Industry-Specific Performance - **Top 5 Industries by Recent IC (October 2025)**: - Social Services: 0.4243*** - Coal: 0.2643*** - Environmental Protection: 0.2262*** - Retail: 0.1888*** - Steel: 0.1812***[4][41][42] - **Top 5 Industries by 1-Year IC Mean**: - Social Services: 0.1303*** - Steel: 0.1154*** - Non-Bank Financials: 0.1157*** - Retail: 0.1067*** - Building Materials: 0.1017***[4][41][42] 3. Industry-Specific Excess Returns - **Top 5 Industries by December 2025 Excess Returns**: - Banking: 4.30% - Real Estate: 3.51% - Environmental Protection: 2.18% - Retail: 1.76% - Machinery: 1.71%[2][45] - **Top 5 Industries by 1-Year Average Excess Returns**: - Banking: 2.12% - Real Estate: 1.93% - Environmental Protection: 1.50% - Retail: 1.46% - Machinery: 1.23%[2][46]

Core Insights - The report indicates that the innovative drug sector is entering a phase of selective stock picking in 2026, following a beta market in 2025. The A-share pharmaceutical industry has risen by 15.9% since the beginning of 2025, underperforming the CSI 300 index by 0.65 percentage points, ranking 17th among industries [2][14]. - The report highlights three key investment directions for 2026: overseas expansion of pharmaceuticals, brain-computer interfaces, and AI in healthcare [2]. Investment Strategy Overview - The innovative drug sector is expected to shift from a broad market rally to a focus on selective stocks in 2026. The average increase for 75 innovative drug sample indices in A-shares reached 54.8%, with Hong Kong's indices doubling [2]. - The report notes that as of December 5, 2025, there were 166 overseas business development (BD) projects, a significant increase from the previous year, with upfront payments reaching $6.3 billion, a growth of over 199% compared to 2024 [2]. Key Investment Directions Overseas Expansion - The report emphasizes the acceleration of Chinese innovative drugs entering international markets, with ADCs and bispecific antibodies being hot topics. The potential for GLP-1R target new drugs remains strong in areas such as long-acting formulations and oral medications [2]. Brain-Computer Interfaces - The report outlines the government's strategic push for brain-computer interfaces as a new economic growth point, with applications in medical rehabilitation for conditions like stroke and spinal cord injuries [2]. AI in Healthcare - The report discusses the establishment of clear short-term and long-term goals for AI in healthcare, covering various applications such as AI health management and clinical decision support systems [2]. Recommended Stocks - The report recommends several companies for investment, including Heng Rui Medicine (600276), BeiGene (688235), Mindray Medical (300760), and others, indicating a diversified approach across the innovative drug and medical device sectors [2].

Quantitative Models and Construction Model Name: DTLC_RL (Decoupled Temporal Contrastive Learning with Reinforcement Learning) - **Model Construction Idea**: The model aims to combine the nonlinear predictive power of deep learning with interpretability by decoupling feature spaces, enhancing representation through contrastive learning, ensuring independence via orthogonal constraints, and dynamically fusing spaces using reinforcement learning[2][11][12] - **Model Construction Process**: - **Feature Space Decoupling**: Three orthogonal latent spaces are constructed to capture market systemic risk (β space), stock-specific signals (α space), and fundamental information (θ space). Each space is equipped with a specialized encoder: TCN for β space, Transformer for α space, and gated residual MLP for θ space[11][12][92] - **Contrastive Learning**: Introduced within each space to enhance robustness by constructing positive and negative sample pairs based on return similarity. The InfoNCE loss function is used to maximize the similarity of positive pairs while minimizing that of negative pairs: $$L_{\mathrm{InfotNCE}}=-E\left[l o g~\frac{e x p\left(f(x)^{\top}f(x^{+})/\tau\right)}{e x p\left(f(x)^{\top}f(x^{+})/\tau\right)+\sum_{i=1}^{N-1}~e x p\left(f(x)^{\top}f(x_{i}^{-})/\tau\right)}\right]$$ where \(f(x)\) is the feature representation, \(x^+\) is the positive sample, \(x^-\) is the negative sample, and \(\tau\) is the temperature parameter[55][56] - **Orthogonal Constraints**: A loss function is added to ensure the outputs of the three spaces are statistically independent, reducing multicollinearity and enhancing interpretability[12][104] - **Reinforcement Learning Fusion**: A PPO-based reinforcement learning mechanism dynamically adjusts the weights of the three spaces based on market conditions. The reward function includes components for return correlation, weight stability, and weight diversification: $$r_{t}=R_{t}^{I C}\big(\widehat{y_{t}},y_{y}\big)+\lambda_{s}R_{t}^{s t a b l e}+\lambda_{d}R_{t}^{d i v}$$ The PPO optimization process includes GAE advantage estimation and a clipped policy loss: $$L^{C L P}=E\left[\operatorname*{min}(r\dot{A},c l i p(r,1-\varepsilon,1+\varepsilon)\dot{A})\right]$$[58][120][121] - **Model Evaluation**: The DTLC_RL model demonstrates strong predictive power and interpretability, with dynamic adaptability to market conditions[2][12][122] Model Name: DTLC_Linear - **Model Construction Idea**: A baseline model for comparison, using a linear layer to fuse the three feature spaces[98][100] - **Model Construction Process**: - The encoded information from the three spaces is concatenated and passed through a linear layer with a Softmax activation to generate fusion weights. The model is trained with a multi-task loss function, including IC maximization, contrastive learning loss, and orthogonal constraints[98][104] - **Model Evaluation**: Provides a benchmark for evaluating the contribution of reinforcement learning in DTLC_RL[98][103] Model Name: DTLC_Equal - **Model Construction Idea**: A simpler baseline model that equally weights the three feature spaces without dynamic adjustments[98] - **Model Construction Process**: The outputs of the three spaces are directly averaged to generate predictions[98] - **Model Evaluation**: Serves as a control group to assess the benefits of dynamic weighting in DTLC_RL[98][103] --- Model Backtesting Results DTLC_RL - **IC**: 0.1250[123] - **ICIR**: 4.38[123] - **Top 10% Portfolio Annualized Return**: 34.77%[123] - **Annualized Volatility**: 25.41%[123] - **IR**: 1.37[123] - **Maximum Drawdown**: 40.65%[123] - **Monthly Turnover**: 0.71X[123] DTLC_Linear - **IC**: 0.1239[105] - **ICIR**: 4.25[105] - **Top 10% Portfolio Annualized Return**: 32.95%[105] - **Annualized Volatility**: 24.39%[105] - **IR**: 1.35[105] - **Maximum Drawdown**: 35.94%[105] - **Monthly Turnover**: 0.76X[105] DTLC_Equal - **IC**: 0.1202[105] - **ICIR**: 4.06[105] - **Top 10% Portfolio Annualized Return**: 32.46%[105] - **Annualized Volatility**: 25.29%[105] - **IR**: 1.28[105] - **Maximum Drawdown**: 40.65%[105] - **Monthly Turnover**: 0.71X[105] --- Quantitative Factors and Construction Factor Name: Beta_TCN - **Factor Construction Idea**: Captures market systemic risk by quantifying stock sensitivity to common risk factors like macroeconomic fluctuations and market sentiment[67] - **Factor Construction Process**: - Five market-related features are selected, including beta to market returns, volatility sensitivity, liquidity beta, size exposure, and market sentiment sensitivity[72] - A TCN encoder processes 60-day time-series data, using dilated causal convolutions to capture short- and medium-term trends. The output is a 32-dimensional vector representing systemic risk features[68] - **Factor Evaluation**: Demonstrates moderate stock selection ability and effectively captures market-related information[73] Factor Name: Alpha_Transformer - **Factor Construction Idea**: Extracts stock-specific alpha signals from price-volume time-series data[76] - **Factor Construction Process**: - Thirteen price-volume features are encoded using a multi-scale Transformer model, with separate layers for short-, medium-, and long-term information. Outputs are fused using a gated mechanism and passed through a fully connected layer for return prediction[77][78] - **Factor Evaluation**: Exhibits strong predictive power and stock selection ability, with relatively low correlation to market benchmarks[81][82] Factor Name: Theta-ResMLP - **Factor Construction Idea**: Focuses on fundamental information to assess financial safety margins and risk resistance[88] - **Factor Construction Process**: - Eight core financial indicators, including PE, PB, ROE, and dividend yield, are encoded using a gated residual MLP. The architecture includes input projection, gated residual blocks, and a final output layer[92] - **Factor Evaluation**: Provides stable stock selection performance with lower turnover and drawdown compared to other spaces[95][96] --- Factor Backtesting Results Beta_TCN - **IC**: 0.0969[73] - **ICIR**: 3.73[73] - **Top 10% Portfolio Annualized Return**: 27.73%[73] - **Annualized Volatility**: 27.19%[73] - **IR**: 1.02[73] - **Maximum Drawdown**: 45.80%[73] - **Monthly Turnover**: 0.79X[73] Alpha_Transformer - **IC**: 0.1137[81] - **ICIR**: 4.19[81] - **Top 10% Portfolio Annualized Return**: 32.66%[81] - **Annualized Volatility**: 23.04%[81] - **IR**: 1.42[81] - **Maximum Drawdown**: 27.59%[81] - **Monthly Turnover**: 0.83X[81] Theta-ResMLP - **IC**: 0.0485[95] - **ICIR**: 1.87[95] - **Top 10% Portfolio Annualized Return**: 23.88%[95] - **Annualized Volatility**: 23.96%[95] - **IR**: 0.99[95] - **Maximum Drawdown**: 37.41%[95] - **Monthly Turnover**: 0.41X[95]