绝对收益产品及策略周报（250901-250905） [Table_Authors] 郑雅斌(分析师) 上周 94 只固收+基金创新高 本报告导读： 股票端采用小盘成长组合+不择时的股债 10/90 和 20/80 月度再平衡策略，2025 年累 计收益分别为 5.88%和 10.99%。 投资要点： 金 融 工 程 周 报 固收+产品业绩跟踪。截至 2025 年 09 月 05 日，全市场固收+基金 规模 17854.15 亿元，产品数量 1179 只，其中 94 只上周净值创历史 新高。上周（20250901-20250905，下同）共新发 1 只产品，各类型 基金业绩中位数表现分化：混合债券型一级（0.09%）、二级（0.05%）、 偏债混合型（0.06%）、灵活配置型（0.00%）、债券型 FOF（0.14%） 及混合型 FOF（0.05%）。按风险等级划分，保守型、稳健型、激进 型基金中位数收益分别为 0.07%、0.06%、0.05%。 大类资产配置和行业 ETF 轮动策略跟踪。1）大类资产择时观点。 2025Q3 逆周期配置模型给出的宏观环境预测结果为 Inflation，8 月 以来沪深 30 ...

Summary of Key Points Core Viewpoint - On September 10, the industry-themed ETF funds experienced a net inflow of 1.12 billion yuan, with a cumulative net inflow of 16.857 billion yuan over the past five trading days, indicating strong investor interest in these funds [1]. Fund Inflows - A total of 132 industry-themed ETFs saw net inflows, with the top performer being the Yongying CSI Hong Kong and Shanghai Gold Industry Stock ETF (517520), which had an increase of 402 million shares and a net inflow of 747 million yuan [1][3]. - The latest scale of the Yongying ETF is 9.308 billion yuan, despite a decline of 1.69% in its price [3]. Fund Outflows - Conversely, 143 industry-themed ETFs experienced net outflows, with the leading outflow being from the Huatai-PineBridge CSI Photovoltaic Industry ETF (515790), which saw a reduction of 506 million shares and a net outflow of 441 million yuan [4][5]. - The latest scale of the Huatai ETF is 14.882 billion yuan, with a price drop of 1.80% [5]. Detailed Fund Performance - Other notable ETFs with significant net inflows include: - Penghua CSI Sub-Segment Chemical Industry ETF (159870) with a net inflow of 448 million yuan [3]. - Guotai CSI All-Index Securities Company ETF (512880) with a net inflow of 363 million yuan [3]. - ETFs with significant net outflows include: - E Fund CSI Artificial Intelligence Theme ETF (159810) with a net outflow of 344 million yuan [5]. - Jiashi Shanghai Stock Exchange Science and Technology Innovation Board Chip ETF (588200) with a net outflow of 341 million yuan [5].

Core Viewpoint - The artificial intelligence sector has experienced a strong surge, with over 10 billion yuan flowing into related ETFs since August [1] Fund Flows - From August 1 to September 1, the net subscription amount for AI-themed ETFs reached 15.319 billion yuan [1] - The E Fund CSI Artificial Intelligence Theme ETF saw a net subscription of 2.568 billion yuan [1] - The Huafu CSI Artificial Intelligence Industry ETF recorded a net subscription of 2.323 billion yuan [1] Market Outlook - According to Huaxia Fund, after a short-term adjustment, the valuation is returning, and the AI market may follow a rotation pattern of "hardware-software-application," presenting a good opportunity for investment after the adjustment [1]

Market Overview - The Shanghai Composite Index rose by 0.84% to close at 3857.93 points, with a weekly high of 3888.6 points [1] - The Shenzhen Component Index increased by 4.36% to 12696.15 points, reaching a high of 12791.18 points [1] - The ChiNext Index saw a significant rise of 7.74%, closing at 2890.13 points, with a peak of 2933.99 points [1] - In contrast, major global indices experienced declines, with the Nasdaq Composite down by 0.19%, the Dow Jones Industrial Average down by 0.19%, and the S&P 500 down by 0.1% [1] ETF Market Performance - The median weekly return for stock ETFs was 2.82% [2] - The highest performing scale index ETF was the Southern CSI Technology Innovation 50 ETF, with a return of 11.65% [2] - The highest performing industry index ETF was the GF National Communication ETF, returning 13.54% [2] - The top five stock ETFs by weekly return included the National Communication Equipment ETF (17.11%), the Fortune CSI Communication Equipment Theme ETF (16.68%), and others [4][5] ETF Liquidity - Average daily trading volume for stock ETFs increased by 55.5%, while average daily turnover rose by 30.9% [6] ETF Fund Flows - The top five stock ETFs by fund inflow included the Jiashi CSI Technology Innovation Board Chip ETF (1.385 billion yuan) and the National Communication Equipment ETF (0.982 billion yuan) [9][10] - The largest outflows were from the Southern CSI Technology Innovation 50 ETF (0.973 billion yuan) and the Southern CSI 500 ETF (0.720 billion yuan) [10] ETF Financing and Margin Trading - The financing balance for stock ETFs increased from 41.1178 billion yuan to 43.9557 billion yuan [11] - The highest financing buy amount was for the E Fund ChiNext ETF, totaling 1.238 billion yuan [11] ETF Market Size - The total market size for ETFs reached 51,130.84 billion yuan, with stock ETFs accounting for 34,987.39 billion yuan [14] - Stock ETFs represented 80% of the total number of ETFs and 68.4% of the total market size [16] New ETF Issuance - No new ETFs were issued last week, but ten new ETFs were established, including the Dachen ChiNext 50 ETF and others [17] Industry Insights - Galaxy Securities noted that the communication AI sector is in an upward trend with low valuations, driven by performance exceeding expectations [17] - Guosheng Securities highlighted that the optical communication industry is entering a new growth phase due to the expansion of global AI computing power, with a shift from performance realization to expectation amplification [19]

Group 1 - The core viewpoint of the articles highlights the significant growth and popularity of ETFs in the current market, driven by improved market sentiment and increased capital inflow [1][7][8] - As of August 22, the total market size of ETFs has surpassed 4.9 trillion yuan, marking an increase of over 1 trillion yuan since the end of last year [1][7] - The wide-based ETFs and industry-themed ETFs have shown strong capital attraction, with the number of industry-themed ETFs exceeding 23, each surpassing 10 billion yuan in scale [2][3] Group 2 - The Huatai-PB CSI 300 ETF saw a single-day scale increase of 112.02 billion yuan on August 22, indicating a strong capital inflow [4][5] - The overall scale of the CSI 300 ETFs has increased by 1,595.70 billion yuan since the beginning of the year, leading other index products [5][6] - The financial technology and securities sectors are particularly favored, with significant inflows into related ETFs, reflecting investor interest in these sectors [2][3][6] Group 3 - The ETF market is expected to continue expanding, playing a dual role as a "weather vane" and "ballast" in future market conditions [1][8] - The low cost, high transparency, and efficient capital absorption capabilities of ETFs make them a preferred investment tool for market participants [7][8] - Regulatory support for the long-term healthy development of the capital market aligns with the growth of ETFs, enhancing market stability and pricing efficiency [8]

Group 1 - The core viewpoint of the articles highlights the significant growth and popularity of ETFs as a primary channel for capital inflow in the current market environment, with total ETF market size surpassing 4.9 trillion yuan, an increase of over 1 trillion yuan since the end of last year [1][7] - Broad-based ETFs and industry-themed ETFs have shown strong capital attraction, with the number of industry-themed ETFs exceeding 23, and specific funds like the Huabao CSI Financial Technology Theme ETF reaching a record high of 110.27 billion yuan, doubling from 46.7 billion yuan at the beginning of the year [2][7] - The securities sector ETFs have also expanded, with the Guotai CSI All-Share Securities Company ETF surpassing 400 billion yuan, reflecting a growing interest in the undervalued brokerage sector amid a market recovery [3][7] Group 2 - The Huabei CSI 300 ETF has emerged as a leader in the current market rally, with a net increase of 273.15 billion yuan on August 22 alone, and a total increase of 1,595.70 billion yuan since the beginning of the year [5][6] - The rapid growth of ETFs aligns with regulatory efforts to promote the long-term healthy development of the capital market, as ETFs facilitate rational capital allocation and enhance market stability through their transparency and liquidity [8]

Group 1: Core Insights - The report highlights that the stock side employs a small-cap growth portfolio combined with a non-timing stock-bond monthly rebalancing strategy, projecting cumulative returns of 5.93% and 11.15% by 2025 [1][4] - As of August 15, 2025, the total market size of fixed income plus funds reached 1,784.66 billion, with 1,177 products, of which 412 achieved historical net value highs last week [2][9] - The report indicates a divergence in performance among various fund types, with median returns for mixed bond type funds being -0.07% for level one, 0.17% for level two, and 0.33% for mixed bond type funds [2][12] Group 2: Asset Allocation and ETF Rotation - The macro environment forecast for Q3 2025 suggests an inflationary trend, with the CSI 300 index, the total wealth index of government bonds, and AU9999 contracts yielding 3.11%, -0.32%, and 1.03% respectively since August [3][4] - Recommended industry ETFs for August 2025 include those focused on artificial intelligence, semiconductors, non-ferrous metals, banking, and major consumer sectors, with a weekly return of 4.01% and a cumulative return of 5.81% for the month [3][4] Group 3: Absolute Return Strategy Performance - The macro-timing driven stock-bond 20/80 rebalancing strategy yielded a return of 0.47% last week, while the stock-bond risk parity strategy returned -0.02% [4][9] - The small-cap growth style within the stock-bond 20/80 combination showed the most significant performance, with a year-to-date return of 11.15% [4][9] - The report notes that the cumulative return for the small-cap growth portfolio, when adjusted for timing strategies, reached 12.81% [4][9]

Quantitative Models and Construction Methods Model: DeepSeek-R1 - **Model Construction Idea**: The DeepSeek-R1 model aims to innovate in AI technology by reducing dependency on high-end imported GPUs and enhancing cost-effectiveness and performance in global markets[5][12][30] - **Model Construction Process**: - The model is based on the DeepSeek-V3 architecture and applies reinforcement learning techniques during the post-training phase to significantly improve inference capabilities with minimal labeled data[33] - The model's performance in tasks such as mathematics, coding, and natural language inference is on par with OpenAI's o1 official version[33] - The team also introduced six distilled small models using knowledge distillation techniques, with the 32B and 70B versions surpassing OpenAI o1-mini in several capabilities[34] - The model's training cost was $5.576 million, only 1/10th of GPT-4o's training cost, and its API call cost is 1/30th of OpenAI's similar services[38] - **Formula**: $$ \text{SUE} = \frac{\text{Single Quarter Net Profit} - \text{Expected Net Profit}}{\text{Standard Deviation of Net Profit YoY Change over the Past 8 Quarters}} $$ where Expected Net Profit = Last Year's Same Quarter Actual Net Profit + Average YoY Change in Net Profit over the Past 8 Quarters[55] - **Model Evaluation**: The model is highly cost-effective and adaptable to different application environments, breaking the traditional AI industry's reliance on "stacking computing power and capital"[38][43] Model Backtesting Results - **DeepSeek-R1 Model**: - **AIME pass@1**: 9.3 - **AIME cons@64**: 13.4 - **MATH-500 pass@1**: 74.6 - **GPQA Diamond pass@1**: 49.9 - **LiveCodeBench pass@1**: 32.9 - **CodeForces rating**: 759.0[36] Quantitative Factors and Construction Methods Factor: Standardized Unexpected Earnings (SUE) - **Factor Construction Idea**: SUE is used to measure the growth potential and latest marginal changes in the prosperity of the industry and individual stocks[57] - **Factor Construction Process**: - SUE is calculated as: $$ \text{SUE} = \frac{\text{Single Quarter Net Profit} - \text{Expected Net Profit}}{\text{Standard Deviation of Net Profit YoY Change over the Past 8 Quarters}} $$ where Expected Net Profit = Last Year's Same Quarter Actual Net Profit + Average YoY Change in Net Profit over the Past 8 Quarters[55] - **Factor Evaluation**: SUE effectively measures future earnings growth and the latest marginal changes in prosperity, representing the future trend changes in the industry[57] Factor Backtesting Results - **SUE Factor**: - **2022**: -29.8% - **2023**: 15.9% - **2024**: 20.1% - **2025 YTD**: 11.0%[65]

Group 1 - The Huafu CSI A500 Index Fund was launched for public offering from July 9, 2025, to September 30, 2025, with a minimum total fundraising amount of 200 million shares [2] - The fund aims to track the CSI A500 Index, which includes 500 securities selected from various industries based on market capitalization and liquidity [2] - The fund's management fee is set at an annual rate of 0.5% based on the previous day's net asset value [4] Group 2 - The fund is managed by Zhang Ya and Li Xiaohua, both of whom have significant experience in fund management [5][6] - Zhang Ya currently manages 7 funds with a total scale exceeding 10 billion, while Li Xiaohua manages 12 funds with a total scale exceeding 5 billion [7] - The Huafu CSI Artificial Intelligence Industry ETF, also managed by Zhang Ya and Li Xiaohua, has seen a net value increase of 6.96% year-to-date, slightly outperforming its benchmark [7] Group 3 - As of July 8, the CSI A500 Index has recorded a year-to-date increase of 1.71% [3]

1. Report Industry Investment Rating No relevant content provided. 2. Core Views of the Report - Big models are entering the AI Agent explosion stage, and Agents may comprehensively drive edge-side intelligence. Human - AI collaboration may transition to the Agent mode, and currently, Agent capabilities are in a stage similar to the transition from GPT3 to ChatGPT, potentially driving edge - side intelligence [2][7]. - More powerful open - source models are emerging, which are expected to trigger an AI application boom. Domestic large models have comparable language abilities to overseas ones, and the inference side has witnessed rapid iteration in 2025. Open - source models are more suitable for enterprise - level applications and may drive an AI application upsurge [2][17]. - The CSI Artificial Intelligence Industry Index selects 50 representative companies as sample stocks, reflecting the overall performance of AI industry companies. It has features such as focusing on AI revenue ratio in component stock selection, scientific weight distribution, balanced industry distribution, and strong performance in capturing industry changes [2]. - The Huafu CSI Artificial Intelligence Industry ETF is a fund benchmarked against the CSI Artificial Intelligence Industry Index, aiming to closely track the target index with controlled tracking deviations and errors [2][52]. 3. Summary According to the Directory 3.1 Big models are entering the AI Agent explosion stage, and Agents may comprehensively drive edge - side intelligence - **Three modes of human - AI collaboration**: Embedding mode (e.g., ChatGPT), Copilot mode (e.g., Microsoft 365 Copilot), and the future Agent mode where humans set goals and provide resources while AI does most of the work [7]. - **Current stage of AI development**: Currently in the reasoning stage, approaching the AI Agent stage. As Agent capabilities improve, AI has expanded from language/text to multi - modality and tool use [10]. - **Driving edge - side intelligence**: In the Agent era, the demand for reasoning - side computing power will increase. Mobile phone manufacturers with their own hardware and operating systems, and Internet giants with operating systems/APP application ecosystems have advantages in developing AI Agents [14]. 3.2 More powerful open - source models are emerging, which are expected to trigger an AI application boom - **Domestic large - model capabilities**: Domestic large models have comparable language abilities to overseas ones, and the inference side has developed rapidly in 2025. For example, DeepSeek R1 represents comparable reasoning capabilities to overseas models [17]. - **Advantages of open - source models**: Open - source models like Deepseek and Llama4 are comparable to closed - source models in terms of capabilities. Their transparency and customizability are more suitable for the diverse needs of enterprise - level applications, potentially driving an AI application boom [17]. 3.3 CSI Artificial Intelligence Industry Index: One - click investment in core AI assets - **Compilation scheme**: Published on 2018/11/21, it selects 50 sample stocks based on AI business proportion and total market value. It focuses on AI revenue ratio to avoid concept speculation and has a unique weight distribution mechanism [19][21]. - **Component stock distribution**: It uses an AI revenue ratio and growth indicator - adjusted market - value weighted method. The top ten component stocks mostly have free - floating market values between 40 billion and 100 billion, and the index has a high concentration, amplifying the growth potential of core targets [23][25]. - **Industry distribution**: With a comprehensive and balanced industry distribution, it has a "hardware - software collaboration, scenario - connected" ecological investment portfolio. It can automatically adjust weights in emerging fields and cover the entire AI industry chain [32][33]. - **Fundamental characteristics**: It has a high - purity investment portfolio through a "two - dimensional screening mechanism", with significant R & D investment and strong growth potential in financial indicators [37][39]. - **Investment value**: It has a leading ability to capture industry changes compared to traditional AI indexes. In early 2025, it outperformed traditional AI indexes and broad - based indexes [45]. 3.4 Huafu CSI Artificial Intelligence Industry ETF Fund Introduction - The Huafu CSI Artificial Intelligence Industry ETF (515980.SH) is issued by Huafu Fund, benchmarked against the CSI Artificial Intelligence Industry Index. It aims to control the daily average tracking deviation and annual tracking error, with current management and custody fees of 0.50% and 0.10% respectively [52]. 3.5 Fund Manager Information - **Fund manager introduction**: Huafu Fund Management Co., Ltd. was established in 2004, with a clear development strategy, a rich product system, and an experienced investment and research team [56]. - **Fund manager profiles**: Zhang Ya, Gao Zhe, and Li Xiaohua have rich experience in fund management, with multiple products under their management and a certain scale [57][58][60].