A股算力租赁板块震荡走强，大位科技、鸿博股份双双涨停，天娱数科涨超7%，宏景科技、并行科 技、奥飞数据跟涨。 ...

Investment Rating - The industry rating is "Outperform the Market" (maintained) [4] Core Insights - Domestic computing power demand is surging, with a focus on computing power leasing and AI infrastructure investment opportunities [1][3] - The scale of domestic intelligent computing power is expected to grow significantly, from 259.9 EFLOPS in 2022 to 1117.4 EFLOPS by 2027, representing a compound annual growth rate (CAGR) of 33.9% [1][14] - The "East Data West Computing" project is accelerating the nationwide layout of computing power networks, with over 250 intelligent computing centers built or under construction as of mid-2024 [1][17] Summary by Sections 1. AI Applications Driving Demand for Computing Facilities - The demand for AI servers is rapidly increasing, with the market size expected to grow from 14.9 billion RMB in 2020 to 143.3 billion RMB by 2028 [10][14] - The cloud computing market reached 616.5 billion RMB in 2023, with a year-on-year growth of 35.5% [10] - Major cloud service providers hold 71.5% of the public cloud market share in China [10] 2. Scarcity of High-End Computing Resources and High Returns from Leasing - Computing power leasing is driven by a shortage of intelligent computing resources, particularly AI chips like GPUs [2][29] - A server equipped with 8 H100 GPUs can yield a net profit margin of 32% to 37% over five years, with a payback period of around three years [2][33] - Recommended stocks in the computing power leasing sector include Hainan Huatie and Gan Consulting, with a focus on investment opportunities in AI infrastructure [2][40] 3. Surge in Demand for Intelligent Computing Centers - The investment scale in the intelligent computing center sector reached 87.9 billion RMB in 2023, a year-on-year increase of over 90% [3][42] - The market size for intelligent computing centers is projected to exceed 288.6 billion RMB by 2028, with a CAGR of 26.8% from 2023 to 2028 [3][42] - The operational model of data centers is primarily divided into retail and wholesale types, with significant investments in AI-driven data centers [3][45]

Group 1 - The A-share computing power leasing concept strengthened in the afternoon session [1] - Hangzhou Steel Group and Hengrun Co., Ltd. reached the daily limit increase [1] - Companies such as Dawi Technology, Ningxia Building Materials, Zhejiang University Network New, and Huizhou Intelligent also saw gains [1]

Group 1: DeepSeek Model and Factor Optimization - The DeepSeek model optimizes and generates stock selection factors based on the Alpha158 factor integrated with Qlib, covering five categories: intraday, volatility, price, volume, and price-volume correlation [3] - By optimizing these factors, DeepSeek significantly improves the predictive ability of most factors, particularly excelling in RankIC and ICIR metrics [3] - Generating new factors from scratch is challenging, but using existing optimized factor expressions as references allows for the generation of new high ICIR factors with fewer iterations [3] - In portfolio optimization, combining the optimized Alpha158 factors with newly generated factors using the Lasso model effectively enhances stock selection capabilities, especially in the full A-share data and the CSI 800 index combination [3] Group 2: New Incremental Policies and Economic Stimulus - On March 23, the Premier reaffirmed confidence in China's economic operation and macro policy implementation, emphasizing increased counter-cyclical adjustments and the potential introduction of new incremental policies, mainly referring to the possibility of central government leverage [4] - Following the National People's Congress, high-level leaders conducted intensive research in key regions such as Guizhou, Guangdong, and Fujian, focusing on technology, openness, foreign trade, employment, and security [4] - The State Council approved the "2025 Key Work Division Plan" on March 12, stressing the need to strengthen support for local governments, particularly in resource assurance, technological innovation, and reform and opening up, likely favoring major economic provinces like Zhejiang, Guangdong, and Fujian [4] - The central bank and other departments plan to stimulate the economy through policies such as reserve requirement ratio cuts, interest rate reductions, promoting foreign trade growth, and facilitating cross-border trade [4] - The "Special Action Plan to Boost Consumption" was issued on March 16, proposing measures to stimulate consumption through demand reform, supply upgrades, and livelihood guarantees, while also mentioning stabilizing the stock and real estate markets [4] Group 3: Hainan Huatie's Computing Power Leasing Business - Hainan Huatie (Huatie Dahuangfeng) has begun laying out its computing power leasing business and announced an investment plan for intelligent computing centers in 2024 [5] - As of December 2024, the company has signed computing power service orders totaling 2.475 billion yuan, and with a new contract of 3.69 billion yuan, the cumulative order amount has reached 6.165 billion yuan, 1.39 times its 2023 revenue [5] - The company’s former controlling shareholder has transferred its shares to Hainan State-owned Capital Holding Company, becoming the new controlling shareholder [5] - Expected net profits for the company from 2024 to 2026 are projected to be 619 million yuan, 846 million yuan, and 1.065 billion yuan, with EPS of 0.31 yuan, 0.43 yuan, and 0.54 yuan respectively [5] Group 4: Haiguang Information's Development in CPU and AI Computing - Haiguang Information, established in 2014, obtained AMD technology authorization in 2016 and initiated CPU product development, leveraging AMD's technology and its own R&D capabilities [7] - The company employs a "sell one generation, verify one generation, and develop one generation" strategy to promote continuous iteration of its Haiguang series CPUs and deep computing units (DCUs) [7] - Haiguang has significantly increased R&D investment, with product performance gradually catching up to global leaders while maintaining a leading position in the domestic market [7] - The Haiguang CPU is based on the x86 architecture, featuring a mature ecosystem and strong compatibility, particularly achieving seamless switching at the software and application levels [8] - Haiguang's DCU products, primarily used in supercomputing centers, have become leaders in domestic supercomputing chips, with ongoing development of the next-generation DCU, expected to significantly enhance AI performance [12] - The domestic information technology innovation industry is accelerating, with increasing domestic production rates supported by fiscal policies [10] - Haiguang is positioned to benefit from the rise of the domestic AI computing industry amid increasing U.S. chip export controls [12] Group 5: Investment Strategies During Earnings Season - Stocks disclosing last year's annual reports are suitable for shorter holding periods, while those disclosing quarterly reports are better for longer holding periods [13] - The excess return rate is typically better for shorter holding periods after annual report disclosures, while it improves with longer holding periods after quarterly report disclosures [13] - In the context of technology stocks experiencing box fluctuations, opportunities for low-cost entry in humanoid robots and DeepSeek + AI computing can be considered [13] - Investors should focus on stocks that perform well post-annual report release, keeping holding periods within T+5 for short-term gains [13] - After quarterly report disclosures, excess returns improve with extended holding periods, making it suitable to select stocks with sustained growth potential and consider holding until T+10 [13] - Maintaining dividend stocks as defensive assets is advisable, especially in a market lacking hotspots, with sectors like steel also considered for supplementary allocation [13]

Core Viewpoint - The A-share market is experiencing significant divergence, with over 3,700 stocks declining while the North Securities 50 index rises nearly 1%, indicating a cautious market sentiment as trading volume decreases by 80 billion compared to the previous day [1] Group 1: Nuclear Fusion Sector - The nuclear fusion sector is gaining attention due to anticipated technological breakthroughs, with stocks like Jiusheng Electric and Lansi Heavy Industry seeing substantial gains. The establishment of a nuclear fusion innovation consortium by the National Energy Group and advancements in magnetic materials from the Shanghai Superconducting Laboratory are driving investment in this trillion-dollar sector [3] - Private equity insiders report that core component companies in the nuclear fusion space have visibility on orders extending to 2026, suggesting strong future demand [3] Group 2: Traditional Sectors - Traditional sectors are also performing well, with coal stocks rising due to expectations of enhanced safety regulations, and the power sector benefiting from increased investment in smart grid technology. Companies like Jiangsu New Energy are experiencing significant stock price increases [4] - The military-civilian integration fund is providing substantial support to the military industry, with stocks like Hongyuan Electronics attracting significant capital [4] Group 3: AI Sector Challenges - The AI sector is facing a downturn, with significant declines in stocks related to computing power leasing and other AI subfields. Capital Online, for instance, plummeted by 16% following a downgrade in global AI server shipment forecasts by Goldman Sachs, raising concerns about the industry's profitability [5] - Some AI concept stocks have dynamic price-to-earnings ratios exceeding 100, indicating a need for valuation adjustments as market sentiment shifts [5] Group 4: Market Signals - Three key signals are emerging from the market: the approaching earnings report season is leading investors to prioritize performance certainty; the upcoming issuance of a trillion-yuan special government bond project is expected to support the infrastructure sector; and foreign capital is increasing its positions in nuclear fusion stocks, indicating a strategic long-term investment focus [6] Group 5: Afternoon Strategy Recommendations - Investors are advised to be cautious of the risks associated with the AI sector and to monitor the sustainability of the nuclear fusion concept. The recommendation is to focus on cyclical stocks with reasonable valuations and expected earnings growth, while reducing positions if trading volume does not exceed 900 billion in the afternoon [7]

Market Overview - On March 25, the market sentiment was low, with over 2,700 stocks declining and more than 2,500 stocks rising. There were 57 stocks hitting the daily limit up and 46 stocks hitting the limit down. Popular sectors like deep-sea technology, computing power, and robotics saw a retreat, while the focus shifted to controllable nuclear fusion and state-owned enterprises [1]. Key Stocks Performance - Southern Precision (002553) achieved a 10% increase, closing at 25.40, with a trading volume of 2.698 billion and a turnover rate of 43.29% over 6 days and 5 consecutive limit ups [2][5]. - Pioneer Electronics (002767) recorded a 9.98% increase, closing at 22.03, with a turnover rate of 31.38% over 3 consecutive limit ups [2]. - Yufu Co. (002427) and Sun Cable (002300) both achieved 3 consecutive limit ups, with increases of 10.08% and 10.06%, respectively [2][3]. Institutional Buying and Selling - The top three net purchases on the daily leaderboard were Southern Precision (1.61 billion), Huayang Co. (1.59 billion), and Jincheng Pharmaceutical (1.22 billion) [4][5]. - The top three net sales were Capital Online, Shaoneng Co., and Shenzhou Information, with net sales of 1.74 billion, 1.36 billion, and 1.29 billion, respectively [6][7]. Sector Highlights - Southern Precision is involved in robotics, bearings, and reducers, with a significant increase in trading activity and a high turnover rate [8]. - Huayang Co. focuses on sodium-ion batteries, photovoltaic, and coal, with a notable trading volume of 9.66 billion and a turnover rate of 3.72% [12]. - Capital Online, associated with computing power leasing and DeepSeek concepts, faced a 20% limit down, with a trading volume of 39.24 billion [14][16]. Financial Performance - Southern Precision reported a revenue of 559 million for the first three quarters of 2024, marking a year-on-year increase of 14.28%, and a net profit of 46.41 million, up 635.79% [11].

Investment Rating - The report maintains a "Buy" rating for several companies including Hainan Huatie, Honglu Steel Structure, China Jushi, and others, indicating a positive outlook on their performance in the market [26]. Core Insights - The AI large model is driving a surge in demand for computing power, with the intelligent computing center market expected to grow from 87.9 billion yuan in 2023 to 288.6 billion yuan by 2028, representing a CAGR of 26.8% [4][14]. - Major internet companies, such as Alibaba, are investing over 380 billion yuan in cloud and AI hardware infrastructure over the next three years, which is more than their total investment in the past decade [4][14]. - The computing power leasing business is currently a mainstream service model, with approximately 35% of the deployed computing power being leased, primarily driven by the shortage of AI chips like GPUs and the high costs associated with building and operating self-owned computing centers [16][20]. Summary by Sections Section 1: Focus on Computing Power Leasing Investment Opportunities - The report emphasizes the growth potential in the computing power leasing sector, driven by the increasing demand for AI capabilities and the limitations of self-built computing centers [4][16]. - Key factors for success in the computing power leasing business include the ability to manage balance sheets, customer resource binding, and access to technical talent [20][21]. Section 2: Company Profit Forecasts and Valuations - The report provides detailed profit forecasts and valuations for various companies, highlighting their expected earnings per share (EPS) and price-to-earnings (P/E) ratios for the coming years [26]. - Companies such as Hainan Huatie and Honglu Steel Structure are noted for their strong market positions and growth potential, with specific EPS forecasts indicating positive trends [26]. Section 3: Weekly Market Review - The report includes a review of the weekly performance of the construction and building materials sectors, noting significant price movements and trends among key companies [28][37]. - The report highlights the performance of specific stocks, with some companies experiencing substantial gains year-to-date, indicating a positive market sentiment [37].

点击注册小程序 阿里巴巴财报数据显示，3QFY2025（截至2024年12月31日）营收同增7.6%，归属于普通股股东净利润同 增239.1%。1-3QFY2025实现营业收入7598.93亿元，同比增长5.6%，实现归属于普通股股东净利润 1170.88亿元，同比增长53.1%，实现非公认会计准则净利润1282.75亿元，同比减少3.6%。单季度拆分来 看，3QFY2025实现营业收入2801.54亿元，同比增长7.6%，实现归属于普通股股东净利润489.45亿元，同 比增长239.1%，实现非公认会计准则净利润510.66亿元，同比增长6.5%。 2025年01月20日，DeepSeek正式发布 DeepSeek-R1，并同步开源模型权重。DeepSeek-R1 在后训练阶段大 规模使用了强化学习技术，在仅有极少标注数据的情况下，极大提升了模型推理能力。在数学、代码、自 然语言推理等任务上，性能比肩 OpenAI o1 正式版。 根据海南华铁2024年5月7日公告，公司拟投资 10 亿元开展智算业务，通过向客户提供 GPU 级的高端算力 资源租赁及增值技术服务取得收益。公司将采购芯片、服务器等算力相关设备 ...

事件二：杭州市余杭区数字经济协会发布通知称，根据杭州市经信局《关于开展阿里巴巴集团云和AI投资 计划本地配套供应商摸排的通知》要求，现面向余杭区摸排意向参与阿里巴巴集团云和AI重大投资计划的 本地配套供应商及核心产品。 华铁算力租赁竞争优势如何？ 1）算力租赁是重资产行业，资产负债表的承接与扩张能力是关键。一是资产负债表的承接能力，假设新 增30亿元租赁负债（分期付款下对应30亿元以上的设备资产），则资产负债率从24Q3末的71.4%提升 3.5pcts至74.9%。二是海南国资控股下华铁扩表能力相较同业民企更强。海南国资入主后，华铁信用等级 评定为AA+，融资成本有望从5-5.5%大幅降低。华铁已申请非公开发行债券规模不超过人民币10亿元；其 中第一期公司债券"25华铁01"发行规模3.6亿元，票面利率2.90%，远低于公司当前融资成本。此外，本次 采购设备的资金主要来自融资租赁公司、银行等金融机构授信，国资赋能下资金来源及成本无虞。 2）较强落地执行力及杭州本土资源禀赋。华铁运营主体在杭州，有利于与大型互联网客户的对接和绑 定，以及技术人才的获得；杭州市政府亦积极推动AI产业链发展，如发布《杭州市人工智能 ...