每周一景： 云南玉龙雪山 点击右上角菜单，收听朗读版 【太阳能 】盈利能力环比改善，行业底部无需过度悲观——光伏行业季报总结 投资建议： 根据国家能源局， 2025 年第一季度，全国光伏新增并网 59.71GW ，同比增长 30.5% ， 其中集中式光伏 23.41GW ，分布式光伏 36.31GW 。我们认为，在碳中和大背景下， 25 年国内需求依 然有望保持稳定， 25 年第一季度国内新增并网量的高速增长，证明光伏的需求具备较强韧性，期待新的 场景和政策进一步促进行业发展；我们预计 2025 年国内光伏装机容量将达到 280GW ，同比小幅增长。 2025 年全球光伏装机需求将有望达到 583GW ，同比增长 10% ，海外新兴市场增速更快。当前行业处 于周期底部，价格已得到比较充分的调整，需求依然表现出韧性，给予光伏行业"增持"评级。 光伏行业 25Q1 毛利率和净利率环比提升。 2024 年，光伏板块合计实现营业收入 9737.40 亿元，同比 下降 21.24% ；合计实现归母净利润 -186.30 亿元，同比下降 116.88% ；板块平均毛利率为 12.84% ，同比降低 8.35 个百分点；平 ...

报告导读： 2025Q1营收、净利润增速为正的公司增多，整体增速也实现提升。按市值 拆分，大市值公司业绩表现更加稳健。按板块拆分，2025Q1 AI、能源IT等多板块实现 复苏。 投资建议：我们认为计算机景气有望持续修复。 2025Q1 增速为正公司增多，业绩边际复苏。 ① 2024 年计算机营收 12693.99 亿元，同比 +4.78% ； 归母净利润 187.96 亿元，同比 -46.82% ；扣非净利润 82.48 亿元，同比 -52.28% 。 2024 年计算 机员工总数为 120.10 万人，同比 -0.63% 。 2025Q1 计算机行业总营收 2860.32 亿元，同比 +15.25% ；归母 28.16 亿元，同比 +193.52% ；扣非 0.36 亿元，同比 +102.25% 。 2025Q1 营收 利润均呈恢复态势。②业绩增速分布上，营收正增速公司占比从 2024 年 49.45% 提升到 2025Q154.27% ；归母净利润正增速从 42.82% 提升到 45.03% ，扣非净利润正增速从 45.86% 下跌 到 42.82% 。③按季度拆分， 2025Q1 计算机公司营收、归母 ...

投资建议：维持行业"增持"评级。 我们认为：在智驾平权和出海两大趋势加持下，汽车行业的成长仍然 不乏亮点；同时由于行业竞争激烈，整车与零部件围绕年降的博弈将持续。我们认为受益的标的包括新能 源整车，以及线控底盘零部件、智能座椅零部件和头部新势力核心供应链公司。 汽车行业成长性尚可。 根据 WIND 及各公司财报，全部汽车行业 A 股公司 2024/2025Q1 营收同比增 速为 6.6%/5.7% 。二级行业方面， 2025Q1 营收成长性排序为摩托车及其他 > 乘用车 > 汽车零部件 > 商用车 > 汽车服务。从细分的三级行业看，营收成长性的亮点在于电动乘用车、汽车电子等。我们认为随 着辅助驾驶的发展，汽车行业的成长亮点仍在。 归母净利润增速口径：全部汽车行业 A 股公司 2024/2025Q1 营收同比增速为 10.9%/11.3% 。 二级 行业方面， 2025Q1 成长性排序为摩托车及其他 > 汽车服务 > 乘用车 > 汽车零部件 > 商用车。从细分 的三级行业看，归母净利润成长性的亮点在于电动乘用车、商用载客车等。我们认为，虽然行业竞争激 烈，优秀企业仍然能够不断提升效率，实现利润率提升。 报告导读 ...

价格已处于底部区间，减值压力有所减小。 光伏行业 2024 全年计提资产减值损失共 480.91 亿元， 24Q4 、 25Q1 单季度计提资产减值损失分别为 242.63 亿元、 37.52 亿元， 25Q1 同比明显下降。我 们认为，产业链价格已经充分调整至底部区间，光伏行业公司已连续计提了较为大额的资产减值损失，后 续存货减值的压力会变小。 负债率小幅上升，经营性现金流同比改善。 光伏板块 24Q4 、 25Q1 末的资产负债率分别为 61.60% 、 61.98% ，环比小幅上升，板块资产负债率自 24Q2 开始处于 60% 以上。 24Q4 、 25Q1 单季度经营 性现金流净额分别为 472.44 亿元、 -33.03 亿元， 25Q1 现金流同比有所改善，仅硅料和组件辅材是同 比下降的，根据历史数据来看， 1 季度一般是全年的低点，企业需要进行一定的原材料备货，我们认为， 由于行业当前处于周期底部，企业开始加强存货管理，更多的实行"以销定产"的策略，现金流表现有望得 到更好的控制。 风险提示： 行业政策波动风险；行业需求下滑风险；竞争加剧风险；新技术替代风险；产品价格波动风 险。 报告导读： 光 ...

Core Viewpoint - The current bottleneck in inference speed is primarily due to memory bandwidth rather than computing power, with the NPU+DRAM stacking technology showing significant improvements in memory bandwidth, indicating a clear industry trend [1][2]. Group 1: Inference Speed and Memory Bandwidth - The Qualcomm Snapdragon 8 GEN 3 has an NPU computing power of approximately 45 TOPs and a memory bandwidth of about 67 GB/s. When running a 7B large model, the calculation capability is limited to approximately 3215 tokens/s by computing power and 4.8 tokens/s by memory bandwidth, with the final speed being constrained by the lower of the two values, highlighting the significant memory bandwidth limitation [2]. - A practical test on a Xiaomi phone using the Qwen3-8B-MNN model showed a decoding speed of 222 tokens with an average response time of 32 seconds, indicating that a user-perceived inference speed should reach 40-50 tokens/s [2]. Group 2: 3D DRAM Solution - The memory limitation for edge AI can be addressed by 3D DRAM. By stacking DRAM and NPU through HB technology, if the memory bandwidth is increased to 800 GB/s, the memory limitation could rise to 57 tokens/s [3]. - Key players in this space include Chinese companies like Zhaoyi Innovation and its subsidiary Qingyun Technology, as well as Taiwanese storage IDM Winbond and mobile AP leader Qualcomm, all focusing on the 3D DRAM+NPU solution, indicating a clear technological trend [3]. Group 3: Hardware and Model Development - The current industry phase suggests that hardware is leading model development, with future growth expected to be driven by model advancements benefiting from hardware improvements. Hardware solutions require extensive stability testing before commercial deployment at scale [3]. - Qualcomm must adopt strategies suitable for AI large model devices to avoid risks associated with a potential "GPU" revolution in mobile AI by the end of 2025 or 2026, as companies prepared with the right hardware and models could experience a significant one-year window of opportunity [3].

Core Viewpoint - The report analyzes the price rhythm of gold from the perspective of trading structure, highlighting the significant role of central bank gold purchases and investment demand in driving gold prices upward in 2024 [1][2]. Group 1: Central Bank Purchases and Demand - Central banks are projected to purchase approximately 1,044 tons of gold in 2024, becoming a crucial driver for gold price increases [1]. - The investment demand for gold is expected to reach 1,179 tons in 2024, indicating a shift in demand dynamics as prices rise [1]. Group 2: Consumer and Industrial Demand - Global gold jewelry consumption is anticipated to decline by 11% year-on-year in 2024, amounting to around 1,877 tons, due to high gold prices suppressing consumer demand [1]. - Industrial demand for gold remains low and stable, projected at only 326 tons in 2024 [1]. Group 3: Trading Structure and Market Dynamics - There has been a significant increase in gold ETF sizes, with domestic gold ETFs seeing a rise of over 50 billion in April, corresponding to a demand for 50-60 tons of physical gold [1]. - A surge in trading volume for A-share gold stocks has been observed, indicating heightened investor enthusiasm for gold investments and a crowded trading environment [1][2]. Group 4: Macro Factors Influencing Gold Prices - The current rise in gold prices is driven by a decline in the credibility of the US dollar and a restructuring of the monetary system, particularly following the freezing of Russian foreign exchange reserves in 2022 [2]. - Increasing uncertainty in the global investment landscape, exacerbated by issues such as the US debt ceiling and unpredictable government policies, is likely to continue pushing investors towards safe-haven assets like gold [2].

Core Viewpoint - The article discusses the historical extreme fluctuations in the US stock market over the past fifty years, highlighting three significant events: Black Monday in 1987, the 2008 financial crisis, and the COVID-19 pandemic in 2020. It emphasizes that the future trajectory of the US stock market may depend on the impact of tariff policies on the economy and the pace of industrial transformation led by AI [1][3]. Group 1: Historical Market Fluctuations - The US stock market has experienced three major extreme fluctuations in the past fifty years: Black Monday in 1987, the 2008 financial crisis, and the COVID-19 pandemic in 2020, which are considered the most typical major shocks in this period [1][2]. - Black Monday in 1987 was characterized by a combination of a Federal Reserve interest rate hike, pressure from a depreciating dollar, and a bubble burst leading to significant sell-offs, resulting in the largest single-day drop in history [2]. - The 2008 financial crisis was triggered by the collapse of the subprime mortgage market, which led to a liquidity crisis, culminating in the bankruptcy of Lehman Brothers and a simultaneous drop in stocks, bonds, and currencies [2]. - The COVID-19 pandemic in 2020 caused a global economic slowdown, with significant negative impacts on production, consumption, and employment in the US, leading to multiple market circuit breakers being triggered within ten days [2]. Group 2: Current Market Dynamics - Since April, the tariff policies of the Trump administration have led to significant volatility in the US stock market, raising concerns about a potential "hard landing" for the US economy amid increasing policy uncertainty and trade tensions [3]. - The impact of tariff policies on the US economy has not yet fully materialized, and if previously postponed tariffs are implemented, the economy may face risks of hard landing or stagflation, which could put additional pressure on the stock market [3]. - The ongoing industrial transformation driven by AI is seen as a critical factor that may support corporate earnings in the US stock market as the trend continues to evolve [3].

大众品：表现分化，结构性增长。 1 ）啤酒： 25Q1 收入同比 +4% 、净利润同比 +11% ， 销量景气修 复吨价承压，成本优势利润率持续改善。 2 ）软饮料： 25Q1 收入同比 +3% 、净利润同比 +1% ，继 续保持较好景气度，龙头亮各有亮点。 3 ）零食： 食品综合板块分化明显，其中新渠道和新品类驱动下 零食成长性凸显， 24Q4 、 25Q1 收入分别同比 +18% 、 +2% ， 25Q1 受高基数和春节提前等因素影 响业绩增速阶段性下滑。 4 ）乳制品： 2024 年（ A 股）收入同比 -7% 、净利润 -27% ，度过调整期 后 25Q1 景气度和毛销差边际改善。 5 ）调味品： 2024 年板块收入边际改善， 25Q1 业绩稳步增长， 收入同比 +3% 、净利润同比 +7% 、毛利率同比 +1.6pct 。 6 ） 餐饮供应链： 竞争加剧导致 2024 年 板块利润承压， 25Q1 收入同比 -5% 、利润同比 -13% 环比降速。 风险提示： 消费复苏不及预期、市场竞争加剧、成本进一步上涨、食品安全问题。 文章来源 本文摘自：2025年5月5日发布的 结构分化，重视成长——食品饮 ...

Core Viewpoint - The article highlights the emergence of "Tech Dragon Pearls" in the Greater Bay Area, focusing on companies with technological advantages and commercialization potential in fields such as low-altitude economy, AI large models, embodied intelligence, and intelligent driving, which are expected to challenge the monopoly of international tech giants and lead to a revaluation of Chinese tech assets [1]. Group 1: Company Summaries - DJI Innovation specializes in the production and sales of drones and handheld photography equipment, holding nearly 80% market share in the global consumer drone market, and offers advanced intelligent driving solutions through its subsidiary [2]. - Self-Variable Robotics focuses on the development of "general embodied large models" and has successfully developed the world's largest parameter scale embodied intelligent general operation model, WALL-A, by October 2024 [2]. - Zhijidongli is dedicated to creating full-size general humanoid robots and has developed innovative products, including bipedal robots, and has implemented a video generation large model for embodied operation algorithms [2]. - Yuanxiang is a leading AI and 3D technology service company that has developed and open-sourced a series of high-performance large models, providing a one-stop platform for 3D content creation and consumption [2]. - Simou Technology utilizes deep learning and machine vision as key technology engines, developing a comprehensive ecosystem of smart manufacturing software and hardware products, including industrial multimodal large models and AI-AOI visual inspection equipment [3]. - Superparameter Technology leverages cutting-edge AI technology to enhance the autonomy and interactivity of Game Agents, becoming the first company globally to achieve large-scale commercialization in the Game Agent field [3]. - Yuanrong Qixing offers leading intelligent driving solutions, with nearly 40,000 units sold by March 2025, and aims to develop RoadAGI in the long term [3]. - Hege Technology provides advanced 3D printing application solutions, integrating a complete process from intelligent data collection to 3D printing materials and post-processing [3].

法律声明 本公众订阅号(微信号: GTHT RESEARCH )为国泰海通证券股份有限公司(以下简称"国泰海通证券") 研 究所依法设立、独立运营的唯一官方订阅号。其他机构或个人在微信平台上以国泰海通研究所名义注册 的，或含有"国泰海通研究"，或含有与国泰海通证券研究所品牌名称相关信息的其他订阅号均不是国泰 海通证券研究所官方订阅号。 本订阅号不是国泰海通证券研究报告发布平台，本订阅号所载内容均来自于国泰海通证券研究所已正式 发布的研究报告 如需了解详细的证券研究信息 请具体参见国泰海通证券研究所发布的完整报告 更多国泰海通研究和服务 亦可联系对口销售获取 重要提醒 本订阅号所载内容仅面向国泰海通证券研究服务签约客户。因本资料暂时无法设置访问限制，根据《证 券期货投资者适当性管理办法》的要求，若您并非国泰海通证券研究服务签约客户，为保证服务质量、 控制投资风险，还请取消关注，请勿订阅、接收或使用本订阅号中的任何信息。我们对由此给您造成的 不便表示诚挚歉意，非常感谢您的理解与配合！如有任何疑问，敬请按照文末联系方式与我们联系。 ...