本周观点 投资建议 相关标的： CPU：海光信息、中科曙光、禾盛新材、中国长城、龙芯中科、兴森科技、深南电路、宏和科技。 国内算力：海光信息、寒武纪、东阳光、协创数据、华丰科技、星环科技、神州数码、百度集团、大位科技、润 建股份、中芯国际、华虹半导体、中科曙光、禾盛新材、润泽科技、浪潮信息、东山精密、亿田智能、奥飞数据、 云赛智联、瑞晟智能、科华数据、潍柴重机、金山云、欧陆通、杰创智能。 海外算力/存储：中际旭创、新易盛、兆易创新、大普微、中微公司、天孚通信、源杰科技、胜宏科技、景旺电子、 英维克等；闪迪、铠侠、美光、SK 海力士、中微公司、北方华创、拓荆科技、长川科技。 风险提示 行业竞争加剧的风险；技术研发进度不及预期的风险；特定行业下游资本开支周期性波动的风险。 敬请参阅最后一页特别声明 1 Agent 驱动的强化学习(RL)时代，CPU 可能比 GPU 更早成为瓶颈。与传统单任务 RL 不同，现代 Agent 系统需 要同时运行成百上千个独立环境实例，"环境并行化"让 CPU 成为事实上的第一块短板。主要源于三大核心逻辑： 1）Multi-Agent 带来 OS 调度压力， Agent 的"推理-执行 ...

投资逻辑： 盈利预测、估值和评级 我们预测， 2025/2026/2027 年 公 司 实 现 营 业 收 入 6.91/12.02/16.23 亿元，同比+40.6%/+73.8%/+35.0%；归母 净 利 润 分 别 为 0.72/1.63/2.42 亿 元 ， 同 比 +7.3%/+125.5%/+48.9%，采用市销率法，给予公司 2026 年 40 倍 PS 估值，目标价 116.06 元，给予"买入"评级。 风险提示 下游应用拓展不及预期风险，原材料价格波动风险，技术路 线迭代风险，地缘政治及汇率波动风险，限售股解禁风险。 0 100 200 300 400 500 600 700 800 23.00 35.00 47.00 59.00 71.00 83.00 95.00 250123 250423 250723 251023 人民币(元) 成交金额(百万元) 成交金额 华曙高科 沪深300 | 公司基本情况（人民币） | | | | | | | --- | --- | --- | --- | --- | --- | | 项目 | 2023 | 2024 | 2025E | 2026E | 20 ...

Investment Rating - The report maintains a "Buy" rating for the company [1] Core Viewpoints - The demand for trendy toys is not a false need; entertainment is a necessity. The "Kidult" trend, driven by demographic changes, media evolution, and consumer psychology, is a long-term trend in the toy industry, exemplified by the increasing market share of companies targeting adult consumers [1][14] - The global entertainment and media industry is projected to reach approximately $3 trillion by 2024, indicating a growing market for diversified entertainment offerings [1][35] Supply-Side Growth Potential - Store Expansion: The company has significant room for growth in retail locations, with a projected increase in global retail stores from 571 in 2025, with a substantial portion located in China [2][58] - Increased Store Traffic: Domestic single-store membership is expected to rise from 145,000 to nearly 200,000, indicating potential for higher customer traffic [2] - Higher Member Spending: The maturity of IPs is expected to drive up member spending, as seen with the SKULLPANDA IP, where the cost to collect all items increased significantly from 5,922 RMB in 2021 to 47,430 RMB in 2025 [2] Demand-Side Competitive Barriers - Artist Talent Barrier: The company has established relationships with mature toy artists, creating a long-term competitive advantage [3] - Marketing Resource Barrier: The company's business model relies on top-tier brand collaborations and celebrity endorsements, which are core competitive barriers [3] - User Data Asset Barrier: The company's direct sales channels enhance its ability to control user data and market feedback, improving operational efficiency [3] Valuation Safety Margin - The company's valuation is expected to be significantly above 10X PE, given its growth stage and the rising influence of China's cultural output on its growth trajectory [3] Profit Forecast and Valuation - The company is projected to achieve revenues of 369.57 billion RMB, 551.72 billion RMB, and 678.11 billion RMB for the years 2025-2027, with adjusted net profits of 123.76 billion RMB, 173.73 billion RMB, and 226.85 billion RMB, reflecting year-on-year growth rates of 295.98%, 40.37%, and 30.58% respectively [4][7] - The target price is set at 359.72 HKD, based on a 25X PE for 2026 [4]

Investment Rating - The report maintains a "Buy" rating for the company [1] Core Views - The demand for trendy toys is not a false need; entertainment is a necessity. The "Kidult" trend, driven by demographic changes, media evolution, and consumer psychology, is a long-term trend in the toy industry, exemplified by the increasing market share of companies targeting adult consumers [1][14] - The global entertainment and media industry is projected to reach approximately $3 trillion by 2024, indicating a growing market for diversified entertainment offerings [1][35] Supply-Side Growth Potential - Store Expansion: The company has significant room for growth in retail locations, with a projected increase in global retail stores from 571 in 2025, with a substantial portion located in China [2][58] - Increased Store Traffic: Domestic single-store membership is expected to rise from 145,000 to nearly 200,000, with the opening of the first U.S. store in September 2023 [2] - Higher Member Spending: The maturity of IPs is expected to drive up the average spending per member, as seen with the SKULLPANDA IP, where the cost to purchase all products increased significantly from 5,922 RMB in 2021 to 47,430 RMB in 2025 [2] Demand-Side Competitive Barriers - Artist Talent Barrier: The company has established partnerships with mature toy artists in Hong Kong, creating a long-term competitive advantage [3] - Marketing Resource Barrier: The company's business model relies on top-tier brand collaborations and celebrity endorsements, which are core competitive barriers [3] - User Data Asset Barrier: The company has strong control over user data and market feedback due to its direct sales channels, enhancing operational efficiency [3] Valuation Safety Margin - The company's valuation is expected to be significantly above 10X PE, given its growth stage and the rising cultural influence of China on the global stage [3] Profit Forecast, Valuation, and Rating - Revenue projections for 2025-2027 are 36.96 billion, 55.10 billion, and 67.74 billion RMB, with adjusted net profits of 12.38 billion, 17.35 billion, and 22.66 billion RMB, reflecting growth rates of 295.98%, 40.21%, and 30.59% respectively [4][7] - The target price is set at 359.72 HKD, based on a 25X PE for 2026 [4]

资产端的"买 A 卖港"与负债端逐步远去的"回本赎回" 2025Q4 主动偏股基金股票仓位回落至 86.30%(A 股 73.96%，港股 12.34%)，且仓位回落均来自港股配置比例的下降， 而 A 股配置比例则延续回升。结合南下来看，四季度公募基金持仓占南下比例下降至 16.46%，且主要源于主动基金配 置比例的下降，而被动基金的占比仍在提升，2025 年末两者的占比已经十分接近。业绩方面，四季度主动基金的业绩 中位数大约为-0.11%，终结了此前连续三个季度为正的趋势，但自 2018 年以来，主动基金单季度持续盈利的最长时 间即为 3 个季度，相应地，约 48%的主动基金跑赢了其业绩基准。因子暴露方面，四季度表现较好的主动偏股基金在 盈利能力因子、成长因子上的暴露显著高于其他分组，同时依然具备高估值、高股价分位数、强动量的特征，这大多 延续了三季度的特征，四季度主动基金因子暴露的主要边际变化在于各基金分组在基本面因子上的暴露差异可能有所 收敛。负债端方面，2025Q4 权益类基金(主动+被动)负债端重获资金净流入，其背后一方面是被动基金负债增量规模 的回升，另一方面是主动基金净赎回规模的明显回落：1）主 ...

Investment Rating - The report maintains a positive recommendation for the industry, indicating an upward trend in fundamentals despite short-term fluctuations in performance [4]. Core Insights - The non-bank sector's public active equity allocation ratio has increased quarter-on-quarter, with a reduction in underweight levels [2]. - The insurance sector has seen a significant increase in active equity holdings, with a quarter-on-quarter growth of 117% in A-share active equity public fund heavy positions [3]. - The brokerage sector's active equity holdings have also risen, with a quarter-on-quarter increase of 14% in A-share active equity public fund heavy positions [5]. - Multi-financial companies, particularly Jiufang Zhituo Holdings, have shown a notable increase in allocation ratios [6]. Summary by Sections Non-Bank Sector - The total heavy stock allocation in the non-bank financial sector reached 40.9 billion yuan, with a configuration ratio of 2.51%, up by 1.03 percentage points quarter-on-quarter, and an underweight ratio of 3.87%, narrowing by 0.92 percentage points [2]. Insurance Sector - The A-share active equity public fund heavy positions in the insurance sector reached 28 billion yuan, with a configuration ratio of 1.72%, up by 0.95 percentage points quarter-on-quarter. The underweight ratio is 0.74 percentage points, down by 0.63 percentage points [3]. - Key individual stock allocations include China Life at 0.08%, Ping An at 1.11%, and others, with varying degrees of over- and under-allocation [3]. Brokerage Sector - The A-share active equity public fund heavy positions in the brokerage sector reached 11.8 billion yuan, with a configuration ratio of 0.73%, up by 0.10 percentage points quarter-on-quarter, and an underweight ratio of 2.67 percentage points, narrowing by 0.29 percentage points [5]. - Major holdings include Huatai Securities and CITIC Securities, with slight increases in their respective ratios [5]. Multi-Financial Sector - Jiufang Zhituo Holdings has seen a significant increase in its allocation ratio, while other multi-financial companies are also recommended for their strong performance potential [6].

—— 我们建议投资者采取哑铃型配置策略：一方面布局具有国资背景、卡位核心频轨资源的系统集成商（确定的贝塔 收益），一方面切入商业火箭与卫星核心零部件及配套环节（如相控阵 T/R 组件、激光通信、3D 打印、测试、核 心网）的民营配套龙头（高弹性的阿尔法收益）。重点关注深度绑定商业火箭/星座供应链的 A 股上市公司，它们 将最先受益于行业从样机研制向流水线生产的范式转移。 中子号延期风险、现金流风险、并购整合风险 敬请参阅最后一页特别声明 1 Rocket Lab 正处于其企业生命周期的关键拐点。作为全球仅有的两家能够同时提供高频轨道发射服务和大规模 航天器制造的私营企业之一（另一家为 SpaceX），Rocket Lab 已超越了单纯的小型火箭发射商的范畴，成功蜕 变为一家垂直整合的端到端太空系统巨头。我们认为其已成功构筑了仅次于 SpaceX 的航天全产业链护城河。 发射业务端：在 SpaceX 几乎垄断全球商业发射市场的背景下，Rocket Lab 的战略价值不仅在于跟随，更在于 制衡与填补。小型发射的绝对统治力与国防属性：公司的 Electron 火箭已成为西方世界小型卫星发射绝对龙 头，其高频次、高可 ...

投资逻辑 商业航天全球产业共振方兴未艾，2026 年中国商业火箭公司将迎来多款中大型火箭的首飞及回收验证，2027 年有望 开始火箭密集发射及卫星大规模组网。我们从一二级产业研究的视角，梳理了商业火箭产业链的各个环节发展现状、 技术方向及投资标的。 商业火箭的发射成功记录以及对应的运载能力能够体现这家公司在未来商业运营中的竞争力。成功完成大载荷或可回 收飞行的商业火箭公司有望在可靠性和成本方面取得先发优势，2026~2027 年是中国商业火箭公司关键的验证窗口 期。从发射历史来看，至今发射成功的商业火箭多为中性和小型固体火箭，天兵科技、蓝箭航天有中大型液体火箭发 射记录。从运载能力来看，在不回收的情况下，800 公里近极轨道载荷能力至少需要做到 2.8 吨才能实现盈利。从回 收经验来看，商业火箭公司中朱雀三号在 2025 年 12 月进行了一子级返回回收场试验，预计 2026 年大部分商业火箭 公司都将进行可回收验证。由于回收可能牺牲载荷运载能力，部分公司可能在发展初期选择以一次性火箭来协助卫星 组网。回收方式方面，宇石空间、大航跃迁、深蓝航天等公司预计进行"筷子"捕获臂回收方式的研发。 发动机的设计能力、推 ...

Investment Rating - The report maintains a positive outlook on the CXO sector for the year 2026, indicating strong demand and recovery in global orders [3][26]. Core Insights - Illumina's release of the Billion Cell Atlas marks a significant advancement in life sciences, transitioning from static genomic data to a comprehensive dataset that integrates AI for drug discovery [3][8]. - Regulatory developments in China and the US are accelerating drug approval processes, with China's implementation of eCTD and the US FDA's support for Bayesian methods enhancing clinical trial efficiency [3][23]. - The CXO sector is expected to see robust growth, supported by major players like Lonza and WuXi AppTec, who are expanding capacity and reporting increased order volumes [3][26]. Summary by Sections Industry Frontiers - Illumina's Billion Cell Atlas is the largest human genome perturbation dataset to date, aimed at accelerating drug discovery through AI [3][8]. - Regulatory dynamics indicate a race for faster drug approvals, with China's NMPA adopting eCTD and the FDA endorsing Bayesian statistical methods for clinical trials [3][23]. - The global order recovery is evident, with a positive outlook for the CXO sector throughout 2026, as major companies report growth and increased demand [3][26]. Capital Trends - GSK's acquisition of RAPT Therapeutics for $2.2 billion focuses on developing an anti-IgE monoclonal antibody for food allergies, highlighting the potential in the allergy treatment market [4][31]. - Novartis has entered a $1.5 billion agreement with SciNeuro to advance a new antibody project targeting Alzheimer's disease, indicating ongoing investment in neurodegenerative treatments [4][36]. Weekly Perspective - The report emphasizes the transition to an AI-driven era in drug discovery, suggesting that previous advantages in research may diminish as the landscape evolves [5][37]. - The demand for CXO services is expected to remain strong, driven by AI-enabled drug discovery and regulatory advancements [5][38].

Report Industry Investment Rating - Not provided in the content Core Viewpoints - The primary market for science and technology innovation bonds continues to grow at a high - speed. The secondary market's performance is closely tied to the capital flow of science and technology innovation bond ETFs, with short - term bonds expected to remain volatile. High - grade medium - and short - term bonds in hard - tech fields such as communications and pharmaceuticals will continue to receive liquidity premiums [2][4] Summary by Directory I. How Aggressive are Science and Technology Innovation Bonds? 1. Primary Issuance Scale and Structure - The primary market for science and technology innovation bonds maintains a high - speed growth trend. From January 12 to January 16, 2026, the new supply scale of science and technology innovation bonds reached 58.92 billion yuan, with bonds with a term of less than 1 year accounting for about 60%. The overall market demand for new science and technology innovation bonds has recovered, but the subscription enthusiasm is still lower than that of non - science and technology credit bonds [2][12] 2. Secondary Trading Activity and Pricing - **Rating and Industry Distribution**: The ratings of outstanding science and technology innovation bonds are highly concentrated. Bonds with an implied rating of AA+ and above account for 73.3%, and AA - rated medium - quality individual bonds account for 22.4%. The industry distribution is dominated by traditional industries, and bonds in construction decoration, public utilities, and comprehensive industries account for 38.7%. Textile and apparel, pharmaceutical biology, communications, power equipment, and non - banking finance industries have an excess spread of over 9bp compared to all credit bonds in the industry [3][19] - **Liquidity**: In late December 2025, the science and technology innovation bond ETF had a volume - boosting market, which strongly pushed up the turnover rate of component bonds. However, at the beginning of 2026, continuous capital outflows caused the turnover rate of science and technology innovation bonds to drop to 1.65% [3][26] - **Yield**: Due to the partial alleviation of negative factors affecting the bond market such as the "seesaw effect" of the stock market and the disappointment of monetary easing expectations, the average yield of high - grade science and technology innovation bonds on the exchange with a term of 1 - 3 years has dropped to 1.94%. The price of 3 - 5 - year bonds fluctuated in a narrow range, which is related to the ETF capital outflows during the week [3][29] - **Internal Pricing of Bond Types**: In the latest week, the spread between the component bonds and non - component bonds of the science and technology innovation bond index has marginally increased to 20.2bp, and the spread between inter - bank bonds and component bonds has also widened to 5.1bp. For 1 - 3 - year bonds, there is still a compression space of 17bp between inter - bank bonds and index component bonds [4][33]