Core Insights - Jefferies indicates that Chinese cloud service providers, AI software companies, and data center operators are undervalued compared to their U.S. counterparts [1] - Alibaba's Qianwen has surpassed 100 million monthly active users within two months of launch, showing rapid growth among students and white-collar workers [1] - The cloud computing ETF Huatai-PineBridge (159273) has seen most of its weighted stocks perform positively, with notable gains from Runze Technology (over 7%), Hengsheng Electronics (over 5%), and Alibaba-W (over 3%) [1] Company News - Alibaba's Qianwen is set to release a significant product iteration this Thursday [1] - Alibaba-W stock has increased by 3.94%, with a market capitalization of 315.11 billion [2] Market Trends - The AI computing sector is experiencing strong performance, with the cloud computing ETF Huatai-PineBridge (159273) rising over 2% and achieving a record high trading price [3] - Jefferies notes that Chinese AI stocks have room for further growth due to increased capital expenditure, improved AI model performance, and favorable policy signals [3] - The AI industry in China is still less mature in monetization compared to the U.S., leading to greater potential for valuation re-rating [3] Industry Analysis - The AI industry is transitioning from a "computing power race" to a focus on "application landing," indicating a maturation of business models [5] - The IDC market is undergoing a supply-demand shift driven by improved chip supply and surging demand for domestic AI applications [7] - The communication industry is experiencing a rotation towards IDC and computing power infrastructure, with expectations of order recovery and performance realization as capital expenditures from major firms become clearer [8]

消息面上，近日，美国投行杰富瑞公司指出，中国的人工智能股票仍有进一步上涨的空间，这得益于资本支出的增加、AI模型性能的提升以及有 利的政策信号。中国的人工智能行业在货币化方面的成熟度仍低于美国，这导致其迄今的估值重估幅度更大，但同时也为未来进一步上涨留下了 空间，因为中国AI企业的盈利水平正在追赶上来。 杰富瑞还强调中美在AI技术方面的差距正在缩小。中国和美国的领先人工智能模型之间的性能差异已从此前的8%缩小至约6%，这得益于诸如北 京的大模型明星企业智谱研发的GLM-4.7模型等新产品的推出，这提升了国内大语言模型的竞争力。 在细分领域层面，杰富瑞表示，中国的云服务提供商、人工智能软件公司以及数据中心运营商的股价相对于美国同行存在折价。 公司新闻方面，上线两个月，阿里巴巴千问C端月活跃用户数（MAU）已突破1亿，在学生和白领人群中增长迅猛。据悉，千问本周四将发布重磅 产品迭代。 今日（1.14），AI算力板块再度强势，云计算ETF汇添富（159273）放量涨超2%，成交价盘中再创历史新高，成交额近5000万元！ （来源：中银证券20260112《从"算力竞赛"到"应用落地"》） 【IDC增量变化：从芯片供给 ...

从具体盘面来看，电商概念板块内，美登科技（920227）斩获"30CM"涨停、凯淳股份（301001） （301001）、光云科技（688365）、值得买（300785）（300785）均斩获"20CM"涨停，生意宝 （002095）（002095）、嘉美包装（002969）（002969）、新华都（002264）（002264）等多股亦涨 停。 人工智能板块同样热度拉满，拉卡拉（300773）（300773）、通达海（301378）（301378）、三维天地 （301159）（301159）、友车科技（688479）、佳缘科技（301117）（301117）等多股收获20%涨停。 《证券日报》记者在采访中了解到，此番两大板块的联袂走强，核心源于上市公司基本面改善、年货节 消费旺季催化、AI技术落地赋能的三重驱动，更叠加政策端重磅利好的共振支撑。 企业端，不少公司已通过"All in AI"战略实现业绩突破，如豪鹏科技（001283）布局AI端侧硬件应用领 域，2025年年度业绩预告显示，公司营收同比增长11.58%至17.45%，净利润同比增幅更是达到113.69% 至141.09%；电商领域的光云科技、值 ...

针对美国企业家马斯克旗下人工智能企业xAI的聊天机器人"格罗克"涉嫌生成非法色情内容，法国巴黎 检方本月2日表示将启动调查。法国数字科技记者、作家玛蒂尔德·萨利乌在接受总台记者采访时表示， 应该追究xAI公司的责任，保护公民权益。 ...

1月14日消息，据21世纪经济报道援引私募排排网最新数据，幻方量化2025年的收益均值达56.55%，在 中国管理规模超百亿的量化私募业绩榜中位列第二，仅次于以73.51%收益均值登顶的灵均投资。目 前，幻方量化管理规模已超700亿元。 而据私募排排网数据，幻方量化近三年的收益均值为85.15%，近五年的收益均值为114.35%。有分析认 为，幻方量化的丰厚业绩为梁文锋旗下的DeepSeek提供了充足的研发资金。 幻方量化是国内最为知名的量化私募巨头之一，由创始人梁文锋在2008年于浙江大学学习信息与通信工 程时创立，是一家有数学、计算、研究和AI基因的对冲基金。2019年幻方量化管理规模破百亿元， 2021年一度突破千亿元。 幻方量化在AI领域布局已久。2016年10月，公司第一个由深度学习算法模型生成的股票仓位便已上线 实盘交易，至2017年底公司几乎所有的量化策略都已经采用AI模型计算。2019年12月，幻方AI即杭州 幻方人工智能基础研究有限公司正式成立，致力于AI的算法与基础应用研究，随后公司斥巨资打造 的"萤火一号"和"萤火二号"陆续交付。 2023年4月中旬，幻方量化宣布将集中资源和力量，全力投 ...

Group 1 - The event "Bay Area Engine: Central-Local Integration" aims to deepen strategic cooperation between the Guangdong-Hong Kong-Macao Greater Bay Area and central enterprises, fostering new productive forces and building a modern industrial system [1] - Guangdong's economy has consistently surpassed significant milestones, with a projected GDP of 14.16 trillion yuan for 2024, maintaining its position as the top province in China for 36 consecutive years [1][2] - The province has formed nine trillion-yuan industrial clusters and has the highest innovation index globally for the "Shenzhen-Hong Kong-Guangzhou" innovation cluster [1] Group 2 - Guangdong has a population of 160 million and over 20.22 million market entities, covering all 31 manufacturing categories, with R&D investment and other innovation metrics leading the nation [2] - Guangzhou is positioned as a strategic hub for central enterprises, particularly in sectors like intelligent connected vehicles and biomedicine, linking global markets [2] - The Guangdong government is establishing a large-scale industrial investment fund to support the growth of central enterprises, enhancing the investment landscape [2] Group 3 - From 2019 to 2024, Guangzhou's equity investment market has seen the establishment of 852 funds, raising 215.43 billion yuan, indicating strong investment attraction [3] - The city has recorded 2,661 investment cases with a total investment amount of 311.32 billion yuan, showcasing its capital aggregation capabilities [3] - The event included project roadshows and the launch of the Central Enterprise Investment Association's special committee, marking a new phase in central-local cooperation [3]

资讯日报：美国 12 月核心 CPI 升幅低于预期 股票市场概览 港股市场表现 美股市场表现 2026 年 1 月 14 日 资讯日报 海外市场主要股市上日表现 | 指数 | 收盘价 | 日常跌 | 間微跌 | 年初至今 STATE | | --- | --- | --- | --- | --- | | | | (%) | (%) | | | | | | | (%) | | 恒生指数 | 26.848 | ---- 0.90 | 2.35 | 4.75 | | 恒生科技 | 5.870 | 0.11 | 3.21 | 6.41 | | 恒生国企 | 9.285 | ... 0.71 | 2.62 | 4.17 | | 上证指数 | 4.139 | (0.64) | 0.44 | 4.28 | | 日经225 | 53,549 | 3.10 | 3.10 | 6.38 | | 新加坡海峡 | 4.807 | 0.85 | 1.32 | 0.00 | | 标普500 | 6.964 | (0.19) | (0.04) | 1.73 | | 纳斯达克 | 23,710 | (0.10) | 0.16 | 2.01 ...

Group 1: AI Integration and Hardware Innovation - AI remains the main theme of CES 2026, with a significant shift from software to physical world integration, empowering various AI terminals [1] - Generative AI is driving a new wave of innovation, with AI smartphones and AI PCs expected to benefit from consumer upgrade cycles, entering an upward trend [1] - AI glasses and embodied intelligent robots have transitioned from the "0 to 1" stage to the "1 to N" stage, with product shipments anticipated to maintain rapid growth [1] Group 2: AI Chip Development - Major tech companies like NVIDIA, AMD, Intel, and Qualcomm showcased advancements in AI chips at CES 2026, with continuous iteration in chip architecture and manufacturing processes [1] - NVIDIA announced the full production of the Vera Rubin platform, integrating six key chips and achieving a tenfold reduction in AI inference costs [1] - AMD's Helios platform, based on MI455X, demonstrated a tenfold performance improvement over previous generations, gaining recognition from leading clients like OpenAI [1] Group 3: Traditional Consumer Electronics - Lenovo introduced innovative AI PCs with rollable screens and foldable designs, enhancing AI interaction experiences through its "hybrid AI" strategy [2] - Dell and HP upgraded their flagship AI PC products, significantly improving edge AI performance [2] - Samsung and Honor showcased advancements in AI smartphones, with Samsung presenting a foldable OLED display and Honor unveiling the world's first robot phone [2] Group 4: New AI Terminals - New AI terminals, particularly AI glasses, are innovating rapidly, with Chinese brands leading the charge [2] - Thunderbird showcased the world's first eSIM smart AR glasses, marking a shift towards independent communication for AR devices [2] - AI models are moving beyond software, enabling personalized applications in emotional companionship, health, and pet care, indicating a new phase for the AI terminal industry [2] Group 5: Automotive and Robotics - NVIDIA's open-source Alpamayo system is lowering the development threshold for Level 4 autonomous driving, while Chinese automakers like Geely and Great Wall are advancing full-stack autonomous driving solutions towards mass production [3] - Innovations in mobility are diversifying, with products like the Strutt smart wheelchair and Verge solid-state battery electric motorcycle reshaping future travel experiences [3] - Embodied intelligence is gaining traction, with NVIDIA's Jetson Thor chip and Isaac platform fostering an open ecosystem, while Boston Dynamics and LG are pushing for large-scale industrial applications [3]

电子商务概念股早间全线走强，板块指数低开高走，放量飙升逾7%，创4年来新高。 年货节将至 电子商务走强 板块内所有个股上涨，值得买（300785）仅约13分钟直线20%涨停，3日2涨停，凯淳股份亦直线快速20cm封 板，壹网壹创、丽人丽妆、焦点科技等也强势涨停或涨超10%。 | 值得买(300785) 2026年01月14日 星期三 PageUp/Down:前后日 空格链:操作 | | | | | --- | --- | --- | --- | | 值得买 分时(静态) | 20.00% | 11:26 | 83.89 | | 68.89 | | 11:26 | 83.89 | | 81.89 | 17.14% | 11:26 | 83.89 | | 79.90 | 14.28% | 11:26 | 83.89 | | 77.90 | 11.43% | 11:26 | 83.89 | | 75.90 | 8.57% | 11:26 | 83.89 | | 73.90 | 5.71% | 11:26 | 83.89 | | 71.91 | 2.86% | 11:26 | 83.89 | | 69.91 | 0. ...

作者： 上交博士，在腾讯codebuddy 实习，发表一作顶会顶刊论文10篇（含best paper 等），开源PFLlib等明星项目，获得社区赞誉。主要研究AI强化 学习、AI合成数据、Agent 记忆等。 在 AI 辅助 Coding 技术快速发展的背景下，大语言模型（LLMs）虽显著提升了软件开发效率，但开源的 LLMs 生成的代码依旧存在运行时错误，增加了 开发者调试成本。 现有基于偏好优化的改进方法，多依赖「通过 / 失败」二元信号构建训练数据，难以知晓「错在哪」，也忽视了模型能力在训练时的动态变化特性。 针对此缺口，在腾讯 CodeBuddy 实习期间，我们提出自适应渐进式偏好优化方法（AP2O），并构建 AP2O-Coder 框架。该方法借鉴人类的「按题型高 效刷题」经验出发，通过「考试 - 分析 - 纠错 - 小测」的系统性流程提升模型代码纠错能力，在多款主流开源模型上实现最高 3% 的 pass@k 性能提 升，同时降低训练数据需求量。 论文标题：AP2O-Coder: Adaptively Progressive Preference Optimization for Reducing C ...