科尔黛伦矿业涨3.5%，Vista黄金涨2.8%，First Majestic Silver、Endeavour Silver涨2.6%，埃氏金业涨 2.4%，纽曼矿业、泛美白银涨超2%，金田、哈莫尼黄金涨1.2%。消息面上，贵金属价格大涨，现货黄 金涨超2%，站上4980美元/盎司；现货白银涨超5%，站上77美元/盎司。(格隆汇) ...

美国白宫国家经济委员会主任哈塞特称，美联储还有很大的降息空间。美股三大指数集体高开，纳指涨 0.23%，标普500指数涨0.24%，道指涨0.16%。 (格隆汇) 英伟达涨近2%，Meta与英伟达建立多年战略合作伙伴关系，将部署数百万颗英伟达芯片。 西部数据涨2.8%，闪迪跌2.4%，西部数据拟出售部分闪迪股份，筹资31.7亿美元用于削减债务。 网络安全公司Palo Alto Networks跌9.4%，下调全年调整后每股盈利指引。 ...

3. 伯克希尔哈撒韦连续第三个季度减持美国银行和苹果，增持纽约时报。 1. 美国三大股指期货齐涨，纳指期货涨0.53%，标普500指数期货涨0.41%，道指期货涨0.26%。 2. 欧股多国股指刷新历史新高，截至发稿，德国DAX指数涨0.83%，英国富时100指数涨0.97%，法国 CAC指数涨0.45%，欧洲斯托克50指数涨0.84%。 4. 段永平试水"AI交易"：Q4卖苹果加仓663.93万股英伟达，建仓CoreWeave、Credo Technology、 Tempus Ai。 5. 英伟达和Meta宣布建立多年战略合作伙伴关系，Meta将部署数百万颗英伟达芯片。 6. 微软计划未来10年向"全球南方"人工智能领域投资500亿美元。 7. 谷歌I/O开发者大会定档5月19至20日，预计将发布Gemini大模型及其他AI产品更新。 8. 谷歌计划在美国和印度之间建设新的光纤线路，以提高网络连接速度和可靠性。 9. 特斯拉避免加州30天销售禁令，自动驾驶功能营销实现合规。 10. 英伟达清仓所持Arm股份，套现约1.4 亿美元。 11. 西部数据拟出售部分闪迪股份，筹资31.7亿美元用于削减债务。 12. ...

近期，华尔街13F报告陆续出炉，金融巨鳄们最新动向最新大曝光。 其中，股神巴菲特在最后任期继续减持苹果、美银等重仓股；高瓴HHLR四季度增持阿里、拼多多等；此前，桥水Q4增持 英伟达、亚马逊等AI热门股。 巴菲特收官"之作" 当地时间2月17日美股盘后，伯克希尔向美国证监会（SEC）披露了最新13F文件。 在巴菲特担任CEO的最后一季度，伯克希尔继续调整科技巨头持仓，减持多只重仓股的同时，首次建仓传统媒体纽约时 报。 截至2025年四季度末，伯克希尔的十大重仓股分别是：苹果、美国运通、美国银行（核心股）、可口可乐、雪佛龙、穆迪 （排名从第七升至第六）、西方石油（排名从第六降至第七）、瑞士安达保险、卡夫亨氏、谷歌母公司Alphabet。 伯克希尔连续第三个季度减持苹果。 第四季度抛售苹果超过1029万股，持股数环比第三季度减少4.32%，季末苹果持仓市值环比减少约28亿美元，持仓在伯克希 尔投资组合中的占比从三季度的22.69%降至22.60%。 以环比减少的市值看，苹果是伯克希尔2025年第四季度减持最多的个股。 纽约时报是伯克希尔四季度唯一新建仓的个股。 伯克希尔买入数量为506.7万股，持仓市值为3.52 ...

与过去需要在多个APP之间比价、查攻略不同，年轻人更青睐"一句话下单"的便捷体验：在旅游方面， 他们让千问帮买最多的景区门票是峨眉山；在观影方面，他们最喜欢让千问帮买《惊蛰无声》电影票； 在日常餐饮中，00后们通过千问下单最多的仍是奶茶。 在购票、出行等需要多步骤判断的场景中，00 后的活跃度更远超其他年龄层。仅景区门票预订一项，就有50%的一句话指令来自00后。 目前，千问春节活动仍在延续，面向所有人推出"每日首单必减"，最低3.8元，同时还能叠加淘宝闪 购、飞猪、大麦等平台的原有优惠，持续推进AI生活成为日常。 千问APP今日公布，通过千问"一句话下单"的用户中，00后占比高达40%，成为最活跃的群体，其中来 自广东、山东、河南三地的00后最为突出。在点外卖、订酒旅、买电影票核心消费场景中，00后占比均 位居首位，"千问帮我"正从尝鲜体验成为年轻人的日常习惯。 半个月前，千问推出"春节大免单"活动，火爆全网。活动期间，全国有超过1.3亿人第一次体验AI下 单，说了50亿次"千问帮我"，千问一跃成为国民级AI助手。其中00后作为数字原住民，率先拥抱这一AI 时代的新生活方式，占比高达40%，特别是广东、山东 ...

格隆汇2月18日｜千问3.5除夕开源发布，在全球AI产业链掀起热潮。截至2月18日，英伟达、AMD、苹 果公司等国际硬件厂商，已实现开发框架乃至芯片层对Qwen3.5的适配；华为昇腾、摩尔线程、沐曦、 海光等国产GPU和平台均官宣Day 0 适配千问3.5新模型。千问3.5总参数3970亿仅激活170亿，性能超越 上代万亿参数模型Qwen3-Max，部署显存占用大降60%，推理最大吞吐提升至19倍。基于底层模型架构 创新，Qwen3-Plus在阿里云百炼API价格低至每百万Token0.8元，以不到5%的低价格获得媲美Gemini 3 的高性能。 千问3.5以小胜大，创下大模型效率和性价比新高，又因其开源而一举成为三方平台争先抢 上的最热门模型。国家超算互联网平台、曙光云、算能Sophnet、OpenRouter、Together AI、Ollama、 ZenMux、Novita AI、Poe等国内外平台，均已第一时间上线Qwen3.5模型，对外提供API服务或体验入 口。 ...

Investment Highlights - Company is a global leader in small nucleic acid drug development with a proven platform since its establishment in 2007 [1] - The company has developed the GalNAc liver-targeting RiboGalSTAR platform and is expanding from liver-targeting to extrahepatic targeting [1] - The company has a comprehensive intellectual property portfolio with 473 patents and patent applications as of the end of 2025 [1] - The research platform integrates the entire technology chain, ensuring efficiency from CMC production to commercialization [1] Clinical Assets - Company is one of the Chinese small nucleic acid enterprises with the most clinical assets, having 7 siRNA pipelines in clinical stages by the end of 2025 [2] - Core assets include RBD4059 (FXI), RBD5044 (ApoC3), and RBD1016 (HBV), with RBD4059 being the fastest progressing FXI siRNA drug for thrombotic diseases [2] - RBD4059 is expected to initiate Phase IIb clinical trials in 2026, targeting multiple thrombotic diseases [2] - Company believes that FXI inhibitors have validated drugability in conditions like stroke and knee replacement surgery, with potential for atrial fibrillation treatment [2] Financial Forecast and Valuation - Company forecasts EPS of -1.79, -2.21, and -2.92 for 2025 to 2027 [2] - The company is rated as outperforming the industry with a target price of 100.0 HKD, indicating a 44.9% upside from the current stock price [2]

Investment Highlights - Company MINIMAX-WP (00100) is rated as outperforming the industry with a target price of HKD 1,109.00, based on a P/S valuation method corresponding to a 75x multiple for 2027 [1] - The company is one of the few domestic firms that successfully integrates foundational model capabilities with global AI-native application commercialization, showcasing its rarity in the AI sector [1] Company Positioning - MINIMAX-WP is one of the earliest companies in China to bet on the native full-modal fusion route, advancing text, speech, and video model development simultaneously since its inception [2] - The company has demonstrated the ability to monetize overseas market scale, with over 73% of revenue from international markets covering more than 200 countries and regions in the first three quarters of 2025 [1][2] - The firm employs a "technology as product" strategy, utilizing a "front store and back factory" model to accelerate the alignment and iteration of models and products [1] Organizational Efficiency - The company maintains a high talent density and efficiency, embedding AI capabilities deeply into internal workflows, with 73.8% of its 385 employees in R&D as of Q3 2025 [2] - The organizational structure is flat, with no more than three levels below the CEO, supporting efficient model and product iteration [2] Competitive Advantage - The company possesses a unique advantage in the integrated understanding and generation of multi-modal models, emphasizing technology as a product, globalization, and high efficiency [2] Financial Projections - Revenue projections for the company are estimated at USD 71 million, USD 228 million, and USD 598 million for the years 2025, 2026, and 2027 respectively, with a CAGR of 170% from 2024 to 2027 [2] - The current trading P/S ratio is 57x for 2027, indicating a potential upside of 31% compared to the target price [2]

Core Viewpoint - The Federal Reserve must analyze data to determine if artificial intelligence is driving productivity growth, enabling faster economic growth without triggering inflation or necessitating policy tightening [1] Group 1: Economic Implications - The Trump administration believes that AI investment is already contributing to productivity increases, potentially creating an economic environment similar to the 1990s tech boom, characterized by moderate inflation and accelerated growth [1] - Daly noted that most macro productivity studies have found limited evidence of significant impacts from AI so far, suggesting that improvements from sector-specific investments may take time to manifest [1] Group 2: Future Considerations - There is a possibility that the economy has not yet reached a critical point where comprehensive changes from AI are evident, indicating that substantial economic transformation may require a longer timeframe to materialize [1]

Group 1 - Tiger Global Management and Adage Capital Partners reduced their holdings in major AI companies, including Nvidia and Amazon, due to concerns over high valuations and insufficient returns on AI investments [1] - Tiger Global decreased its Microsoft shares from 6.5 million to 5.47 million, maintaining a stake valued at $2.6 billion, making it one of the largest investors in Microsoft [2] - The firm also reduced its Amazon holdings by 9.35%, bringing its total shares to 10 million, valued at $2.3 billion, and decreased its Nvidia position to a value of $2.1 billion [2] Group 2 - Adage Capital reported minor reductions in its stakes in Microsoft, Alphabet, Amazon, and Nvidia, with decreases ranging from 1% to 3% [2] - Adage increased its Oracle holdings by approximately 19%, reaching 1.87 million shares valued at about $365 million [2]