市场有风险，投资需谨慎。本文不构成个人投资建议，也未考虑到个别用户特殊的投资目标、财务状况或需要。用户应考虑本文中的任何 意见、观点或结论是否符合其特定状况。据此投资，责任自负。 据私募排排网，美国证券交易委员会(SEC)官网披露了但斌掌舵的东方港湾旗下海外基金2025年四季度 的美股持仓数据。数据显示，2025年末东方港湾共持有10只美股标的，持股市值合计13.16亿美元，折 合人民币超91亿元，相比2025年三季度末约12.92亿美元的持股市值略有增长。四季度，东方港湾清仓 了Coinbase、奈飞、AsteraLabs、BitMineImmersionTechnologies、博通、台积电等美股标的。此外，变 化较大的是，谷歌取代英伟达，跃升为东方港湾的第一大重仓股。 风险提示及免责条款 ...

Group 1 - SoftBank Group is in discussions to invest up to $30 billion in OpenAI, which is seeking up to $100 billion in new capital [1][4] - SoftBank holds approximately 11% stake in OpenAI after a recent $22.5 billion investment [2] - The company's shares rose over 8% following the news, later stabilizing at a 3.7% increase [1] Group 2 - SoftBank is actively repositioning itself around artificial intelligence, having sold stakes in Nvidia and T-Mobile to fund its investment in OpenAI [3] - OpenAI requires substantial funding for AI model development, computing needs, and talent retention, with a potential valuation of up to $830 billion if the funding is successful [4] - OpenAI plans to slow down hiring due to advancements in AI that allow for increased output with smaller teams [5]

1月28日，美国证券交易委员会(SEC)官网披露了但斌掌舵的东方港湾旗下海外基金"Oriental Harbor Investment Fund"2025年四季度的美股持仓 数据。 0 1 东方港湾美股持仓超91亿，谷歌取代英伟达成第一重仓股 本文首发于公众号"私募排排网"。 （点击↑↑ 上图查看路演回放详情 ） 数据显示， 2025年 末 东方港湾 共 持有 10 只美股 标的 ，持股市值合计 13.16 亿美元， 折合 人民币 超 91 亿元 ，相比2025年三季度 末约 12.92亿美元的 持股市值略有增长。 2025年四季度，东方港湾在美股进行了一定幅度的调仓换股，更加聚焦美股的科技巨头。 四季度，东方港湾清仓了Coinbase、奈飞、Astera Labs、BitMine Immersion Technologies、博通、台积电等美股标的。 此外，变化较大的是， 谷歌取代英伟达，跃升为东方港湾的第一大重仓股 。此前多个季度，东方港湾的第一大重仓股为英伟达。 2025年四季度， 东方港湾加仓了谷歌约 40.55%的股份，叠加该股在四季度涨幅接近29%，从而导致谷歌的持仓占比大幅提升，并远高于对英 伟 ...

1月28日上午，英伟达创始人兼CEO黄仁勋现身深圳南山区的英伟达半导体(深圳)有限公司。(英伟达深圳公司长啥样？深圳商报记者实地探访) 经过一上午的闭门会议，中午12时，黄仁勋离开办公室，继续他的2026年中国之行。 图片来源深圳特区报 最后一买单，发现仅需800多元，黄仁勋对大伙说："深圳的牛肉，很新鲜，很好吃！" 这已经是他连续第三年在中国传统年关之际到访中国，此前分别到访了上海、北京。 在深圳商报微信公众号每日推送的【任意 篇微信头条稿件】(下方留言) 商报君精选留言后,随机抽取当天任一头条的留言 点赞蜀高者 获得福利 + 颜色随机发放哦~ 留言点赞统计截至留言次日上午10:00 届时商报君将私信收集领奖地址。 数量有限, 抓紧时间! 111 据此前媒体报道，1月27日黄仁勋带着6位朋友来到深圳福田CBD，吃了一顿深圳打工人最爱的牛肉火锅！ 来源深圳商报、深圳特区报 编辑吴怡漪责编黎莉校审谭录岗 ...

Company Performance - Seagate Technology experienced a significant stock increase of 19.1% after reporting a larger profit than analysts expected, driven by demand from artificial intelligence applications [2] - Nvidia's stock rose by 1.6%, contributing positively to the S&P 500, supported by a favorable report from ASML regarding chip-making machinery [3] - Amphenol's stock fell by 12.2% despite reporting stronger profits than forecasted, following a prior surge of 23% in the year [5] - Apple’s stock decreased by 0.7% ahead of its upcoming profit report, making it the heaviest weight on the S&P 500 due to its size [6] Market Overview - The S&P 500 index saw a minor decline of 0.57 points, closing at 6,978.03, while the Dow Jones Industrial Average rose by 12.19 points to 49,015.60, and the Nasdaq composite increased by 40.35 points to 23,857.45 [7] - The U.S. dollar stabilized after a recent decline, with Treasury Secretary Scott Bessent stating that the government is not intervening in the currency market and aims for a "strong dollar" [7] Industry Trends - Companies are under pressure to deliver solid profit growth following record stock price increases, as stock prices typically align with corporate earnings over the long term [6] - ASML's optimistic revenue forecast for 2026 has alleviated concerns about a potential bubble in AI demand, indicating sustained interest in AI technologies [4]

Core Insights - The article highlights a significant surge in AI-related stocks, particularly with Jinji Co., which saw a 20% limit up, and the Clawdbot AI Agent project gaining immense popularity with over 58,000 stars on GitHub within a month, marking a "lobster moment" for AI [2][3] Price Adjustments in Cloud Services - Google Cloud announced a price increase for global data transmission services starting May 2026, with North American rates doubling from $0.04 to $0.08 per GB, and similar increases in Europe and Asia [3] - This price adjustment reflects a shift in the long-standing trend of decreasing prices in the cloud services industry, driven by rising demand for AI computing power and increasing upstream costs [3] AI Agent Project Clawdbot - Clawdbot has rapidly gained traction, with a daily increase of 62% in GitHub stars and nearly 9,000 members in its Discord community, being compared to a "ChatGPT moment" for 2026 [3][4] - The project has received endorsements from notable figures in the tech industry, further fueling discussions and interest [3] Impact on AI Industry - Analysts predict that as AI integrates into workflows, the demand for tokens will experience explosive growth, benefiting the entire AI industry chain [4] - The recent price hikes in cloud services are expected to contribute to the growth of the computing rental sector, which has already seen significant stock price increases [6] Developments in Cloud Computing - Amazon Web Services (AWS) announced a 15% price increase for its EC2 machine learning capacity blocks, marking the first price adjustment in nearly 20 years [6] - NVIDIA's investment of $2 billion in CoreWeave aims to enhance AI computing capacity, indicating a robust demand for AI cloud infrastructure [6] Advancements in AI Models - Alibaba launched its Qwen3-Max-Thinking model, boasting over 1 trillion parameters and outperforming several leading models in key performance benchmarks [7] - The pricing strategy of AWS reflects the high demand for AI computing resources, while domestic giants like ByteDance, Alibaba, and Tencent are expected to significantly boost AI cloud infrastructure demand [7]

Core Viewpoint - Semiconductor stocks are expected to gain further as investments in AI infrastructure accelerate, with major tech companies spending hundreds of billions on data centers and AI chip technology [1]. Group 1: Nvidia - Nvidia is a leading choice due to its dominance in GPUs and CUDA software, which are essential for generative AI [3]. - AI processing semiconductor revenue surpassed $200 billion last year, with a projected CAGR of 14% through 2033, reaching a total addressable market of $486 billion; Nvidia is expected to maintain a 75% market share through 2030 [4]. - Nvidia is expanding its business lines, including a $20 billion partnership with Groq for inference capabilities, which may enhance its leadership in the AI chip market [5]. - Despite its strong position, Nvidia's stock is trading at its lowest valuation in over a year, indicating potential investor concerns about competition from AMD and Broadcom [6]. - The AI chip market is still expanding, allowing for multiple winners, and Nvidia is recommended as a buy-and-hold investment for long-term growth [7]. Group 2: Taiwan Semiconductor Manufacturing (TSMC) - TSMC is the largest chip manufacturer globally, holding nearly 70% market share, and serves as a key partner for companies like Nvidia and AMD [8][9]. - TSMC's diverse manufacturing capabilities position it well to benefit from the growing AI chip market, as it can produce various chip types [10]. - The company is set to capture significant market share in the AI chip sector as hyperscalers increase their capital expenditures on AI infrastructure [11]. - TSMC is investing in new manufacturing facilities and expanding its geographic presence, indicating potential for further growth in the ongoing chip supercycle [12]. Group 3: Micron Technology - The rise of large language models and generative AI has created a demand for high-bandwidth memory (HBM) and storage solutions, benefiting Micron Technology [13]. - Micron's revenue in its DRAM division increased by 69% year-over-year, with NAND sales rising by 22% in the fiscal first quarter of 2026 [14]. - Micron's earnings per share (EPS) are projected to triple this year, indicating strong pricing power in the memory and storage market [15]. - Despite its strong outlook, Micron is undervalued with a forward P/E of 11, suggesting it may be an attractive investment opportunity [16].

机器之心编辑部 在技术如火如荼发展的当下，业界常常在思考一个问题：如何利用 AI 发现科学问题的新最优解？ 一个普遍的解法是「测试时搜索」（Test-time search），即提示一个冻结的（不更新参数的）大语言模型（LLM）进行多次尝试，这一点类似人类在做编程作业时 的「猜」解法，尤其是进化搜索方法（如 AlphaEvolve），会将以往的尝试存入缓冲区，并通过人工设计、与领域相关的启发式规则生成新的提示。 可是，尽管这些提示能够帮助 LLM 改进以往的解法，但 LLM 本身并不会真正提升，就像一个学生始终无法内化作业背后的新思想一样。 实际上， 能够让 LLM 真正进步的最直接方式是学习。 尽管「学习」和「搜索」都能随着算力扩展而良好地增长，但在 AI 的发展历史中，对于围棋、蛋白质折叠等这类困难问题，「学习」往往最终超越了「搜索」。 因为， 科学发现本质是：超出训练数据与人类现有知识的 out-of-distribution 问题。 为此， 斯坦福大学、英伟达等机构联合提出一种新方法：在测试时进行强化学习（RL），即让 LLM 在尝试解决特定测试问题的过程中持续训练自己。 论文链接：https://w ...

1月28日，全球最高市值公司英伟达的CEO黄仁勋出现在位于深圳市南山区的英伟达半导体（深圳）有限公司（以下简称"英伟达深圳公司"）楼下，身 穿黑色夹克和黑色裤子。 证券时报记者在现场了解到，黄仁勋在上午参加完英伟达深圳公司的活动后，于中午12时左右下楼，与现场人群互动并签名，随后黄仁勋乘车离开。 此 前在2025年1月，黄仁勋曾在深圳公司的年会上派发"万元大红包"，包括十多个1万元和2.5万元的红包，以及五个4万元的红包，当时黄仁勋在现场表示， 下次到访时红包金额将达到10万元。 据了解，本次"中国行"黄仁勋已经陆续到访上海、北京、深圳三大城市，他分别参加了三地分公司举办的新年活动，并参与面向核心供应商的答谢活动。 除了参加年会外，本次中国行的主题是"逛吃"。在上海，黄仁勋出席了公司年会并到菜市场买菜，在北京他打卡了北京王府中环半山腰品尝云南菜，在深 黄仁勋在书上签名。 宋春雨/摄 圳他吃了八合里牛肉火锅。 不过证券时报记者观察到，在黄仁勋访华期间，一条隐秘的"周边经济链"也悄然浮现——25000元"买合影"、99999元买"签名红包"，黄仁勋中国行正在成 为一部分人的"致富手段"。 现身北上深，餐厅店员称" ...

1月28日，全球最高市值公司英伟达的CEO黄仁勋出现在位于深圳市南山区的英伟达半导体（深圳）有 限公司（以下简称"英伟达深圳公司"）楼下，身穿黑色夹克和黑色裤子。 黄仁勋在书上签名。 宋春雨/摄 图源："八合里潮汕鲜牛肉火锅"小红书账号 吃了牛肉火锅之后，黄仁勋夸赞深圳的牛肉"很新鲜，很好吃"，并在三个餐盘上进行了签名留念。1月 28日上午11时，八合里店员向证券时报记者表示："黄仁勋昨天点了牛骨清汤锅底，还有匙肉、肥牛等 招牌菜。这个点我们刚营业，暂时还没有感受到'芯片大佬'带来的热度，中午可能人会更多一些，已经 有不少打电话来问黄仁勋'同款菜'的顾客了。" 图源："八合里潮汕鲜牛肉火锅"小红书账号 证券时报记者在现场了解到，黄仁勋在上午参加完英伟达深圳公司的活动后，于中午12时左右下楼，与 现场人群互动并签名，随后黄仁勋乘车离开。此前在2025年1月，黄仁勋曾在深圳公司的年会上派发"万 元大红包"，包括十多个1万元和2.5万元的红包，以及五个4万元的红包，当时黄仁勋在现场表示，下次 到访时红包金额将达到10万元。 据了解，本次"中国行"黄仁勋已经陆续到访上海、北京、深圳三大城市，他分别参加了三地分公司举办 ...