*内容仅为呈现不同市场观点与研究视角，并不意味着本公众号对文中观点结论认可。 在过去二十年的中国投资史上，段永平的名字几乎等同于 "稳定性"的代名词。从早年重仓网易的成名一战，到后来长期持有苹果、贵州茅台， 他的投资行为始终贯穿着一条清晰的主线： 长期主义、重仓出击、极低换手。 他倾向于围绕少数几家确定性极高的消费与平台型巨头反复加 码，这种"买入并持有"的哲学，让他被市场公认为"巴菲特价值投资体系在中国最忠实的实践者"。在很长一段时间里，跟随段永平的持仓逻 辑，几乎成为了许多价值投资者的一种隐性策略。 然而，当时钟拨向 2025 年第四季度，这位一贯以"静水流深"著称的投资人，却交出了一份明显偏离历史路径的成绩单。根据 2026 年初披露 的持仓数据显示，他大幅减持了长期占据第一大重仓股位置的苹果，与此同时，将英伟达（Nvidia）的仓位一口气拉高逾 11 倍，并在 AI 产业 链的多个垂直领域进行了轻仓试水。 这并不是一次普通的战术调仓，更像是一次对时代变量的正面回应。如果把时间轴再拉长一点，这个变化恰好发生在一个重要的宏观背景之后： 沃伦 ·巴菲特正式淡出伯克希尔的日常投资决策。当"老师"的时代落幕，那 ...

英伟达与Meta达成多年期协议，将向后者出售数百万芯片，盘前股价上涨2%。 来源：滚动播报 ...

本文源自：金融界AI电报 美股盘前，部分明星科技股走强，英伟达(NVDA.O)、美光科技(MU.O)、Meta Platforms(META.O)涨超 1%。 ...

每经AI快讯，2月18日，美股部分大型科技股盘前走强，英伟达涨1.68%，美光科技涨1.43%， MetaPlatforms涨1%。 ...

Core Viewpoint - The article discusses the evolving landscape of investment in hard technology, highlighting the shift in investment logic and the importance of understanding market dynamics and differentiation in projects [12][20][29]. Group 1: Investment Trends - Current investment cycles in hard technology resemble the TMT era, characterized by massive financing and significant losses [8][9]. - The investment logic is shifting, with a focus on projects that either require continuous financing or are profitable from the start [6][12]. - The capital market has recently regained vitality, with renewed interest in AI and technology sectors, although challenges remain in securing funding for early-stage projects [13][15]. Group 2: AI and Globalization - The article emphasizes the importance of the AI value chain, noting that while major players like Nvidia thrive, many AI application companies face difficulties in domestic markets and are compelled to seek international opportunities [15][16]. - Chinese companies often need to adopt a "shell" strategy to appeal to international investors, complicating their growth prospects [17][18]. - The article points out that supply chain companies working with global brands tend to perform better than those serving domestic brands [18]. Group 3: Investment Decision-Making - Successful investment requires identifying unique project characteristics and understanding market sentiment rather than solely focusing on technology [25][28]. - The article warns against over-analysis, suggesting that investors should act decisively and focus on projects that can survive in a competitive landscape [23][29]. - It highlights the importance of being aware of the competitive landscape and the need for projects to differentiate themselves to attract investment [28][30]. Group 4: Mergers and Acquisitions - The article discusses three notable mergers, illustrating the differences between American and Chinese acquisition strategies, with the latter often relying on talent acquisition rather than technology purchases [32][34]. - It notes that many Chinese companies lack both "face" and "substance," complicating the merger process [33]. - The article concludes that the Hong Kong capital market serves as a crucial buffer between Chinese and American markets, especially for tech companies seeking IPOs [36][37].

Nvidia doesn't have to miss Wall Street's Q4 estimates for its share price to fall.Many Nvidia (NVDA +1.12%) investors have probably circled Feb. 25, 2026, on their calendars. That's the date the giant chipmaker is scheduled to announce its fiscal 2025 fourth-quarter results. Nvidia plans to release its results following the market close and hold its quarterly conference call with analysts shortly afterward. It's easy to predict that Nvidia's share price will surge after its Q4 update. The company's results ...

此举为两家公司的动荡篇章画上了句号。英伟达曾于2020年同意以400亿美元收购Arm，这笔交易原本 有望成为芯片行业有史以来最大规模并购。但该交易几乎立即面临监管机构和客户的反对。Arm的技术 支撑着全球大多数先进半导体，其独立性被视为至关重要的资产。 2022年2月，双方宣布终止协议。由软银集团控股的Arm随后继续推进其首次公开募股计划。 作为全球市值最高的公司，英伟达已成为科技行业的重要投资者。英伟达新增持仓覆盖芯片制造、 EDA工具、通信基础设施三大关键领域，分别为英特尔(INTC.US)、新思科技(SNPS.US)与诺基亚 (NOK.US)。 智通财经APP获悉，英伟达(NVDA.US)已清空其持有的全部Arm(ARM.US)股份。五年前，英伟达曾试 图收购这家芯片技术公司，但最终以失败告终。 根据一份监管文件，英伟达出售了110万股Arm股票，以Arm周二收盘价计算价值约1.4亿美元。这笔交 易发生在去年第四季度，至此英伟达在Arm的持股已归零。 英特尔作为全球传统芯片巨头，英伟达此前已斥资约50亿美元战略入股，持股近4.91%，此次纳入公开 持仓进一步强化AI算力底层协同；新思科技是全球EDA行业 ...

智通财经APP获悉，投资大佬段永平美股持仓主体H&H International Investment披露了截至2025年12月 31日的第四季度持仓报告(13F)。据统计，H&H第四季度美股持仓总市值达175亿美元，上一季度总市值 为147亿美元。H&H的持仓组合中新增3只个股，增持了6只个股，减持了5只个股，清仓了0只个股。其 中前十大持仓标的占总市值的99.66%。 (资料图片) ...

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。 ...

Core Insights - Meta has entered a large-scale agreement to utilize millions of NVIDIA chips, including the new standalone CPU and the next-generation Vera Rubin system, in its AI data centers [1][2] - The financial terms of the agreement have not been disclosed, but it is expected to be in the hundreds of billions range, with a significant portion of Meta's capital expenditure directed towards this collaboration [1][2] - This partnership marks a significant expansion of the existing collaboration between Meta and NVIDIA, which has been ongoing for at least a decade [1] Group 1 - The standalone CPU is a major highlight of this collaboration, making Meta the first company to deploy NVIDIA's Grace CPU as a standalone chip in data centers [1][2] - The next-generation Vera CPU is planned for deployment by Meta in 2027, as part of a broader commitment to invest $600 billion in data centers and supporting infrastructure in the U.S. by 2028 [2] - Meta plans to build 30 data centers, with 26 located in the United States [2] Group 2 - Two major AI data centers are currently under construction: the 1 GW Prometheus project in New Albany, Ohio, and the 5 GW Hyperion project in Richland Parish, Louisiana [3] - The agreement includes NVIDIA's Spectrum-X Ethernet switches for high-speed GPU interconnectivity within large AI data centers [3] - Meta will also leverage NVIDIA's security capabilities to support AI functionalities in WhatsApp [3] Group 3 - Meta is not solely reliant on NVIDIA, as it is also considering using Google's Tensor Processing Units (TPUs) in its data centers by 2027, which previously caused NVIDIA's stock to drop [3] - The company is developing its own chips and utilizing AMD products, especially as AI companies seek alternative supply channels amid NVIDIA's supply constraints [3] - NVIDIA's current Blackwell GPUs have been in short supply for several months, while the next-generation Rubin GPUs have recently entered production [3] Group 4 - Through this agreement, Meta has secured a sufficient supply of chips and will engage in deep joint design with NVIDIA's engineering team to optimize and accelerate cutting-edge AI model training and inference [4] - Meta is developing a new cutting-edge model, codenamed Avocado, as a successor to its Llama series AI technology [5] - Recent fluctuations in Meta's stock price have raised concerns on Wall Street regarding its AI strategy, with significant price movements following the announcement of ambitious AI investment plans [5]