【中信科技产业】海外AI叙事或重回乐观情形，重视海外算力链新一轮上涨机遇！ 核心判断 近期海外推理与训练算力需求同步走强，亚马逊云、谷歌云相继涨价，台积电上修Capex。 尽管当前AI应用大规模商业化能见度仍有限，但在模型与应用密集催化下，未来3–6个月海外算力需求有望进一步 上行，算力"泡沫论"担忧或阶段性缓解，产业链有望迎来新一轮修复。 推理侧：Agent落地 【中信科技产业】海外AI叙事或重回乐观情形，重视海外算力链新一轮上涨机遇！ 核心判断 近期海外推理与训练算力需求同步走强，亚马逊云、谷歌云相继涨价，台积电上修Capex。 尽管当前AI应用大规模商业化能见度仍有限，但在模型与应用密集催化下，未来3–6个月海外算力需求有望进一步 上行，算力"泡沫论"担忧或阶段性缓解，产业链有望迎来新一轮修复。 推理侧：Agent落地抬升推理算力消耗 MoltBot等新一代Agent显著提升对电脑操作与复杂任务的处理能力，带来更高推理算力消耗。 Anthropic持续推出Claude Code、Agent Skills等产品，拓展Agent应用场景。 OpenRouter数据显示，2026年1月初以来Token调用量持 ...

公众号记得加星标⭐️，第一时间看推送不会错过。 前所未有的无监督训练数据的可用性，以及神经网络的扩展规律，导致用于服务/训练低层逻辑模型 （LLM）的模型规模和计算需求出现了前所未有的激增。然而，主要的性能瓶颈正日益转移到内存 带宽上。 过去20年，服务器硬件的峰值浮点运算能力（FLOPS）以每两年3倍的速度增长，超过了DRAM和互 连带宽的增长速度，后两者分别仅以每两年1.6倍和1.4倍的速度增长。这种差距使得内存而非计算成 为人工智能应用（尤其是服务应用）的主要瓶颈。 本文分析了编码器和解码器Transformer模型，并展示了内存带宽如何成为解码器模型的主要瓶颈。 我们提出重新设计模型架构、训练和部署策略，以克服这一内存限制。 引言 近年来，训练大型语言模型 (LLM) 所需的计算量以每两年 750 倍的速度增长。这种指数级增长趋势 是人工智能加速器发展的主要驱动力，这些加速器致力于提升硬件的峰值计算能力，但往往以牺牲其 他部分（例如内存层次结构）的简化为代价。 然而，这些趋势忽略了训练和服务人工智能模型过程中一个新兴的挑战：内存和通信瓶颈。事实上， 许多人工智能应用的瓶颈并非计算能力，而是芯片内部/芯 ...

芯片是现代电子技术的基石。目前，其信息处理能力依赖于在硅片上构建的高密度晶体管集成电 路。为了追求更强的算力，人类沿着摩尔定律不断推进制程工艺，推动了多个产业变革。随着可穿戴设 备、电子织物、脑机接口等新兴领域的蓬勃发展，人们希望能发展出不同于硬质硅基芯片的新型柔性信 息处理器件，以有效满足电子设备柔性化、轻量化、微型化的应用需求。 复旦大学彭慧胜/陈培宁团队打破传统芯片硅基研究范式，成功在柔软、弹性的高分子纤维内，制 造出大规模集成电路，创造出一种全新的信息处理器——纤维芯片。与传统芯片相比，这种新型芯片具 有高度柔软、适应拉伸扭曲等复杂形变、可编织等独特优势，有望为脑机接口、电子织物、虚拟现实等 新兴产业变革发展提供有力支撑。相关成果于1月22日发表于《自然》主刊。 芯片从"硬质"变"软线" 如何在细如发丝的纤维上实现强大的信息处理功能，而又不影响其柔软、可拉伸、可编织的本性？ 这是纤维电子领域公认的"硬骨头"。 在智能可穿戴设备方兴未艾的今天，一个主要矛盾长期存在：我们的身体和衣物是柔软的，而赋予 它们"智能"的核心部件——芯片却是硬质的。这导致最终的智能织物、植入式设备都难以摆脱"外挂"硬 质信息处理 ...

Core Viewpoint - Apple reported strong quarterly results but CEO Tim Cook indicated a need to reset expectations due to supply constraints and rising component costs [1][4][5]. Financial Performance - Apple achieved revenue of $143.8 billion, a 16% increase year over year, and diluted EPS of $2.84, up 19% year over year, with a net income of $42.1 billion [6]. - The gross margin was reported at 48.2%, exceeding guidance, and operating cash flow reached a record $53.9 billion [6]. - Segment sales included iPhone at $85.3 billion (+23%), Services at $30.0 billion (+14%), Mac at $8.4 billion (-7%), iPad at $8.6 billion (+6%), and Wearables/Home/Accessories at $11.5 billion (-2%) [6]. Supply Chain and Component Costs - Cook highlighted that Apple is in "supply chase mode" due to advanced chip constraints and rising memory prices, which are expected to persist for several years [2][8]. - The memory market is experiencing record demand, with companies like SanDisk seeing stock increases of 1,230% in the past six months [3]. - Apple is facing challenges in balancing supply and demand, with Cook noting that demand is currently outpacing Apple's planning [7][8]. Market Outlook - For the upcoming March quarter, Apple anticipates revenue growth of 13% to 16% year over year and a gross margin between 48% and 49% [6]. - Analysts remain optimistic about Apple's stock, with average price targets suggesting significant upside potential, ranging from $280 to $330 [19]. Pricing Strategy - Apple's pricing strategy for the iPhone has historically shown resilience, with demand remaining strong even at higher price points [14][15]. - Consumer sentiment indicates that while many perceive iPhones as overpriced, a notable percentage still consider them worth the investment despite financial constraints [17].

以下文章来源于氪睿研究院 从表面数据看，2026年全球IPO市场正在回升。多国交易所的上市储备项目增加，AI、硬科技、能源与先进制造成为高频关键词，市场开始重新讨论"IPO 窗口期"的打开。这一现象，容易被解读为资本市场风险偏好的修复。 但如果进一步拆解结构，会发现这一轮IPO并不符合以往"周期复苏"的典型特征。 首先，上市企业的类型发生了显著变化 能够顺利推进IPO的公司，集中于少数行业和赛道，且普遍具备高资本密度、长周期投入和强政策关联特征；而大量轻资产、应用导向或依赖叙事驱动的 企业，仍然停留在上市门外。 其次，IPO的定价逻辑正在发生位移 过去二十年，资本市场更倾向于为"增长潜力"定价；而在当前高利率与地缘政治结构化的背景下，资本开始优先评估企业的战略必要性、现金流可验证性 以及长期资本承载能力。 这意味着，IPO正在从一种"市场奖励机制"，转变为一种战略资产筛选与定价机制。 在美国市场，算力、AI基础设施、航天与国防相关企业获得显著溢价，本质上反映的是资本对"未来关键基础设施"的提前定价；在中国市场，硬科技、新 质生产力与产业链关键环节加速进入资本市场，IPO更多承担的是产业升级与技术自主的制度功能 ...

来源：环球网 针对这一传闻，黄仁勋回应称："我们将对OpenAI进行一笔巨额投资。我坚信OpenAI，他们所取得的成 就令人惊叹，无疑是当下最具影响力的公司之一。我非常享受与萨姆（OpenAI CEO萨姆·奥尔特曼）共 事。"他还补充道，目前萨姆正在推进这一轮融资，英伟达肯定会参与其中并投入大量资金，这极有可 能成为英伟达史上最大的一笔交易。（青山） | I INV VEHICHING I I DULUSURIA THE ECONOMIC TIMES tech | | --- | | Subscribe to ETPrime English Edition . Today's ePaper | | Home BUDGET DETPrime Markets [[Market Data Masterclass News Industry SME Politics Wealth Tech Al Careers Opinion NRI Pan | | Al Web Stories IT Tech & Internet Funding Startups Tech Bytes Newsletters . Blogs & ...

公众号记得加星标⭐️，第一时间看推送不会错过。 一个简化的AI加速器架构展示了这两个领域如何共存。在计算层，加速器通过高带宽铜缆链路向上连 接到L1计算交换机。这些是典型的纵向扩展连接：短距离、高密度，并针对以最小延迟传输海量数 据进行了优化。L1交换机之间也通过铜缆互连，形成一个紧密耦合的网络结构，使得多个加速器在 软件层面上几乎可以像一个大型设备一样运行。 随着流量向上层级传输，它会汇聚到与更广泛的数据中心网络连接的二层网络交换机。在这个层级， 光插拔设备占据主导地位，因为系统必须支持更远的传输距离、更高的端口数量以及可扩展的带宽增 长。 这两个领域面临的日益严峻的挑战是，尽管电信号串扰器（SerDes）仍在不断发展，但其系统层面的 限制却日益增多。在硅芯片上，SerDes 的容量持续从 112G 扩展到 224G PAM4 及更高。然而，随 着数据速率的提升，包括封装、基板、PCB 走线、连接器和电缆在内的电气通道逐渐成为瓶颈。为 了在远距离传输中保持信号完整性，需要越来越强大的均衡和数字信号处理（DSP）能力，这会导致 每比特功耗增加，并增加热负载。 对于拥有数千条SerDes通道的大型AI交换机和加 ...

Core Insights - The rising adoption of generative AI models from OpenAI and Anthropic is significantly impacting major cloud computing platforms, with trillions of dollars committed to future infrastructure projects by these companies [1] Group 1: OpenAI and Microsoft - OpenAI's partnership with Microsoft has provided the latter with a first-mover advantage in integrating generative AI, with ChatGPT being heavily utilized across Microsoft's Azure cloud services [3][5] - The increasing use of OpenAI's software has led to a surge in AI workloads on Azure, driving demand for incremental cloud services [4] Group 2: Amazon's Role in AI Infrastructure - Amazon Web Services (AWS) has entered a $38 billion GPU leasing deal to support OpenAI, highlighting the competitive landscape among cloud providers [7] - Amazon has invested $8 billion in Anthropic, positioning itself strategically in the AI sector, with Anthropic utilizing AWS's GPU clusters and custom-designed chips [8][9] - If Amazon's AI accelerators can compete effectively with Nvidia and AMD's GPUs, AWS could gain significant pricing power and increase customer retention [10][11] Group 3: Google Cloud's Position - Google Cloud has experienced impressive growth, with OpenAI and Anthropic as key customers, leveraging its computing power and custom chips [12][13] - Anthropic's use of Google Cloud's Tensor Processing Units (TPUs) is expected to enhance Google Cloud's competitive position in the AI infrastructure market [14] - As OpenAI addresses its capacity challenges, Google Cloud is likely to benefit from increased user adoption and ongoing data center expansion [15]

Advanced Micro Devices ( NASDAQ:AMD ): A multi-year deal to supply around 6 GW of GPU capacity, potentially valued up to $300 billion. OpenAI would also receive warrants for a 10% stake in AMD if certain targets are met.Microsoft ( NASDAQ:MSFT ) Azure: An incremental $250 billion commitment from OpenAI to purchase Azure cloud services for AI training and inference over several years.OpenAI’s position today is anchored in a dense network of partners and funding discussions that go far beyond any single Nvidi ...

Core Insights - Meta Platforms is positioned as a leading AI stock for 2026, driven by significant advancements in its advertising business and productivity enhancements through AI technologies [1] Group 1: Advertising Business Transformation - Meta's ad revenue increased by 24% year over year in Q4, reaching $58.1 billion, with AI playing a crucial role in enhancing revenue and profits [2] - The company revamped its ad ranking model and doubled GPU usage for AI training, resulting in a 3.5% rise in ad clicks on Facebook and over 1% increase in ad conversions on Instagram [3] Group 2: Productivity Improvements - The implementation of agentic coding has led to a 30% increase in output per engineer since early 2025, with power users experiencing an 80% year-over-year productivity boost [4] - CFO Susan Li indicated that growth is expected to accelerate in the latter half of 2026 [5] Group 3: Product Innovations - Sales of Meta's AI-powered smart glasses tripled in 2025, with CEO Mark Zuckerberg comparing their potential impact to that of smartphones [5][6] - Meta is committed to developing personal superintelligence, which is anticipated to significantly enhance user experience by understanding individual preferences and relationships [7] Group 4: AI Infrastructure Development - Meta established a new division, Meta Compute, aimed at creating custom silicon and energy sources for AI, which is expected to reduce reliance on third-party chips and lower energy costs [8][9] - The Andromeda ad retrieval engine is now compatible with various GPUs, including Nvidia and AMD, enhancing operational flexibility [9] Group 5: Business-to-Business Revenue Growth - Agentic AI is also contributing to B2B revenue, with Meta's business AIs on WhatsApp facilitating over 1 million weekly conversations between customers and businesses in Mexico and the Philippines [10][11] - Plans are in place to expand the availability of these AI agents to more markets in 2026 [11] Group 6: Reality Labs Financial Outlook - Reality Labs reported a $6 billion loss in Q3, which impacted overall profits, but losses are expected to stabilize in 2026, with a focus on AI glasses and wearables [12][13]