AI图片 瑞银警告称，一场关于车企成本的"完美风暴"正在形成。 1月27日，据追风交易台消息，瑞银在最新研报中称，随着2026年初刺激政策的全面退潮和5%购置税的征收，需求端本就疲软。然而， 供给端却迎来了大宗 商品（铜、铝、锂）和关键零部件（存储芯片DRAM）价格的剧烈反弹。 据该行测算，一辆典型的中型智能电动车的成本通胀高达人民币4000至7000元。瑞银称，在当前竞争激烈、利润微薄的市场环境下，由于车企很难将这部分 成本转嫁给消费者，这波成本上涨足以将车企的利润完全吞噬（fully erode carmakers' margin）。 瑞银表示，综合计算，仅金属原材料一项，纯电动车（BEV）的单车成本通胀就高达 人民币5600元（其中锂贡献了大部分涨幅，高达109%的价格反弹）。 被忽视的通胀源：DRAM价格飙升180% 具体测算如下： | 铝（Aluminum）： | 单车用量约200公斤。过去三个月铝价上涨导致单车成本增加约人民币600元。 | | --- | --- | | 铜（Copper）： | 单车用量约80公斤。铜价上涨导致单车成本增加约人民币1200元。 | | 锂（Lithium）： | ...

Core Viewpoint - DeepSeek has launched the DeepSeek-OCR 2 system, which utilizes the DeepEncoder V2 method to enable AI to understand images in a human-like logical sequence, potentially transforming document processing and complex visual understanding applications [1][12]. Group 1: Technical Innovations - The DeepEncoder V2 method allows AI to dynamically rearrange image segments based on their meaning, rather than following a rigid left-to-right scanning approach, mimicking human visual perception [1][5]. - DeepSeek-OCR 2 achieved a score of 91.09% in the OmniDocBench v1.5 benchmark, representing a 3.73% improvement over its predecessor [1][10]. - The model maintains high accuracy while controlling computational costs, with visual token counts limited to between 256 and 1120, aligning with Google’s Gemini-3 Pro [2][8]. Group 2: Performance Metrics - In practical applications, the model demonstrated a reduction in repetition rates, decreasing from 6.25% to 4.17% for online user logs and from 3.69% to 2.88% for PDF data processing, indicating its high practical maturity [2][10]. - The reading order edit distance metric improved significantly from 0.085 to 0.057, validating the effectiveness of the logical reordering capabilities of DeepEncoder V2 [10]. Group 3: Architectural Changes - The architecture of DeepEncoder V2 replaced the original CLIP components with a compact LLM-style architecture (Qwen2-0.5B), introducing learnable query vectors known as "causal flow tokens" [6][8]. - The design retains a bidirectional attention mechanism for visual tokens while employing a causal attention mechanism for causal flow tokens, allowing for intelligent reordering of visual information [7][8]. Group 4: Future Implications - The release of DeepSeek-OCR 2 signifies not only an upgrade in OCR performance but also a significant exploration of architecture, suggesting a promising path towards unified multimodal encoders capable of feature extraction across images, audio, and text [12].

AI图片 英伟达正通过重磅布局服务器CPU领域，进一步巩固其在AI基础设施市场的统治地位。 公司CEO黄仁勋确认将推出代号为"Vera"的全新CPU，并首次将其作为独立产品推向市场，旨在消除AI供应链中的算力瓶颈，全面支持代理人工智能（Agentic AI）的发展。 1月26日，据彭博报道，英伟达已承诺向新兴云服务商CoreWeave追加20亿美元投资， CoreWeave将成为首家部署Vera CPU作为独立基础设施选项的客户 。 黄仁勋在接受彭博Ed Ludlow采访时表示， Vera是一款"革命性"的CPU。 分析称，此次合作不仅延续了双方的长期关系，更标志着英伟达在计算堆栈层面的重大战略转折。 报道称，英伟达此举意在解决服务器CPU已成为AI供应链主要瓶颈的问题。 通过提供独立的Vera CPU，英伟达为客户提供了更具成本效益的高端计算替代方 案，使其不仅能在英伟达GPU上运行计算堆栈，也能在英伟达CPU上处理工作负载，这将对现有的服务器处理器市场格局产生直接影响。 据报道，随着Vera CPU技术细节的曝光以及生产计划的推进，英伟达正展示其在"后Blackwell时代"的雄心。 这一战略不仅针对高端 ...

在北京朝阳区一家繁忙的连锁咖啡店里，早高峰的节奏正如精密齿轮般运转。 一位店员熟练地接过订单，与此同时，吧台角落那颗不起眼的摄像头正捕捉着客流数据；后台的库存系统在实时监测咖啡、牛奶等物料的消耗量。 支撑这一系列井然有序场景的正是天数智芯的 国产 边端 AI算力 产品 。 事实上，这不仅是一家咖啡店的日常，更是 国产 AI算力设备在现实商业中扎根生长的典型横切面。 将视线从这间咖啡店拉升，我们看到的是一片更为壮阔的商业前景。弗若斯特沙利文预计到 2029年中国通用GPU市场规模有望攀升至7153亿元，未来5 年复合增长率将高达29.5%。 正是在国内市场规模爆发的前夕， 天数智芯、壁仞、摩尔线程等国产 GPU 厂商相继完成上市，只为拿到那张通往七千亿市场的入场券，以应对接下来更 为残酷的规模化战役。 当 "上市蓄力"完成，资本市场的聚光灯也让行业的隐痛无处遁形：落地困难、生态割裂依然是摆在现实的难题。 对此，天数智芯给出的答案是 一份横跨三年的 四代架构 路线 图和一系列边端新品 ： 1月26日， 天数智芯 一口气 亮出了 " 天数 天枢、 天数 天璇、 天数 天玑、 天数 天权 " 四 代架构 ，明确了在 ...

图片由AI生成 白银在纽约中午刷新历史高位至接近118美元的水平，随后快速回吐涨幅。 媒体分析称，这清楚地反映出，在连续多日由动量推动、追涨情绪主导的行情之后，这轮上涨已经变得极为脆弱。 此次大幅回落提高了一个风险，即这轮行情 中的投机阶段可能已经暂时告一段落。 尽管如此，从更大的背景来看，美元走弱以及地缘政治紧张情绪，仍然对贵金属整体构成支撑。 波动率失控与恐慌性买入 数据显示，白银市场目前比散户抱团股更加疯狂，波动率目前已处于"完全失控"状态。 据LSEG Workspace数据，渴望入场的投资者以及空头正在恐慌性买入上行波动率期权。白银波动率指标VXSLSLV目前徘徊在106附近，盘中一度触及124， 意味着市场正在为每日约7%的价格波动定价。这种波动率水平对于Meme股票而言已属极端，对于白银这类硬资产更是史无前例。 分析认为，白银这轮抛物线式上涨已经明显显露出疲态， 价格走势越来越像一次"冲顶式见顶"。 美东时间周一，白银价格亚太盘初以来持续走高， 北京时间02:03刷新历史高位至117.7132美元，涨幅一度多达14%，创下自2008年以来的最大单日涨幅。 但随后的回落来得非常突然：在周一美股收盘 ...

Core Insights - Morgan Stanley's 2026 market outlook highlights four major themes: AI technology diffusion, future energy, a multipolar world, and social change, presenting ten key predictions for investors to understand the evolving market landscape driven by technology [1] Group 1: AI Technology and Market Dynamics - The development of global AI technology is expected to show a bifurcated pattern, with the U.S. leading in advanced models while other regions lag behind, creating a "two worlds" scenario in AI development [2] - There will be an exponential growth in computing power demand driven by the proliferation of AI applications, which will outpace supply growth, fundamentally changing the economics of data centers [3] Group 2: Policy and Geopolitical Implications - The Trump administration is predicted to implement stronger policies than expected, focusing on securing domestic supplies of critical minerals, supporting manufacturing return, increasing military spending, and lowering consumer costs [4] - There will be increasing pressure for AI technology transfer and a push for national self-sufficiency, affecting trade dynamics and leading countries to enhance their domestic AI capabilities [5] Group 3: Energy and Infrastructure - Rising global energy costs will trigger a backlash against data center growth, prompting the introduction of low-cost energy support policies and encouraging data centers to adopt off-grid power strategies [6] - Major AI companies will accelerate the integration of energy infrastructure to secure reliable and cost-effective energy sources, enhancing energy efficiency through AI [7] Group 4: Manufacturing and Investment Trends - China is expected to increase its share in the global manufacturing market in key tech-intensive sectors, while the U.S. manufacturing landscape will shift towards domestic production as technology diffusion diminishes the low-cost labor advantage [8] - Latin America is entering a new investment cycle driven by policy shifts, geopolitical changes, and peak interest rates, with this bull market being investment-led rather than consumption-driven [9] Group 5: Workforce and Economic Transformation - Companies and governments will initiate extensive retraining programs to address employment changes driven by AI, responding to the political sensitivity surrounding job displacement caused by AI applications [10] - Transformative AI is anticipated to reshape the economy and asset valuations, with early signs of price declines across multiple sectors by the second half of 2026, leading to increased wage inequality and upward pressure on capital expenditures and interest rates [11][12]

图片由百度AI生成 腾讯正在通过其最擅长的社交打法，试图在日益拥挤的AI赛道中复制当年的移动支付奇迹。 1月26日，腾讯董事会主席马化腾在公司年会上定调2025年为"AI大年"，并宣布旗下AI应用"元宝"将于2月1日启动春节分10亿元现金活动 。马化腾明确表 示，希望通过此次高额补贴， 重现当年微信红包"摇一摇"绑定数亿用户的盛况。 这标志着腾讯已将AI产品的用户获取与活跃度提升至集团战略级的优先地 位。 与此同时，腾讯元宝低调开启了名为"元宝派"的社交功能内测，马化腾透露该产品此前属于"绝密项目"。根据媒体报道，"元宝派"允许用户创建群组，AI在其中 不仅担任助手，还深度参与社交互动。此举意味着腾讯正试图打破当前AI应用仅作为提效工具的局限，将其拉入高频的社交场景，通过"人机共生"的交互模式 构建新的流量护城河。 分析人士指出，腾讯此次"左手重金补贴，右手社交裂变"的组合拳，显示出其在AI商业化落地上的紧迫感。在经历了去年的行业混战后，腾讯正试图利用其在 社交链上的垄断优势，加速完成对用户AI使用习惯的"闪电式改造"，以应对字节跳动、百度等竞争对手在AI应用侧的激烈挑战。 10亿现金重注，意在复刻支付战役奇 ...

Core Viewpoint - Morgan Stanley anticipates that the upcoming January FOMC meeting will maintain interest rates unchanged, focusing on the tone of the statement [1][2] Group 1: FOMC Meeting Insights - The Federal Reserve is expected to keep the federal funds rate target range at 3.50%-3.75%, indicating a tactical adjustment rather than a return to a tightening cycle [2] - The statement is likely to upgrade the economic growth assessment from "moderate" to "robust" and remove references to "increased risks to employment," suggesting reduced concerns about the labor market [2][4] - Morgan Stanley predicts a dissenting vote from a board member advocating for a 50 basis point rate cut [2] Group 2: Market Strategy and Liquidity - Despite the Fed's pause on rate cuts, the short-term financing market remains loose, with repo rates normalizing below the interest on reserve balances (IORB), indicating an "excessively ample" cash situation [5] - The Fed is expected to maintain reserve levels by purchasing $40 billion in Treasury bills monthly, with projections for the SOMA account holdings to exceed $600 billion by the end of 2026 [6] Group 3: Currency Outlook - Morgan Stanley has revised its outlook on the foreign exchange market, now projecting a stronger U.S. economy with a GDP growth forecast of 2.4% for 2026, delaying the anticipated rate cuts [8] - Despite this, the firm maintains a moderately bearish view on the dollar due to synchronized global growth and undervaluation of the Japanese yen, which is expected to converge [9] Group 4: Asset Class Focus - In the mortgage-backed securities (MBS) sector, the significant $200 billion purchase plan by government-sponsored enterprises has led to a narrowing of MBS spreads, prompting a neutral stance from Morgan Stanley [11] - The municipal bond market shows solid fundamentals but is considered expensive, with low yield ratios compared to corporate bonds, raising concerns about sustainability if the Fed signals ambiguity rather than a clear dovish stance [11]

Core Viewpoint - Nvidia plans to enter the Windows consumer laptop market by launching Arm architecture-based System on Chips (SoCs), aiming to disrupt the long-standing x86 architecture monopoly and reshape the PC processor competitive landscape [1][2]. Group 1: Nvidia's SoC Launch - Nvidia is set to introduce two SoC models, N1 and N1X, which will integrate CPU and GPU into a single design, moving away from the traditional "x86 CPU + discrete GPU" configuration [1]. - This strategic shift is seen as an attempt to replicate Apple's success with custom Arm chips in the Mac ecosystem, potentially challenging Intel and AMD's dominance in the Windows laptop market [1][5]. Group 2: Early Adoption by Lenovo - Lenovo is leading the early adoption of Nvidia's Arm chips, having developed six laptop models based on the N1 and N1X processors, including various configurations such as the IdeaPad Slim 5 and Yoga Pro 7 [3]. - Notably, Lenovo's "Legion 7 15N1X11" gaming laptop indicates Nvidia's ambition to penetrate the high-performance consumer market, not just the lightweight segment [3]. Group 3: Dell's Participation - Dell is also preparing to launch laptops featuring the Nvidia N1X chip, with the Alienware gaming laptop and XPS high-end models expected to debut as early as 2026 [4]. - The collaboration between Nvidia and these manufacturers suggests that at least eight Nvidia-powered laptops are currently in development [4]. Group 4: Performance Insights - Although complete specifications for the N1 and N1X have not been officially released, leaked Geekbench data suggests that the N1X may feature up to 20 CPU cores and GPU capabilities comparable to the desktop-level RTX 5070 [4]. - The architecture is believed to be similar to Nvidia's GB10 Superchip used in the DGX Spark mini AI computer, indicating a strong performance potential [4]. Group 5: Industry Implications - Nvidia's entry into the PC chip market is anticipated to diversify CPU options for Windows laptops, marking the potential end of the era dominated by Intel and AMD x86 processors [2][5]. - The move aligns with trends where Apple leads with custom Arm chips, Qualcomm advances in Windows on Arm, and AMD explores Arm processor options, indicating a shift towards a multi-architecture future in the PC industry [2][5].

日本正深陷一场严峻的金融两难：在日元崩盘与国债市场解体之间，政策制定者似乎已无路可退。 随着日本债券收益率飙升且货币持续承压， 市场正密切注视着一个可能改变全球汇市格局的信号——美国是否正在准备"亲自下场"协助日本。 华尔街见闻此前提及，日本首相高市早苗周日发出严厉警告，承诺政府将采取"一切必要措施"应对市场的投机性和极度异常波动。 这一表态紧随上周五市场的剧烈震荡，当日美元兑日元汇率一度重挫约1.75%，日元创下五个月来最大单日涨幅。 市场普遍猜测，这一逆转的催化剂源自纽约联储极其罕见的"询价"（rate check）动作。 值得注意的是，此次行动极为罕见。 据纽约联储网站数据，自1996年以来，美国仅在三个不同场合干预过汇市，最近一次是在2011年日本地震后，与G7国家 联手卖出日元以稳定市场。 据早安汇市分析指出，由于时差问题，在东京的深夜时间，日本财务省干预汇率的工作是可以请求纽约联储"接棒"代理的，这时纽约联储干预动用的也是日本 财务省的外汇储备。 而本次纽约联储rate check代表美国财政部的意志，需要美国财长贝森特（甚至需要特朗普）签署确认，因而已经上升到跨国联合干预行动的层级。 据媒体援引 ...