小熊跑的快
Search documents
美国债确实是最近最大的影响因素
小熊跑的快· 2025-05-22 03:08
Core Viewpoint - The primary focus for the U.S. is addressing the issues surrounding government debt, which is significantly influencing various tariffs, policies, and market sentiment [1] Group 1: U.S. Treasury Auctions - A large amount of U.S. government debt is maturing at the end of June, leading to a critical 90-day period for stabilization [1] - The recent auction of 20-year Treasury bonds was notably poor, resulting in a new high for bond yields [1] - The U.S. Treasury auctioned $16 billion of 20-year bonds, with the final yield at 5.047%, marking the second instance of yields surpassing 5% since the bond's introduction five years ago [1] - The yield from this auction was 24 basis points higher than the 4.810% from April, and the auction's bid-to-cover ratio was 2.46, the lowest since February [1] Group 2: Market Reactions - The pressure on U.S. assets is significant, with indications that the dollar index is likely to continue declining [1] - U.S. stock markets have begun to decline as well, reflecting broader market concerns [1] - Traditionally, rising bond yields would lead to falling gold prices; however, gold is currently rising as an alternative to the dollar [1] - There has been a divergence in the behavior of Treasury bonds and gold, with both previously moving in tandem during times of risk aversion [1] Group 3: Investment Sentiment - The range of assets considered safe havens is narrowing, indicating a shift in investment strategies [2]
裁员遍及海内外
小熊跑的快· 2025-05-21 07:35
所以我们看到很多海内外大厂eps 纷纷超预期?营业利润率也在微涨。多少收入贡献的?多少裁员贡献 的?比如google肯定要依靠裁员继续提升eps的,它目前还有18w人。meta也有这个打算。微软目前人均 贡献收入比google 低,目测也要大裁特裁。只有英伟达人均收入最高,目前没有裁员的压力和计划. A股上市公司呢? 其实也挺猛。最后一列是24年财报 减少人数比上22年公司披露的总人数。表里只看到tmt。主要想知道 ai带来的影响,对码农和后台的影响。 但是绝对数第一还真不是tmt,比如晶澳,晶科....那都是万人起步。表装不完,后面有一些类似中国移 动这样的公司 人员是增加的,占比较小。 数据:东方财富 最近都在讨论 因为AI导致的裁员。 全球都被美国几个大厂的 裁员计划震惊了。动则几万几万的裁员。 因为总人数是10几 20万人。所以哪怕3%,也有7000-8000人被裁。 今年meta google 微软还会是裁员重灾区。 ...
大摩看多中国资产
小熊跑的快· 2025-05-20 08:22
Core Viewpoint - The article emphasizes that China's AI development model differs from the U.S., focusing on "efficiency-driven, engineering implementation" rather than "compute power supremacy" [1] Group 1: AI Development and Investment Landscape - Major foreign banks are optimistic about Chinese cloud providers and chip foundries, indicating a positive outlook for international investments in these sectors [1] - The AI investment value chain is shifting from foundational hardware to application layers, platform layers, and vertical industry implementations [1] - The DeepSeek incident is identified as a pivotal moment that has prompted a reassessment of the "cost efficiency model" and spurred an open-source movement [1][2] Group 2: Economic Implications of AI - The AI revolution is expected to enhance China's long-term potential GDP growth by addressing structural challenges such as aging population and slowing productivity growth [3] - AI capital expenditure is projected to be a primary driver in the short term, contributing approximately 0.2 to 0.3 percentage points to annual GDP growth [3] - By 2024, AI is estimated to create an equivalent labor value of approximately 6.7 trillion RMB, assuming AI penetration aligns with IMF predictions for emerging markets [3][4] Group 3: Return on Investment and Market Dynamics - By 2030, China's AI industry is projected to yield a total return of about 806 billion RMB, achieving a 52% return on invested capital (ROIC) [4] - The e-commerce sector is expected to contribute the largest share to AI growth, estimated at 271 billion RMB, followed by advertising and local services [4] - The data center market in Beijing and surrounding areas is anticipated to experience the highest investment returns, with a supply-demand reversal expected within 6 to 12 months [4] Group 4: AI Hardware and Ecosystem - Despite U.S. chip restrictions, China's AI computing power continues to advance, supported by a combination of existing GPU inventory and domestic GPU development [5] - The self-sufficiency rate of Chinese AI GPUs is projected to reach 34% by 2024 and 82% by 2027 [5] - Huawei and Cambricon are identified as key domestic GPU suppliers, with most chips manufactured by local foundry SMIC [6] Group 5: Global Market Outlook - The global cloud AI market is expected to grow at a CAGR of 28% from 2024 to 2027, reaching approximately $239 billion, with China projected to account for 20% of this market [6]
COMPUTEX 2025
小熊跑的快· 2025-05-19 13:03
Core Insights - The article discusses the advancements in AI technology as presented by NVIDIA CEO Jensen Huang at COMPUTEX 2025, highlighting the evolution of AI from perception to reasoning and physical AI [1] AI Evolution Path - The evolution of AI is categorized into four stages: 1. Perception AI: Understanding patterns like speech and image recognition 2. Generative AI: Transitioning from understanding to generating content across multiple modalities 3. Reasoning AI: Focusing on complex reasoning capabilities, utilizing techniques like "Chain of Thought" and "Tree of Thought" 4. Physical AI: Understanding physical concepts such as inertia and causality, crucial for the next AI era [1] GB300 and Blackwell Architecture - The GB300 system, based on the new Grace Blackwell architecture, has been in production since early this year, with significant upgrades including a 1.5x increase in inference performance and a 2x increase in network capability [2] - The system features 100% liquid cooling and maintains the same physical footprint as previous models, with a single node performance of approximately 40 petaflops [2] NVLink and CoWoS-L Technology - NVIDIA has developed a new collaborative process with TSMC called CoWoS-L to create larger chips, enhancing performance through NV-Link technology, which offers a data transfer speed of 7.2TB/s [3] - The NV-Link architecture connects multiple GPUs within a single rack, achieving a bandwidth of 130 terabytes/s, necessitating liquid cooling due to high power requirements [3] NVLink Fusion - NVLink Fusion is introduced to allow partners to build semi-custom AI infrastructure solutions, enabling integration of custom ASICs into NVIDIA's ecosystem [4] - This technology facilitates the mixing of NVIDIA components with partner-specific chips, enhancing the flexibility of AI infrastructure [5] DGX Spark and Workstations - DGX Spark has entered full production, designed for AI-native developers, offering 1 petaflops of computing power and 128GB of memory for prototyping and early development [6] - NVIDIA also launched desktop-level DGX supercomputers, capable of running AI models with up to 1 trillion parameters, suitable for home use [6] Enterprise AI Solutions - The RTX Pro Enterprise server integrates x86 architecture and supports various AI agents, showing significant performance improvements over previous models [7]
阿里答卷
小熊跑的快· 2025-05-15 23:08
阿里云季度收入加速增长至18%, 达到301.27亿元,创下三年来的最快增速,其中AI相关产品收入连续第七个季度保持三位数的同比增 长。 • 经营利润为人民币1409.05亿元(194.17亿美元),同比增长24%。 • 归属于普通股股东的净利润为人民币1294.70亿元(178.41亿美元)。净利润为人民币1259.76亿元 (173.60亿美元),同比增长77%。 对外商业化收入同比增长17%。上个季度13%。 I DC 最新数据, 阿里云国内市场排名第一,份额连续三个季度回升(阿里和火山都不错)。 IDC分 析,阿里云加大研发投入,在通义系列大模型、AI基础设施等领域全面发力,市场份额和同比增速显 著提升。 截至4月底, 阿里通义已开源200余个模型 ,全球下载量超3 亿次。 此外 基建带来的折旧出现了,海内外一样。 经调整 EBITA 利润率环比下降 1.9 个百分点(上个季度10%), 这是由于我们加大了技术和产品开发 方面的投入以把握快速增长的 AI 需求,同时,随着我们加大基础设施投入以满足需求爬坡,折旧与摊 销费用的上升也对利润率产生了一定影响。 本季度资本开支246亿,环比下降28%。 回溯 ...
腾讯ai说了啥
小熊跑的快· 2025-05-14 23:44
昨天腾讯收入微超市场预期.capex 比彭博预测高,但环比下降了一些。 市场最关心对AI的表述。 对ai的定位: A:通用型智能体AI可以通过调用工具和多步骤操作完成复杂任务,腾讯正在"元宝"、"艾玛"等原生AI产 品中培育这种能力。随时间推移,AI将从简单问答逐步进化到具备长链条思考、推理和跨应用操作能 力。在微信生态中,腾讯还可以构建另一种代理AI,通过整合社交关系链、公众号、视频号、小程序 等组件,形成与其他通用型智能体AI差异化的产品。比如用户可以通过AI直接调用微信生态内数百万 小程序的商业服务能力,这在其他平台难以复制,未来可能成为腾讯的差异化产品。 关于商业模式,广告业务已通过AI提升精准投放获得增长,AI能够增强广告精准投放能力,直接转化 为更多的广告收入。交易场景与广告的联动效应也日益显著。广告直接促成交易,进而体现广告价值。 而GPU租赁目前优先级较低,属于新兴领域,更像是转售业务,当GPU供应紧张时,租赁业务并非腾讯 的首选。 订阅制并非中国AI业务的主要模式,目前所有人都在免费提供AI服务,因此国外的订阅模式 并不适用于国内电商领域的主流AI商业模式。 对GPU的采购: Q:管理层如何 ...
cutting同时 推理价格涨了
小熊跑的快· 2025-05-13 23:30
Core Viewpoint - The year 2025 is projected to be a significant year for layoffs due to AI, with companies implementing single-digit percentage layoffs while simultaneously experiencing an increase in rental prices for inference chips [1] Group 1 - Companies are expected to implement layoffs in the single-digit percentage range in 2025 due to the impact of AI [1] - The rental prices for inference chips, specifically A10 and T4, have been rising recently, indicating a shift in demand despite layoffs [1] - A10 and T4 chips are limited to inference tasks and cannot be used for training, highlighting a specific market niche for these products [1]
cutting workforce 裁员
小熊跑的快· 2025-05-13 23:12
学长前辈调侃 还好,都是低个位数裁员。 虽然但是,目前就业确实挺难。 前几年进个微软 挺轻松的,现在学校不错,offer难度挺大。 最近的新闻是 微软宣布裁员3%。年初meta宣布计划裁员5%。 Google一直在裁,24年一口气裁1.2w,裁了6%。 23-24年是为了降本增效裁员,25年里面的前辈说,开始因为AI裁员了。Ai能替代较多岗位。其实赔偿 也不算多。好在创投还比较活跃,出来后很多去了创新公司。 上午刚跟 业内大厂的人咨询 就业的问题。 谈到了2023年以来的西海岸裁员的事儿。 ...
关于agent的2个事
小熊跑的快· 2025-05-13 10:17
Group 1 - OpenAI announced a significant update potentially related to the GPT-4.5 model, personalized assistant systems, or multimodal capabilities [1] - OpenAI launched HealthBench, a medical open-source testing benchmark developed by 262 global doctors, featuring 5,000 real health dialogues and 48,562 evaluation criteria, aimed at standardizing medical AI assessments [1] - The introduction of HealthBench provides essential evaluation standards and data support for the development of medical agents, enabling quantifiable assessments of their accuracy, communication quality, and contextual awareness in medical scenarios [1] Group 2 - Manus, a domestic AI intelligent entity, has opened registration for all users, offering daily free task execution and a reward of 300 points, with new users receiving an additional 1,000 points [1] - A previously criticized company has made a comeback in the overseas market, with subscription fees set at $19, $39, and $199, corresponding to basic functionality, priority computing power scheduling, and enterprise-level API access [2] - The year 2025 is widely regarded as the "year of agents" in the industry, with various companies competing in both vertical and general applications, indicating a shift from computational power to application development [2]