Core Viewpoint - Nvidia's market position has been challenged globally, but it has unexpectedly benefited from a surge in H20 chip sales in China, with major tech companies placing significant orders [1][3]. Group 1: Sales Surge - In Q1 2025, Chinese tech giants like ByteDance, Alibaba, and Tencent collectively ordered at least $16 billion worth of H20 chips, nearly matching Nvidia's total sales in China for 2024, which was $17 billion [1][4]. - In 2024, Nvidia shipped over 1 million H20 chips to China, generating approximately $12 billion in revenue, while in just three months of 2025, the orders approached the previous year's total [4]. Group 2: Market Dynamics - Initially, Chinese companies were cautious about H20 chips due to performance issues and high prices, leading to a wait-and-see attitude [5]. - By the second half of 2024, market confidence rebounded, evidenced by a surge in orders at the beginning of 2025 [6]. Group 3: Supply and Demand - As of early April 2025, H20 chip inventory was nearly depleted, with new shipments expected only by mid-April, indicating a supply-demand imbalance [7]. Group 4: Drivers of Demand - The rapid growth of China's AI industry and the unique positioning of H20 chips have driven demand, particularly for low-cost AI models [8]. - Major companies are investing in AI servers based on H20 chips to support cloud services and the deployment of DeepSeek-R1 models [8]. Group 5: Competitive Landscape - Despite the current success, H20 chips face significant challenges from U.S. export restrictions and domestic regulatory pressures [10][11]. - The potential tightening of U.S. export controls poses a risk to H20's supply chain, while new energy efficiency standards in China could limit sales if H20 chips do not comply [13]. - Local competitors like Huawei are rapidly advancing, with their chips offering similar pricing and localized advantages, which could threaten H20's market position [14].

Core Insights - The rise of AI-native organizations represents a dual revolution in technology and organizational structure, posing significant challenges to traditional internet giants [1][2] - The competition is not only about technological capabilities but also about organizational forms, cultural genes, and talent strategies [2][3] Group 1: Characteristics of AI-native Organizations - AI-native organizations integrate AI as a core driver of products, services, and business processes, rather than as an added feature [2] - They possess self-developed core technologies, with rapid iteration speeds that outpace traditional companies, exemplified by OpenAI's swift transition from GPT-3 to GPT-4 within two years [2] - Product design inherently relies on AI capabilities, making it impossible for products to exist independently of AI [3] - The focus has shifted from "data and computing power" to "algorithms and community," emphasizing algorithm breakthroughs and scenario innovations as keys to market recognition [4] - Organizational structures are fluid, with flat, self-organizing teams that enable rapid decision-making and resource responsiveness [5] - A geek culture and strong founder cohesion drive these organizations, emphasizing technical idealism and long-term value [6] Group 2: Challenges for Traditional Internet Giants - Traditional tech giants face a core issue: how to evolve their organizations to maintain competitiveness in the AI-native wave [2][9] - Despite having significantly more resources, traditional companies struggle to replicate the technical sharpness of AI-native organizations like DeepSeek [1][9] - The lack of visionary leadership and a clear pursuit of algorithmic efficiency hampers traditional firms' ability to compete effectively [9] - The user engagement battle is intensifying, with AI-native applications rapidly gaining traction and threatening traditional applications' user time [10] Group 3: Strategic Responses from Major Companies - Major companies are attempting to integrate AI-native capabilities into their core businesses, recognizing the potential for scalable applications [11][21] - ByteDance is restructuring its AI organization to enhance agility and innovation, with a focus on AI-native talent [19][20] - Tencent is migrating its AI product lines to a more integrated structure, emphasizing collaboration with AI-native models [21] - Alibaba plans to invest over 380 billion yuan in AI infrastructure and aims for a comprehensive transformation across its core businesses [22] Group 4: Future Directions and Organizational Evolution - The evolution of organizational forms will be crucial as companies transition from traditional data-algorithm-traffic models to a model-data-agent framework [27] - Companies must focus on enhancing their organizational learning speed to convert technological breakthroughs into business cycles effectively [27] - The historical challenges of organizational inertia must be addressed to facilitate meaningful transformation in response to AI-native competition [25][26]

原标题：以科技创新增强抵御外部冲击的底气 我们已与美国打了8年贸易战，而这8年正是中国经济发展"含新量"不断上升的8年。无论是所谓"小 院高墙"，还是"脱钩断链"，都没能阻挡中国经济发展和科技进步。中国科技创新实力正在从量的积累 迈向质的飞跃，从点的突破迈向系统能力提升，在集成电路、人工智能、人形机器人等领域创新成果密 集井喷，带动了产业发展，提振了市场信心，增强了抵御风浪的能力。 如今，面对美政府变本加厉的霸凌行径，我们更要以破釜沉舟的决心和勇气坚定不移做"困难而正 确"的事，闯出科技创新"华山一条道"。从某种意义上来说，外部的遏制打压是一种"反向激励"，它倒 逼我们走出"舒适区"，摆脱创新惰性，激发内在潜能与活力。从华为的绝地反击到DeepSeek的横空出 世，无不验证了"哪里有封锁，哪里就有突围"。在攻关核心技术的实验室里，在产教融合的课堂上，在 产业升级的车间中，每一份创新的努力，都积蓄起中国勇开顶风船的更大能量。 外部环境越是严峻复杂，我们就越要扎实推动科技创新和产业创新深度融合，加快现代化产业体系 建设，让中国全产业链优势无可替代。我们可以依托持续科技创新不断提升产品附加值，增强出口产品 的国际竞 ...

Market Reaction to Tariffs - U.S. stocks experienced significant declines following President Trump's announcement of 10% tariffs on all trading partners, leading to corrections and bear markets for major equity indexes, with the Nasdaq entering bear market territory [1] - On April 9, Trump announced a temporary reduction in tariff rates to 10% for 90 days, resulting in a historic surge across U.S. markets [2] Market Performance - The S&P 500 surged over 9%, marking its third-largest single-day gain since World War II, while the Dow posted its strongest percentage gain since March 2020, and the Nasdaq Composite had its best single-day performance since January 2001 [3] - The Nasdaq-100-based ETF, Invesco QQQ Trust, Series 1 (QQQ), increased by approximately 12% on April 9 [3] Semiconductor Sector Performance - Semiconductor stocks were particularly notable, with six of the top eight performers in the Nasdaq-100 being semiconductor companies, and the VanEck Semiconductor ETF (SMH) advancing 17.2% on April 9 [4] - Top-performing chip stocks included Microchip Technology Inc (up 27.1%), Arm Holdings PLC (up 24.8%), Advanced Micro Devices Inc (up 23.8%), ON Semiconductor Corp (up 22.7%), Marvell Technology Inc (up 21.9%), and NXP Semiconductors NV (up 21.1%) [5] AI Market Developments - Companies are increasingly showcasing their AI product roadmaps, positively impacting chip stocks, with OpenAI valued at $300 million in a new $40 billion fundraising round led by SoftBank [6] - OpenAI's CEO announced that GPT-5 is set to launch in "a few months," indicating advancements beyond initial expectations [7] - Meta launched the first version of its open-source Llama 4 AI model family, with the most powerful model still under development [8] - Other key players in AI include Google with Gemini 2.5, Anthropic with Claude AI models, Elon Musk's xAI with Grok models, and China's DeepSeek preparing new open-source models [9] Semiconductor Industry Challenges - NVIDIA, a leading semiconductor company, faced a 17% stock drop in 2025, with concerns over margin pressure as it scales production of next-gen Blackwell AI chips [11] - In 2023 and 2024, semiconductor stocks outperformed software companies, but interest in data center infrastructure stocks has cooled in 2025, leading to increased scrutiny of AI stocks [13] - The U.S. is tightening restrictions on semiconductor exports to China, particularly targeting high-performance chips for AI systems and advanced computing [13] Investment Opportunities - Despite challenges, developments in the AI field have kept the semiconductor sector active, with recent valuation corrections seen as beneficial [14] - Investing in semiconductor ETFs such as SMH, iShares Semiconductor ETF (SOXX), SPDR S&P Semiconductor ETF (XSD), and Invesco Semiconductors ETF (PSI) could be advantageous if market recovery continues [14]

" 这次，充当中间商的云大厂没能赚到差价。 " 作者丨徐晓飞 编辑丨周蕾 "一天十几个客户电话进来，有的一打一个多小时，我中午饭都没顾上吃。"云大厂政企销售小顾哑着嗓子 告诉雷峰网，开源大模型DeepSeek出现后，市场反馈热烈，云大厂这边也接到不少客户来电，咨询相关 的API接入和线下部署方案。 不过小顾坦言，DeepSeek眼下这么火爆，主要是因为它帮国人把大模型的信心又找回来了，但最终能转 化成多少商业收益，能为云大厂带来多少拉动，还要打个问号。 究其原因在于，DeepSeek虽然降低了大模型的成本门槛，但它的能力跟ChatGPT差别不大，也面临和 ChatGPT一样的困境：应用场景实在有限。 DeepSeek爆火后，作为产业链上的核心一环，云大厂们迅速跟进，手里的"战略牌"一张连着一张打了出 来：组织变阵，业务调整，大笔投入，战略重构……整个一个目不暇接。 然而，这一热潮是真正的价值机遇，还是昙花一现的浮躁之风，跟还是不跟？成为摆在云大厂面前的一道 难题。 雷峰网通过调研了解到，眼下一众国内云大厂可分成两派：一派是推出DeepSeek一体机硬件部署方案 的，有阿里、火山、华为、百度等；一派是没推出的， ...

点 击 蓝 字 关 注 我 们 本文全文共 1322 字，阅读全文约需 4 分钟 近期，DeepSeek大火，电信、云计算、芯片、金融、汽车、手机等多领域已有超过200家头部企业宣布接 入DeepSeek。聚焦汽车行业，吉利、岚图、智己、长城、广汽、长安、奇瑞等20多个车企纷纷宣布与 DeepSeek深度融合，标志着汽车智能化进程的加速。 DeepSeek介绍 近期，DeepSeek接连发布开源大语言模型DeepSeek-V3、推理模型DeepSeek-R1，凭借低成本、高性能、 开源等特性在全球范围内迅速爆火出圈，国内外众多企业纷纷宣布接入DeepSeek模型，打破了传统AI市场的 格局，为整个行业带来新的活力和可能性。 DeepSeek热潮来袭，国内车企纷纷加急大模型"上车" 借助DeepSeek的理解与推理能力，提升车辆座舱语音交互、感知决策等多方面能力，为用户带来更加智 能化、个性化的用车体验；从技术实现路径看，车企主要采用直接接入、多模型联合协同部署、模型深度融合 与蒸馏三大接入方式。 大模型在接入智驾方面有前景，但仍具较大挑战 一方面，DeepSeek为智驾带来了算法、算力和数据多维度的借鉴，助力 ...

Group 1 - The article discusses the competitive landscape of AI-generated content (AIGC) and highlights that traditional internet giants are not the primary players leading the current wave of AIGC technology [1] - Companies like Baidu and ByteDance have quickly launched their AI products, but smaller firms exhibit greater agility and cost control compared to larger corporations [2][3] - The uncertain commercial prospects of chatbot products lead larger companies to be more cautious in resource allocation, as these products do not generate the same addictive engagement as mobile games or short videos [3][4] Group 2 - The emergence of DeepSeek has prompted major companies to reassess the commercial value of AI to consumer (AI to C) applications [5][6] - Tencent views DeepSeek as a starting point for scaling AI, while other major players are also heavily investing in AI technologies [6][7] - The article emphasizes that the competition among large companies will intensify as they strive to develop better technologies and products in the AI to C space [8] Group 3 - Despite launching later than competitors, ByteDance's chatbot product, Doubao, achieved significant user engagement through effective marketing strategies [15][16] - Tencent's AI assistant, Tencent Yuanbao, has seen rapid user growth after integrating DeepSeek, indicating a shift in its AI strategy [19][20] - Alibaba's Quark has been upgraded to become the group's flagship AI application, aiming to integrate various user functions through AI [24][25] Group 4 - The article raises questions about the commercial viability of AI to C applications, noting that the subscription model used by OpenAI may not be applicable in the domestic market [29][30] - The current focus for AI applications is on building market recognition and user retention rather than immediate monetization [30][31] - Major companies are exploring the potential of AI applications, with significant investments in GPU procurement to support their AI initiatives [33][34] Group 5 - The competition for AIGC market share is likened to previous battles for digital traffic, with companies needing to establish brand recognition to attract users [35][37] - The article notes that while there is no clear commercial model for consumer-level AI applications, the demand for customized industry models is growing [36] - Companies are rapidly iterating their AI products to enhance functionality and user experience, with Tencent's Yuanbao undergoing frequent updates [39][41]

Core Viewpoint - The article discusses the impact of geopolitical tensions and supply chain constraints on the AI supply chain, particularly focusing on TSMC's CoWoS capacity and its implications for the semiconductor industry [1][2][3]. Group 1: TSMC's CoWoS Capacity and Forecasts - Morgan Stanley predicts that TSMC's CoWoS capacity allocation for 2026 will be a crucial catalyst for the global AI supply chain [1]. - Goldman Sachs has revised its CoWoS shipment forecasts for 2025 and 2026 down to 585,000 and 923,000 units, respectively, due to supply chain limitations and geopolitical uncertainties [1]. - TSMC's monthly CoWoS capacity is expected to increase significantly from 70,000-75,000 units at the end of this year to approximately 105,000-125,000 units by mid-2026 [1][2]. Group 2: Geopolitical and Economic Impacts - The article highlights the potential long-term effects of Trump's tariff policies on the global economy and future AI capital expenditures [2]. - Despite the geopolitical challenges, demand for AI inference in China remains strong, leading to an increase in the spot prices of NVIDIA graphics cards [2]. - Goldman Sachs maintains a capital expenditure forecast of $40 billion for 2025 but has adjusted the 2026 forecast down from $46 billion to $45 billion due to slower CoWoS capacity expansion [2]. Group 3: Competitive Landscape and Market Dynamics - The emergence of DeepSeek, a Chinese startup, is noted for significantly reducing AI usage costs, which could disrupt NVIDIA's dominance in the AI GPU market [3][4]. - If DeepSeek successfully challenges NVIDIA, it may lead to a broader expansion of the AI ASIC chip market, benefiting TSMC's customer base, which includes major companies like Google, Meta, and Apple [5]. - Despite the challenges posed by DeepSeek and geopolitical tensions, NVIDIA and TSMC have not yet revised their AI GPU and CoWoS demand forecasts [5][6]. Group 4: Future Capacity Projections - TSMC's CoWoS monthly capacity is projected to reach 75,000-80,000 units by 2025, with further increases expected in subsequent years, reaching 150,000 units by 2029 [6]. - The breakdown of CoWoS capacity by type indicates a significant ramp-up in CoWoS-S and CoWoS-L segments, with respective targets of over 20,000 and 45,000 units by 2025 [6].

Nvidia CEO Jensen Huang appears to have struck a deal with the Trump administration to avoid export restrictions on the company’s H20 AI chips.The H20, the most advanced Nvidia-produced AI chip that can still be exported from the U.S. to China, was reportedly spared thanks to a promise from Huang to invest in new AI data centers in the U.S. According to NPR, Huang made the proposal during a dinner at Trump’s Mar-a-Lago resort sometime last week. Nvidia declined to comment.Many in the semiconductor industr ...

自今年二月起，DeepSeek 便因其开源旗舰级推理模型DeepSeek-R1 而引发全球瞩目——该模型性能 堪比全球前沿推理模型。其独特价值不仅体现在卓越的性能表现，更在于仅使用约2000块NVIDIA H800 GPU 就完成了训练（H800 是H100 的缩减版出口合规替代方案），这一成就堪称效率优化的 典范。 几天后，Elon Musk 旗下xAI 发布了迄今最先进的Grok-3 模型，其性能表现略优于DeepSeek-R1、 OpenAI 的GPT-o1 以及谷歌的Gemini 2。与DeepSeek-R1 不同，Grok-3 属于闭源模型，其训练动用 了惊人的约20万块H100 GPU，依托xAI "巨像"超级计算机完成，标志着计算规模实现了巨大飞跃。 xAI "巨像" 数据中心 Grok-3 展现了无妥协的规模扩张——约200,000块NVIDIA H100 显卡追求前沿性能提升。而 DeepSeek-R1 仅用少量计算资源就实现了相近的性能，这表明创新的架构设计和数据策展能够 与蛮力计算相抗衡。 效率正成为一种趋势性策略，而非限制条件。DeepSeek 的成功重新定义了AI扩展方式的讨 论。我 ...