12月8日消息，近日由美国共和党参议员皮特与民主党参议员克里斯领衔、跨党派议员联合提出的该法 案， 要求美国商务部在未来30个月内禁止向中国、俄罗斯等国家出口任何性能优于当前许可标准的先 进AI芯片！ 例如，英伟达等曾依赖中国市场出口的美国芯片厂商直言这次管制失败。由于无法向中国销售高端 GPU，他们损失了大量订单。随着美国高端芯片封锁，中国市场需求转向国内，促使本土芯片厂商加快 研发步伐。预计到2026年，中国本土AI芯片在国内市场的占有率将升至约40%。 转自：国芯网 声明：本文仅为信息交流之用，不构成任何投资建议，股市有风险，投资需谨慎。 加入"中国IC独角兽联盟"，请点击进入 是说芯语转载，欢迎关注分享 这意味着，即便英伟达、AMD等企业已获部分特供芯片（如H20、MI308）的出口许可，其下一代产品 （如H200）也将被彻底挡在门外。此举被视为国会"对华鹰派"对特朗普政府潜在政策松动的一次强力 制衡。 据路透社报道，就在法案提出前夕，美国白宫正内部评估是否允许英伟达向中国出口更强大的H200 AI 芯片——此举虽可缓解美企营收压力，却被批评者视为"严重偏离2022年以来确立的对华技术封锁方 针"。 ...

Core Viewpoint - The article discusses the recent bipartisan legislative proposal in the U.S. to impose stricter export controls on advanced AI chips to China and other countries, highlighting the implications for both U.S. and Chinese semiconductor industries [2][4]. Group 1: Legislative Actions - A bipartisan bill led by U.S. Senators Peter and Chris aims to prohibit the export of advanced AI chips to China and Russia for the next 30 months [2]. - This legislation serves as a counterbalance to potential policy relaxations by the Trump administration regarding technology exports to China [4]. - The proposal reflects a continuation of the U.S. policy of alternating between tightening and loosening restrictions on semiconductor exports to China, dating back to early 2025 [4]. Group 2: Impact on U.S. Companies - Companies like NVIDIA and AMD, which previously received export licenses for certain chips, will now face restrictions on their next-generation products, such as the H200 AI chip [4]. - The inability to sell high-end GPUs to China has resulted in significant order losses for U.S. chip manufacturers, indicating a failure of the export control strategy [5]. Group 3: China's Semiconductor Industry - Despite the tightening of U.S. export controls, China's AI development continues to thrive, driven by domestic innovation and a shift towards local chip alternatives [4][5]. - The market share of domestic AI chips in China is projected to rise to approximately 40% by 2026, as local manufacturers accelerate their research and development efforts [5].

值得注意的是，AMD 在今年早些时候与 OpenAI 达成了一项重大合作协议，其中后者将部署 6 吉瓦（GW）级 AMD Instinct GPU，同时 AMD 允许 OpenAI 以每股 1 美分（现汇率约合 0.071 元人民币）的价格购买 1.6 亿股 AMD 股票，等于给了 OpenAI 约 10% 股份。 谈到英伟达等友商时，苏姿丰认为，她并不对英伟达的竞争报以担忧，甚至不担心谷歌、亚马逊等正在打造自家 芯片的互联网巨头，她对此解释道："真正让我夜不能寐的问题是：我们如何在创新上跑得更快"？ 不过她认为，AI 仍处于"婴儿期"，AMD 必须为未来准备好更多芯片。 当主持人问她，科技行业是否处于"AI 泡沫"时，苏姿丰回应道："从我的角度看，绝对不是。AI 行业未来需要来 自 AMD 等公司的海量芯片，现在外界对'泡沫'的担忧有些被夸大了"。 苏姿丰的这番话听起来非常大胆，符合她的个人风格。她从 2014 年出任 AMD CEO 以来，将公司市值从 20 亿美 元（IT之家注：现汇率约合 141.57 亿元人民币）提升至 3000 亿美元（现汇率约合 2.12 万亿元人民币），并推出 Zen 架构处理 ...

Investment Rating - The report does not explicitly state an overall investment rating for the semiconductor industry or specific companies within it Core Insights - The semiconductor industry is experiencing a significant AI super cycle, with companies like AMD expecting substantial growth in AI workloads and custom ASICs capturing 20-25% of the total addressable market [2][15] - KLAC anticipates growth in its TSMC business and higher investment from INTC, indicating a positive outlook for memory and advanced packaging sectors [3][16] - CoreWeave reports an "insatiable" demand for compute resources, highlighting the shift in AI workloads from training to more compute-intensive inferencing [17] - ARM is making strides in data center CPU development, with expectations of increased revenue opportunities through its CSS model [5][18] - MCHP has raised guidance due to strong order flow and backlog, indicating robust demand across multiple sectors [6][19] - INTC is cautious about supply constraints peaking in 1Q26, while also seeing increased demand driven by AI applications [9][20] Summary by Company AMD - AMD is in a ten-year AI super cycle, expecting significant capital deployment from large companies [2] - The company anticipates multiple GW-scale customers beyond OpenAI, indicating strong market confidence [15] KLAC - KLAC expects low to mid-single-digit growth in C1H:26 due to memory pull-ins and reassured investors about N2 monetization continuing into C2026 [3][16] CoreWeave - CoreWeave's backlog is approximately 55 billion, indicating a strong demand for AI compute resources [17] ARM - ARM is progressing on a data center CPU chip for a major customer and sees significant revenue potential through its CSS model [5][18] MCHP - MCHP reported phenomenal orders in November and expects CQ1 to be significantly above normal seasonal levels [6][19] INTC - INTC reiterated that supply constraints are expected to peak in 1Q26, with a focus on AI-driven demand for CPUs [9][20] PI - PI sees growth from deeper apparel adoption and logistics deployments, with a focus on improving gross margins [10][21] AMBQ - AMBQ's customer base is stable, with expectations for diversified growth in 2027 [11][22] ENTG - ENTG is focused on improving gross margins and may consider consolidating capacity to enhance sentiment [12] SMTC - SMTC is optimistic about its CopperEdge ramp and the adoption of LPO technology among hyperscalers [13][23] Anthropic - Anthropic is constrained by compute capacity, indicating strong demand for AI resources [14] Impinj - Impinj is focused on expanding its RAIN market presence, with significant growth opportunities in logistics and food sectors [21] Semtech - Semtech is preparing for a ramp in its ACC technology to support major hyperscaler deployments [23]

For months, the Trump administration has been thundering about America’s artificial Intelligence (AI) revival. But now, the spotlight is swiveling to a new space – robotics. The shift did not come with fanfare, just a steady drumbeat of meetings between Commerce Secretary Howard Lutnick and the country’s biggest robotics founders. Those conversations, insiders say, are laying the groundwork for a potential executive order that could ignite a full-scale national robotics push. Agencies are scrambling to al ...

NAND wafer 合约价格大幅提升，AI 从云到端 拥抱新机遇 [Table_ReportDate] 2025 年 12 月 7 日 证券研究报告 行业研究 [行业周报 Table_ReportType] | [Table_StockAndRank] 电子 | | | --- | --- | | 投资评级 | 看好 | | 上次评级 | 看好 | [Table_Author] 莫文宇 电子行业首席分析师 执业编号：S1500522090001 邮 箱：mowenyu@cindasc.com 杨宇轩 电子行业分析师 执业编号：S1500525010001 邮箱：yangyuxuan@cindasc.com 王义夫 电子行业分析师 执业编号：S1500525090001 邮 箱：wangyifu@cindasc.com 信达证券股份有限公司 CINDA SECURITIES CO.,LTD 北京市西城区宣武门西大街甲127号金隅 大厦B座 邮编：100031 [Table_Title] NAND wafer 合约价格大幅提升，AI 从 云到端拥抱新机遇 [Table_ReportDate] 2025 年 12 ...

Core Viewpoint - The competition between Nvidia and AMD in the AI infrastructure market is intensifying, with AMD showing strong performance in 2025 and potential for further growth in 2026 [1][3]. Nvidia's Position - Nvidia holds over 90% market share in the data center GPU space, benefiting from its established CUDA software platform, which is crucial for training large language models [3][5]. - Nvidia's stock is trading at a forward P/E of 24, lower than AMD's 34, and it has experienced faster revenue growth, with a 62% increase last quarter compared to AMD's 36% [5]. - The company has a market capitalization of $4,433 billion and a gross margin of 70.05% [4]. AMD's Potential - AMD's data center revenue is significantly lower than Nvidia's, but capturing market share in the AI infrastructure space could lead to substantial growth [6]. - AMD has established a niche in the inference market, where it can compete more effectively against Nvidia, as the inference market is expected to grow larger than the training market [7]. - AMD has partnered with OpenAI, which includes a potential 10% stake in AMD and a supply agreement for up to six gigawatts of GPUs, valued at over $200 billion [9][10]. - AMD aims for over 35% compound annual revenue growth over the next three to five years, with more than 60% growth in data center revenue and an 80% increase in AI revenue [11]. Conclusion - Both Nvidia and AMD are expected to perform well in 2026, but AMD may have the edge due to its partnerships and growth potential in the AI market [13].

图片来自AI生成 【数智周报将整合本周最重要的企业级服务、云计算、大数据领域的前沿趋势、重磅政策及行研报 告。】观点华为任正非：大量建设大模型是正确的探索，未来算力一定过剩；AI重在"应用"，着眼于未 来3-5年 华为创始人兼总裁任正非与ICPC（国际大学生程序设计竞赛）主席、教练及获奖选手座谈会纪要公 布。在回答关于计算资源的问题时，任正非直言，"以后是算力过剩的，不是你想象的算力不足"。他认 为，建数千个、数百个大模型都是正确的探索。华为在解释超节点时提到了950、960、970等型号， 但"我们需要多少个'970'，970在哪儿能用？需要多少台，集群怎么连接"，任正非表示，这是一种线性 技术推演，是一定能实现的算力假设。而需求是不是一种线性结构尚未可知，"万一是非线性的呢"。因 此，他说："算力过剩的时代一定会到来的。" 谈及当下火热的AI，任正非的观点很明确：AI重在应用，不重在发明。"发明AI只是一家IT公司，应用 AI是会强大一个国家。"任正非表示，IT公司对人类的贡献只2%，AI在产业上的贡献则会占到98%。仅 以洗煤、炼钢为例，洗煤精度提高0.1%，乘以中国40亿吨的煤炭产量，价值惊人；高炉炼 ...

公众号记得加星标⭐️，第一时间看推送不会错过。 AMD之所以面临更高的发热量和更低的每瓦性能，原因很简单：Zen 5及其所有前代产品都是为数据 中心市场而设计的。知情工程师不断向我们透露，Infinity Fabric架构在移动领域功耗过高，难以实 现高效设计。此外，AMD也缺乏强有力的小核心策略。 福雷斯特·诺罗德领导的数据中心团队之所以能够蓬勃发展，得益于以数据中心为先导、覆盖面广且 功能强大的核心战略。而消费者和商业移动业务部门则不得不接受 Zen 5 的不足之处。 4nm工艺很难与英特尔18A和台积电3nm工艺竞争。 上周，AMD向金融和行业分析师展示了其未来几年的人工智能、数据中心和消费级产品计划。尽管 在数据中心市场取得了巨大成功，人工智能领域也取得了一定的进展，并且未来发展路线图清晰明 确，但我仍想指出其消费级产品路线图的一个不足之处，尤其是在2026年方面。 计算与图形事业部高级副总裁兼总经理Jack Huynh及其团队正朝着正确的方向前进；X3D台式机的 成功对微软来说是一项重大胜利。然而，在消费业务的某些关键领域，尤其是在消费级和商用笔记本 电脑领域，不太可能在2026年就立即取得成功。公 ...

美国参院两党议员4日提出一项法案，旨在将现行对先进半导体的出口管制措施正式立法，要求行政部 门未来至少30个月内拒发任何先进芯片出口至中国大陆、俄罗斯等国家的许可。此举可能导致英伟达无 法向中国大陆出口其先进AI芯片；AMD与Google的芯片也将受到规范。 这项被称为《安全芯片法案》（SAFE Act）的提案，要求美国商务部在未来两年半内，停止核发高于 现有许可性能标准的芯片出口许可。 SAFE法案提出的时机，恰巧发生在英伟达执行长黄仁勋与参院银行委员会共和党议员闭门会谈的隔 天，该委员会负责出口管制政策。黄仁勋前一天也和美国总统川普私下会面，他表示双方讨论出口管制 议题，但未透露细节。 法案获参议院外交委员会跨党派议员支持，被视为维持美国在人工智能领域对中国优势的关键举措。 这也对英伟达构成新挑战，该公司一直在游说川普政府与国会放宽出口限制，以维持其在中国AI芯片 市场的领先地位。此项法案提出前一天才传出，英伟达才刚成功游说国会暂缓另一项出口管制法案 GAIN AI Act（保障国家人工智能获取与创新法案）。 目前川普政府正评估是否允许英伟达对中国销售H200芯片。若同意放行，将显著偏离自2022年起实施 ...