文 | 第一新声 11月26日凌晨1点，英伟达罕见发布紧急声明，回应近期市场对AI硬件发展路线的热议。业界普遍认 为，此举与谷歌第七代Ironwood TPU正式上市，以及Meta考虑采用TPU方案等行业动态密切相关。 值得一提的是，尽管谷歌成功开发了TPU系列芯片，但其仍在采购英伟达GPU。这凸显出当前AI芯片市 场的复杂态势。专用芯片虽能在特定场景展现优异性能，但通用平台在实际工作负载适配、灵活扩展性 和生态系统方面仍具优势。 01 谷歌用Ironwood+Axion重构AI基建 如果说过去十年，AI基础设施基本等同于"GPU+云+自行管理架构"的组合，那么现在Google正试图用 一套集"定制硬件、云服务、托管支持、专用芯片、高速网络、生态协同与弹性扩展"于一体的方案，重 新定义AI基础设施的核心形态。 其中，Ironwood作为Google的第七代TPU，是其迄今为止性能最强、能效最高的加速器，专门针对大规 模模型训练、推理、模型服务及智能体（Agent）工作流设计。 一方面，Ironwood实现了性能与能效的双重突破，10倍性能提升搭配高效能耗与大规模扩展能力，对 大模型训练、推理及服务场景极为友好 ...

The Gemini app reached 650 million monthly active users at the end of the third quarter, with total queries also increasing threefold from the previous quarter. Alphabet is also leveraging AI extensively across its offerings, as evidenced by the over 1.3 quadrillion monthly tokens processed by the end of the third quarter, representing over 20x year-over-year growth.In the third quarter of fiscal 2025, Alphabet's revenues were up 16% year-over-year to $102.3 billion. This was the first quarter in the compan ...

Core Insights - Alphabet is focusing on AI as a core compute product and enterprise security layer, amidst pressures and valuation questions [2][3] - The company is investing heavily in in-house technology, including the Ironwood TPU and Axiom CPU, to control the entire tech stack [3][6] - Google is integrating AI across various consumer-facing products like YouTube and Gmail, enhancing its capabilities throughout the stack [4][7] AI Strategy - Google's strategy contrasts with competitors like Microsoft and AWS, as it aims to control its technology from hardware to software, including its own large language model, Gemini [6][12] - The integration of AI into products is seen as a significant growth opportunity for Google, with a focus on innovation across its applications [12][17] Financial Performance - Google reported a cloud backlog of $155 billion, reflecting an 82% year-over-year increase, indicating strong demand for its services [13] - The company is adopting a capital expenditure-heavy investment strategy, which is deemed sensible given its strong balance sheet and the competitive landscape [14][17] Market Position - The current investment climate is characterized by a need for substantial spending to stay competitive, with overinvestment viewed as a preferable strategy compared to underinvestment [15][17] - Enterprises are likely to evaluate cloud providers based on their AI capabilities, positioning Google favorably in the market [17]

Core Viewpoint - NVIDIA's main competitor in the AI hardware race is Google, not AMD or Intel, as highlighted by the recent launch of Google's Ironwood TPU, which significantly enhances its competitive position against NVIDIA [2][10]. Group 1: Ironwood TPU Specifications - Google's Ironwood TPU features 192GB of HBM memory with a peak floating-point performance of 4,614 TFLOPs, representing a nearly 16-fold improvement over TPU v4 [5][4]. - The Ironwood TPU Superpod can contain 9,216 chips, achieving a cumulative performance of approximately 42.5 exaFLOPS [5][4]. - The inter-chip interconnect (ICI) technology allows for a scalable network, connecting 43 modules, each with 64 chips, through a 1.8 PB network [3]. Group 2: Performance Improvements - Compared to TPU v5p, Ironwood's peak performance has increased by 10 times, and it shows a 4-fold improvement over TPU v6e in both training and inference workloads [4][6]. - The architecture of Ironwood is specifically designed for inference, focusing on low latency and high energy efficiency, which is crucial for large-scale data center operations [6][7]. Group 3: Competitive Landscape - The AI competition is shifting from maximizing TFLOPS to achieving lower latency, cost, and power consumption, positioning Google to potentially surpass NVIDIA in the inference market [10]. - Google's Ironwood TPU is expected to be exclusively available through Google Cloud, which may lead to ecosystem lock-in, posing a significant threat to NVIDIA's dominance in AI [10]. Group 4: Industry Insights - The increasing focus on inference queries over training tasks indicates a shift in the AI landscape, making Google's advancements in TPU technology particularly relevant [6][10]. - NVIDIA acknowledges the rise of inference technology and is working on its own solutions, but Google is positioning itself as a formidable competitor in this space [10].

Core Insights - Google has launched its seventh-generation Tensor Processing Unit (TPU), named Ironwood, designed specifically for AI workloads, marking a significant advancement in AI computing capabilities [1][2][3] - The new TPU offers a 10X peak performance improvement over the previous generation and more than 4X better performance per chip for both training and inference tasks [3] - Google is positioning itself in the "age of inference," where the focus shifts from training AI models to utilizing them for practical applications, anticipating a surge in demand for AI computing [5][9] Product Launch and Features - Ironwood TPUs will be available for Google Cloud customers soon, alongside new Arm-based Axion virtual machine instances that enhance performance per dollar [2] - The Ironwood TPU is optimized for high-volume AI inference workloads, which require quick response times and the ability to handle numerous requests [4] Market Position and Growth - Google Cloud generated $15.2 billion in revenue in Q3, reflecting a 34% year-over-year increase, with an operating income of $3.6 billion and an operating margin of approximately 24% [8] - The cloud computing sector is competitive, with Microsoft Azure and Amazon Web Services also expanding their AI capabilities, but Google is leveraging its decade-long experience in TPU development to gain an edge [7][9] Strategic Partnerships - AI companies like Anthropic are expanding their use of Google's TPUs, with a new deal granting access to 1 million TPUs, which is crucial for their goal of reaching $70 billion in revenue by 2028 [6]

Group 1 - Yu Minhong confirmed the departure of Sun Dongxu from Dongfang Zhenxuan, highlighting Sun's contributions to the company's development [1] - Dongfang Zhenxuan was established in October 2021 with a registered capital of 10 million RMB, and is involved in various sectors including broadcasting and cultural operations [1] - Sun Dongxu's company has successfully registered multiple trademarks and copyrights related to its brand and products [1] Group 2 - Xiaopeng Motors addressed concerns about its IRON robot by publicly demonstrating its internal mechanical structure during a product launch event [3] - Tesla shareholders approved Elon Musk's compensation plan, allowing him to potentially earn up to $878 billion in stock over the next decade [6] - Xiaomi announced the discontinuation of its calling service effective December 3, 2025, due to the availability of similar features in other applications [6] Group 3 - Xiaohongshu has obtained a payment license through its subsidiary, enabling it to conduct internet payment services [8] - ByteDance's Doubao PC product head Qi Junyuan has left the company, with reports of interest from multiple dollar funds [9] - Wang Yibo's contribution to Lehua Entertainment's revenue is reported to exceed 60% in 2024, although the company claims this percentage is gradually decreasing [9] Group 4 - Huawei's Mate 70 Air launched with a market price increase of 300 RMB shortly after its release [9] - Fat Donglai Supermarket rewarded 20 employees with a total of 200,000 RMB for their bravery during an emergency incident [11] - Rabbit Mom responded to a product quality issue by confirming that the affected batch of shampoo was destroyed before reaching the market [14] Group 5 - Nissan announced the sale of its global headquarters for 97 billion JPY (approximately 4.5 billion RMB) as part of its business restructuring efforts [14] - Nissan established a joint venture for vehicle import and export in China, marking a significant step in its local operations [15] - Google is reportedly negotiating to increase its investment in AI startup Anthropic, which may exceed a valuation of 350 billion USD [15] Group 6 - Yushi Technology completed a Pre-A+ round financing of several million RMB to enhance its HPC and AI infrastructure [16] - Jabil announced the acquisition of Hanley Energy Group for approximately 725 million USD, with potential additional payments based on future performance [16] - Leap Bio completed nearly 100 million RMB in A+ round financing, focusing on cell therapy for various diseases [17] Group 7 - Google is set to launch its most powerful AI chip, Ironwood, aimed at enhancing AI model training and real-time applications [17] - iFlytek released the Spark X1.5 deep reasoning model, achieving performance metrics comparable to leading international models [18] - Yushu Technology introduced a full-body remote operation platform, showcasing its robot's capabilities in household tasks [19] Group 8 - ByteDance launched a free music app, Tomato Music, to diversify its music product offerings and cater to different user needs [21] - The China Association of Automobile Manufacturers reported a significant decline in vehicle imports, with September figures showing a 25.6% year-on-year drop [21]

Core Points - The U.S. job market shows signs of deterioration, leading to a collective decline in major tech stocks, with the S&P 500 down 1.12%, Nasdaq down 1.9%, and Dow Jones down 0.84% [1] - OpenAI's CFO Sarah Friar mentioned the need for a financial ecosystem involving banks, private equity, and federal government support for massive chip investments, which sparked concerns about an "AI bubble" [3][4] - OpenAI's CEO Sam Altman clarified that the company does not seek government guarantees for its data center investments and emphasized the importance of a national strategic computing reserve for public benefit [6][7] Company Performance - Major tech stocks experienced declines, with Nvidia down 3.65%, Apple down 0.14%, Microsoft down 1.98%, and Amazon down 2.86% [9] - Duolingo, a language learning platform, saw a significant drop of 25% due to disappointing earnings guidance, marking its largest single-day decline [9] - Chinese stocks showed mixed results, with Alibaba up 1.69% and JD down 0.28% [9] Economic Indicators - The Challenger job-cut report indicated that U.S. employers cut over 150,000 jobs in October, the highest number for that month in over two decades, influenced by cost-cutting and AI adoption [8] - Vanguard's economist noted that the report's predictive ability for future labor market conditions has historically been poor, but current layoffs pose greater risks compared to 2022 [8] Other Industry News - Google Cloud launched its seventh-generation TPU, "Ironwood," which offers over four times the performance of its predecessor, enabling the connection of up to 9,216 chips in a single cluster [11] - The U.S. government updated its critical minerals list, adding copper, silver, and uranium, which caused significant fluctuations in the mining sector [12]

Group 1 - Google is launching its most powerful chip, Ironwood, which is the seventh generation of its Tensor Processing Unit (TPU) and will be fully available in the coming weeks [1][3] - The new Ironwood TPU can connect up to 9,216 chips in a single pod, offering over four times the speed of its predecessor and eliminating data bottlenecks for demanding models [1][3] - AI startup Anthropic plans to use up to 1 million new TPUs to run its Claude model, indicating Google's competitive positioning against rivals like Microsoft and Amazon in building AI infrastructure [3] Group 2 - On the same day, Google announced plans for its largest investment in Germany, focusing on infrastructure and data center construction, as well as renewable energy and waste heat utilization projects [3] - The investment will expand Google's operations in Munich, Frankfurt, and Berlin, with details to be revealed in a press conference on November 11 [3] - Following the news, Google's stock saw a pre-market increase of over 1.5%, although the gains later moderated [3][4] Group 3 - Domestic PCB companies in China have experienced stock price surges, with leading PCB firm Dongshan Precision hitting its daily limit up, and other companies like Shenghong Technology and Pengding Holdings also seeing significant increases [6] - Reports indicate that Google has been in discussions with major domestic PCB manufacturers for potential orders related to AI chip production, which could significantly boost the PCB industry [5][6] - A report from Guojin Securities highlights a strong demand for AI-related ASICs and increased orders for AI-PCB companies, driven by major North American firms ramping up AI investments [8]

Group 1 - The U.S. stock market opened slightly lower, with the Dow Jones down 0.18%, S&P 500 down 0.16%, and Nasdaq down 0.17% [1] - Datadog's stock surged over 18% due to strong earnings performance [1] - Snap's stock increased over 17% after Q3 revenue exceeded expectations and a $400 million deal with Perplexity AI [1] - DoorDash's stock fell over 11% as Q3 profits did not meet expectations [1] Group 2 - Google is launching its most powerful AI chip, Ironwood, which will be fully available in the coming weeks [2] - Ironwood is Google's seventh-generation Tensor Processing Unit (TPU) and is reported to be over four times faster than its predecessor [2] - Major clients, including AI startup Anthropic, plan to use up to 1 million new TPUs to run their Claude model [2] - The new Ironwood TPU can connect up to 9,216 chips in a single pod, eliminating data bottlenecks for demanding models [2] - Google is in a high-stakes competition with Microsoft, Amazon, and Meta to build future AI infrastructure [2] - Google's TPUs are custom chips that offer advantages in price, performance, and efficiency compared to NVIDIA's GPUs [2] Group 3 - Chinese concept stocks mostly rose, with Xpeng Motors up over 6%, Baidu Group up over 3%, and Alibaba up over 2% [4]

Group 1 - The US stock market opened slightly lower on November 6, with the Dow Jones down 0.18%, S&P 500 down 0.16%, and Nasdaq down 0.17% [1] - Datadog's stock surged over 18% due to strong performance in its earnings report [1] - Snap's stock rose over 17% after exceeding revenue expectations for Q3 and securing a $400 million deal with Perplexity AI [1] - DoorDash's stock fell over 11% as its Q3 profits did not meet expectations [1] Group 2 - Google is launching its most powerful AI chip, Ironwood, which will be fully available in the coming weeks [2] - Ironwood, the seventh generation of Google's Tensor Processing Unit (TPU), is reported to be over four times faster than its predecessor [2] - Major clients, including AI startup Anthropic, plan to use up to 1 million new TPUs to run their Claude model [2] - The new Ironwood TPU can connect up to 9,216 chips in a single pod, eliminating data bottlenecks for demanding models [2] - Google is in a high-stakes competition with Microsoft, Amazon, and Meta to build future AI infrastructure [2] - Unlike most large language models that rely on NVIDIA GPUs, Google's TPUs are custom chips that offer advantages in price, performance, and efficiency [2] Group 3 - Chinese concept stocks mostly rose, with Xpeng Motors up over 6%, Baidu up over 3%, and Alibaba up over 2% [4]