Core Viewpoint - The forum highlighted the potential for China's economy to achieve stable medium- to long-term growth driven by total factor productivity improvements and the rise of new economic sectors [1][2]. Economic Outlook - China's average GDP growth over the past few years has been around 5%, supported by structural changes such as strong export performance despite U.S. tariffs and a transition from old to new economic sectors [2][3]. - The share of consumption in GDP is expected to increase relative to investment from 2022 to 2024, indicating a shift in the economic growth model [2]. - The real estate sector's contribution to GDP has decreased from over 20% at its peak to 15%, with ongoing adjustments expected [2]. Industry Development - Traditional consumer goods and manufacturing markets are saturated, while new consumer categories and services show significant growth potential [3]. - The export structure has shifted from low-end products to mid-range machinery and high-end products, contributing to export resilience [3]. Social Considerations - To ensure equitable economic development, efforts should be made to reduce disparities in social security and per capita consumption between urban and rural areas [3]. Future Growth Projections - The average potential growth rate for China is projected to remain above 4% over the next decade, with a target for per capita GDP to double [3][4]. - Historical data indicates that total factor productivity has experienced rapid growth during key periods of economic reform, characterized by increased marketization, optimized resource allocation, and heightened competition [3].

Core Insights - The WISE 2025 conference in Beijing emphasized the transformation of AI applications and the globalization of technology, highlighting the shift from traditional industry practices to immersive experiences in business [1][3]. Company Overview - GMI Cloud is a North American AI Native Cloud service provider and one of the first six Reference Cloud Partners of NVIDIA [2][6]. - The company focuses on AI infrastructure for overseas markets, offering three main product lines: computing hardware, cluster management, and inference services [7]. AI Application Trends - The current state of AI application development is described as "armed to the teeth," with a significant increase in active users, particularly in North America, where over 90% of knowledge workers are proficient in using AI tools [8][10]. - The demand for AI services in overseas markets has surged, driven by completed user education and a growing need for AI inference capabilities [10]. Challenges in AI Deployment - Key challenges in AI deployment include service timeliness, scalability, and stability, particularly as traditional software expansion methods are inadequate for AI applications [11]. - Rapid technological iteration in AI has led to fluctuating token prices, complicating the operational landscape for companies [12]. GMI Cloud's Strategic Initiatives - GMI Cloud is investing in building AI factories in collaboration with NVIDIA, aiming to enhance cluster throughput and support AI application efficiency [12]. - The company is iterating its cluster and inference engines to cater to different customer needs, with the cluster engine designed for technically capable clients and the inference engine for lightweight applications [12][14]. Inference Engine Features - The GMI Cloud Inference Engine supports global deployment and automatic scaling across clusters and regions, addressing the challenges faced by companies during peak traffic [16]. - It features a three-layer architecture for resource scheduling, ensuring efficient workload management based on sensitivity to latency and cost [16]. Future Outlook - By 2026, the paradigm of AI globalization is expected to evolve from a one-way technology output to a model of global value resonance, emphasizing a dual empowerment ecosystem for resources, technology, and demand [23].

Group 1 - The A-share market saw a strong performance on December 8, with all three major indices rising, particularly the ChiNext index which increased by over 2% [2] - The CPO sector experienced a significant surge, with stocks such as Dongtian Micro and Zhishang Technology hitting the 20% daily limit, while Yitian shares rose over 10% [2] - Historical highs were reached by companies like Zhongji Xuchuang and Tianfu Communication in the CPO sector [2] Group 2 - Amazon Web Services (AWS) introduced a new generation of AI training chips called Trainium at its annual cloud computing event re:Invent, and announced plans for the next generation product, Trainium 4 [2] - Morgan Stanley's latest research indicates that Google's TPU production is expected to increase significantly, which could lead to substantial revenue growth for Alphabet [2] - The forecast for Google's TPU production in 2027 was raised from approximately 3 million units to 5 million units, an increase of 67%, and the 2028 forecast was adjusted from about 3.2 million units to 7 million units, a rise of 120% [2]

计算机 分析师：唐月 登记编码：S0730512030001 tangyue@ccnew.com 021-50586737 证券研究报告-行业月报 强于大市(维持) 计算机相对沪深 300 指数表现 资料来源：中原证券研究所，聚源 -17% -11% -4% 2% 8% 15% 21% 28% 2024.12 2025.04 2025.08 2025.12 计算机 沪深300 相关报告 《计算机行业年度策略：AI 应用加快，全球格 局重塑中》 2025-12-01 《计算机行业分析报告：十五五规划建议的信 息科技领域内容解读》 2025-10-31 《计算机行业月报：鸿蒙迎来重要升级，AI 算 力需求多元化趋势明显》 2025-10-23 联系人：李智 预训练规模 ——计算机行业月报 发布日期：2025 年 12 月 08 日 风险提示：国际局势的不确定性；海外 AI 产业竞争格局变化带来市 场调整风险。 本报告版权属于中原证券股份有限公司 www.ccnew.com 请阅读最后一页各项声明 第1页 / 共50页 手机端 AI 应用加速，DeepSeek 将加大 投资要点： 电话： 0371-65585629 ...

Core Viewpoint - The annual meeting of the Shandong Think Tank Alliance focuses on high-quality development and seeks new paths to support the construction of a modern strong province, emphasizing the importance of think tanks in national strategy and regional development [1][3]. Group 1: Think Tank Development - The development of think tanks in China faces both historical opportunities and significant responsibilities, necessitating proactive changes and innovative approaches to enhance their contributions [3]. - Think tanks are encouraged to align their research with national strategic deployments and provincial priorities, producing strategic and forward-looking outcomes [3]. - The integration of artificial intelligence into think tank operations is highlighted as a key area for innovation, with a focus on establishing new research organizational forms suitable for the AI era [3]. Group 2: Educational Institutions' Role - Educational institutions, particularly Qilu Normal University, are recognized for their efforts in think tank research and construction, which are seen as vital for enhancing their overall strength and social service capabilities [5]. - The construction of new-type think tanks is essential for modernizing national governance systems and promoting high-quality regional development, with a strong emphasis on collaboration and long-term commitment [5]. - The university's think tank development is closely tied to the education sector, aiming to support the construction of a strong education province and contribute to regional economic and social development [5]. Group 3: Alliance Activities - During the annual meeting, the Shandong Think Tank Alliance held a council meeting to elect new vice presidents, secretaries, and approve new member units, reflecting the alliance's growth and collaborative spirit [7]. - The integration of party leadership into think tank construction is emphasized as a fundamental guarantee for progress, with a focus on aligning research outcomes with national and regional needs [7]. - The alliance has produced high-level think tank results that have received significant attention and adoption by central media and relevant departments, providing strong intellectual support for high-quality economic and social development [7].

此前，亚马逊还被曝出将启动新一轮裁员计划，预计裁减多达3万个企业岗位。这是自2022年底以来亚马逊规模最大的一次人员调整。亚马逊全球员工总数 约为154万，其中，企业岗位员工约35万。 裁员将波及多个核心部门，包括人力资源、设备与服务、运营以及亚马逊云服务等。知情人士透露，随着亚马逊财务优先事项的调整，裁员人数可能会有所 变动。 与此同时，亚马逊首席执行官安迪·贾西（Andy Jassy）近年来持续推进组织"扁平化"改革，减少管理层级，并设立匿名投诉热线以识别低效流程。 据雷峰网报道，近期亚马逊中国区内部持续动荡，中国副总裁杨钧（职级L8）已于一周前离职。 整个亚马逊全球开店大中华区仅有1位L10和4位L8高管，杨钧是4位L8之一，其离职在内部引发震动。杨钧离职后，其管理团队被拆分合并至其他部门。 据悉，杨钧已入职亚马逊十多年，长期负责亚马逊全球开店企业购在亚太区的业务，以及亚马逊全球开店在韩国、东南亚的整体跨境电商业务。 ...

Group 1 - The ChiNext Index (399006) has seen a significant increase, with a rise of over 3% during the session, currently up by 2.67%, indicating a potential three-day winning streak [1] - Notable stocks include Tianfu Communication (300394) up by 19.89%, Maiwei Co. (300751) up by 18.23%, and Jiangbolong (301308) up by 14.60% [1] - The Ministry of Industry and Information Technology held a meeting emphasizing the need to improve carbon peak and carbon neutrality management systems, accelerate technological upgrades in key industries, and enhance the supply of carbon-related technologies and standards [1] Group 2 - Guolian Minsheng Securities reported that despite increased volatility in the global tech sector, major US and Chinese tech giants showed stable operational performance, with AI-related business growth accelerating [2] - Companies like Microsoft, Amazon, and Alibaba recorded over 20% growth in cloud revenue, with management expressing optimism about the future of AI [2] - Zhongyin International Securities predicts that the A-share market will continue its upward trend supported by valuation and profit stabilization, with geopolitical risks easing and expectations of a Federal Reserve rate cut potentially stimulating the market [2] Group 3 - The MACD golden cross signal has formed, indicating positive momentum for certain stocks [3]

Group 1 - The core viewpoint of the news highlights significant growth in the cloud computing and robotics ETFs, with the Tianhong Cloud Computing ETF (517390) showing a notable increase in trading volume and market performance [1][3] - The Tianhong Cloud Computing ETF (517390) experienced a 3-month scale growth of 15.21 million yuan, indicating strong investor interest and market confidence [1] - The Tianhong Cloud Computing ETF tracks the CSI Hong Kong-Shenzhen Cloud Computing Industry Index (931470), which rose by 3.45%, with key stocks like Xinyi Technology (300502) and Zhongji Xuchuang (300308) showing substantial gains of 8.05% and 7.55% respectively [1] Group 2 - The Robotics ETF (159770) also demonstrated significant performance, with a trading volume of 333 million yuan and a net subscription of 43.5 million shares during the trading session [1][3] - Over the past week, the Robotics ETF (159770) saw a scale increase of 74.42 million yuan, and in the last month, it gained 23.6 million shares, reflecting strong demand [3] - The Robotics ETF closely follows the CSI Robotics Index (H30590), which rose by 0.97%, with notable stock performances from companies like Haimao Star (688559) and Jingye Intelligent (688290) [1][3] Group 3 - The news also mentions the launch of the "Yufeng" intelligent regulatory platform by Yipay at the Digital Intelligence Technology Ecological Conference, showcasing advancements in AI and cloud computing for risk management [5] - A joint venture between UBTECH and the Beijing Humanoid Robot Innovation Center established Tianyou Robotics Company, focusing on the humanoid robotics sector, with a registered capital of 10 million yuan [6] - CICC expresses optimism for technology growth investments in 2026, emphasizing opportunities in the mechanical industry driven by technological innovation and domestic demand recovery [6]

Investment Rating - The report does not explicitly state an investment rating for the industry or specific companies involved Core Insights - AWS is accelerating its in-house AI development with the introduction of the next-generation AI training chip Trainium 3 and the upcoming Trainium 4, alongside the Nova model series [1][12][18] - The full-stack AI strategy of AWS encompasses chips, models, frameworks, services, and applications, indicating a significant competitive push against Microsoft and Google [1][12][18] - Trainium 3 has been optimized for matrix computation, memory bandwidth, and interconnect topology, allowing for larger training workloads at reduced costs, reportedly saving up to 50% compared to mainstream GPU solutions [2][13][5] - The integration of Nvidia NVLink Fusion technology in Trainium 4 reflects AWS's strategy to reduce reliance on a single GPU vendor while collaborating with Nvidia for performance and ecosystem compatibility [2][14][18] - The Nova 2 series includes various models targeting different applications, such as Nova 2 Lite for cost efficiency, Nova 2 Pro for advanced reasoning, and Nova 2 Sonic for real-time voice interactions [3][20][18] - AWS's shift from GPU dependency to a full-stack AI platform is evident, as it aims to embed AI deeper into enterprise workflows through its AI Factories and proprietary models [3][16][18] Summary by Sections Chip Development - AWS launched Trainium 3 and is developing Trainium 4, which will incorporate Nvidia NVLink Fusion technology, enhancing its competitive edge in AI training [1][2][14] Model and Service Offerings - The introduction of the Nova 2 series aims to provide a comprehensive technology stack for enterprise applications, with models designed for various use cases [3][20][18] Competitive Landscape - AWS is positioning itself to compete more aggressively with Microsoft and Google, which have established strong narratives in the AI space through partnerships and proprietary technologies [3][16][18]

近日，京东云JoyBuilder模型开发平台迎来全新升级，成功支撑业界顶尖模型GR00T N1.5的千卡训练， 成为行业首家支持具身智能千卡级LeRobot开源训练框架的AI开发平台，且训练效率较开源社区版本提 升3.5倍。基于软硬件深度调优和算法层面的突破，大幅提升了模型训练效率与稳定性，1亿+数据的千 卡训练时间从15小时缩短至22分钟，加速具身智能迈向规模化落地。 1、JoyBuilder三大优势，全面加速具身智能模型训练 通过分布式元数据管理与智能预取，支持高并发访问。在1024卡集群上，读取带宽超过400GB/s，保障 数据持续高速供给。 其次，作为行业首家支持具身模型千卡级的开源训练框架的AI开发平台，京东云JoyBuilder基于开源 LeRobot框架，首次实现GR00T模型的千卡分布式高效训练，训练效率较开源社区版本提升3.5倍，1亿 +数据千卡训练时间从15小时缩短至22分钟、提速40倍，大幅降低了千卡训练的使用门槛与运行成本。 近年来，具身智能成为AI领域的重要研究方向，但训练一个高效的具身模型，需要同步处理图像、文 本和连续动作信号等众多环节，在工程落地方面存在着数据供给读取慢、算力效 ...