来源：科技日报 科技日报记者 崔爽 12月25日，以"昇思MindSpore为超节点而生的AI框架"为主题的昇思人工智能框架峰会在浙江杭州召开。记者从会上获悉，昇思MindSpore AI框架下载量超 过1300万，覆盖全球156个国家和地区，社区贡献者超5.2万。昇思MindSpore聚焦超节点技术创新，打造昇思HyperParallel架构，加速新模型结构和新训推范 式创新，引领AI框架迈入"超节点时代"。 华为中央软件院总裁谢桂磊在致辞中表示，人工智能发展浪潮奔涌向前，昇思MindSpore致力于打造超节点亲和、全场景融合、架构开放、敏捷使能的人工 智能框架，以更前沿的技术、更易用的体验、更开放的社区，助力千行万业智能化转型。 对此，昇思MindSpore开源社区技术委员会主席金雪锋介绍，昇思HyperParallel架构把超节点看成一台"超级计算机"进行编程和调度，充分发挥超节点架构优 势，实现HyperShard声明式并行编程、HyperMPMD异构非规则并行、HyperOffload多级智能卸载等特性，加速Agentic AI（代理式人工智能）、全模态等新 模型结构和训推范式创新。 现场，8名开发 ...

Group 1 - The core viewpoint of the article is the launch of XinMM-AM1, the first Agentic AI model in the automotive finance industry by Yixin Group, which aims to enhance customer acquisition, risk control, and operational efficiency, potentially driving the industry into the next decade [1][2] - XinMM-AM1 has approximately 30 billion parameters, with a response latency of less than 200ms and a single card throughput of up to 370 tokens/s, facilitating low-cost large-scale deployment and business services [1] - The model serves as the "core brain" and orchestrator for automotive finance, enabling full-channel interaction, multi-modal perception, global collaboration, and comprehensive security compliance, thus automating decision-making processes [1] Group 2 - Yixin Group is an AI-driven fintech platform that became the first company in China's automotive finance sector to file a generative AI model in 2024 and has implemented large-scale AI applications across all business scenarios [2] - With the launch of the Agentic model, Yixin aims to accelerate AI empowerment in the industry and continue building a smart automotive finance ecosystem [2]

Core Viewpoint - Oracle's positive outlook on its cloud business has further fueled the booming global AI industry, leading to significant stock price increases for Oracle and other tech companies like Nvidia [2][3]. Cloud Market Growth - The global cloud computing market is projected to reach $692.9 billion in 2024, with a year-on-year growth of 20.2%, while China's cloud market is expected to grow to 828.8 billion yuan, a staggering 34.4% increase [3]. - By 2030, the global cloud market could approach $2 trillion, with China's market potentially exceeding 3 trillion yuan [3]. AI Infrastructure Investment - Major tech companies are heavily investing in data centers to support AI model training and inference, with Meta planning to invest at least $600 billion by 2028 [5]. - OpenAI has indicated plans to spend trillions on data center construction to meet increasing computational demands [5]. Oracle's Position in AI - Oracle's remaining performance obligations (RPO) surged to $455 billion, a 359% year-on-year increase, driven by large-scale cloud contracts with top AI companies like OpenAI and Meta [6]. - Oracle aims to adapt quickly to AI demands, leveraging its strong customer base and financial resources to provide reliable cloud infrastructure [6]. Competitive Landscape - The AI landscape is becoming more competitive, with traditional cloud providers facing challenges from new entrants focused on AI capabilities [6]. - Despite Oracle's significant contracts, concerns exist regarding its reliance on a few major clients and the long-term nature of its contracts, which may introduce uncertainty [7][8]. AI Demand Dynamics - The demand for AI infrastructure is expected to grow significantly, with inference needs outpacing training requirements, as highlighted by Oracle's CTO [11]. - The transition from training to inference represents a substantial opportunity for cloud service providers, as ongoing user interactions will require continuous computational power [12]. Chinese Cloud Service Opportunities - China's cloud service market is experiencing explosive growth, with Alibaba Cloud reporting a 26% revenue increase and significant investments in AI infrastructure [13]. - The AI cloud market in China is projected to reach 223 billion yuan by mid-2025, with a forecasted growth of 148% driven by generative AI [13]. Future Outlook - The AI infrastructure race is expected to gain momentum in 2025, with both domestic and international markets showing strong growth potential [16][17]. - The demand for AI-related services in China is anticipated to grow rapidly, although challenges such as supply chain uncertainties and intense competition may impact market growth [15].