Core Insights - The article emphasizes the successful path taken by JD Cloud in the context of domestic chip and software development, highlighting the shift from policy-driven to market-driven approaches by 2025 [1][14] - JD Cloud's methodology of "gradual true replacement" is gaining traction as a practical model for domestic innovation, focusing on real business scenarios to validate technology [4][6] Group 1: Domestic Chip and Software Development - By 2025, domestic chips are expected to account for 40% of the market, with a notable increase in the performance of domestic software stocks [1] - The AI industry in China is transitioning from "usable" to "usable and effective," with AI workloads projected to dominate cloud computing by 2029 [1][10] Group 2: Challenges in Domesticization - The main challenges in domesticization are categorized into three areas: usability, controllability, and trustworthiness [2][3] - Usability issues arise from the need for stability in high-demand environments, particularly with the coexistence of X86 and ARM architectures [2] - Controllability concerns involve hidden costs related to software restructuring and personnel training, which can delay domesticization efforts [3] - Trustworthiness is critical, as businesses cannot afford the risks associated with complete system overhauls [3] Group 3: JD Cloud's Approach - JD Cloud's strategy includes multi-chip management to ensure system availability despite individual chip failures [5] - The gradual replacement strategy allows for controlled risk and cost management by starting with small-scale pilot projects [5] - JD Cloud leverages real business scenarios to refine its hardware and software solutions, ensuring continuous upgrades [5][9] Group 4: Technological Advancements - JD Cloud's JoyScale AI computing platform integrates various domestic chips, providing efficient computing solutions validated through extensive real-world testing [12] - The JoyBuilder model development platform enhances training and inference speeds while significantly reducing application costs [13][14] - Data security is prioritized through the use of national encryption standards and secure sandbox technologies, ensuring compliance and safety [14] Group 5: Market Perception and Future Outlook - The market's perception of domesticization is shifting from mere availability to the effectiveness and value of solutions [14] - By 2025, the focus will transition from policy-driven initiatives to commercially viable solutions that withstand extreme testing scenarios [14] - Companies like JD Cloud, which continuously enhance their capabilities, are expected to thrive in the competitive landscape of digital sovereignty [14]

京东云这条路走对了。 2025年，国产化已从政策议题演变为资本市场的核心主线。TrendForce数据显示，2025年国产芯片预期占比将提升至40%，加之近来国产软件板块持续发力，相关概 念股逆势表现活跃。在"安全可控"成为数字基建底层逻辑的背景下，市场不再满足于概念炒作，而是聚焦于真正具备落地能力的技术路径与商业闭环。 当大模型迈入全栈国产化时代，中国AI产业正从"能用"向落地深度应用要"好用"跃迁。Gartner预测到2029年，AI工作负载将占据云计算的50%，这意味着未来几年 AI将成为云技术发展的核心驱动力。数智化水平的提升，不仅倒逼算力基础设施升级，更对系统的可用性、可控性和可信度提出前所未有的挑战。 在此过程中，京东云以"渐进式真替真用"为方法论，依托万亿级真实业务场景打磨技术底座，其路径选择正引发行业广泛关注。 国产化三大关卡：可用、可控、可信 多位信创领域专家指出，当前国产化推进的最大瓶颈并非技术空白，而是生态割裂与迁移成本过高。"很多企业愿意支持国产，但不敢'一步到位'。"一位行业人士 坦言，"一旦核心系统停摆几小时，损失可能远超替代收益。" 三是"可信"鸿沟。企业无法承受"推倒重来"的风 ...

Core Insights - Gartner identifies AI supercomputing platforms as the top strategic technology trend for 2026, integrating various computing paradigms to manage complex workloads [1] - By 2028, over 40% of leading enterprises are expected to adopt hybrid computing paradigms, a significant increase from the current 8% [1] - The shift towards hybrid computing is driven by the need to mitigate risks associated with reliance on single types of chips, making it a necessity for AI strategies [1] Industry Challenges - The transition in computing architecture is necessary as AI's role grows, requiring a shift from CPU-centric to GPU-centric infrastructures [2] - Utilization rates of computing resources in unoptimized hybrid environments are generally below 40%, leading to inefficiencies [2] - The complexity of hybrid architectures increases exponentially from experimental to production systems, often exceeding the capabilities of most technical teams [2] Company Solutions - JoyScale AI computing platform aims to address the challenges of hybrid computing by providing a comprehensive service model rather than merely aggregating resources [3] - JoyScale enhances AI task deployment density and overall resource utilization by 70% through intelligent scheduling and compatibility with various domestic computing resources [5] - The platform is designed to meet stringent security and compliance requirements, ensuring data safety and performance stability [6] Market Adoption - As AI applications mature, competition in computing infrastructure is shifting from scale to systematic competition, with a focus on resource utilization and cost [7] - JoyScale has been adopted by numerous leading state-owned enterprises, demonstrating its effectiveness in pooling dispersed GPU resources and significantly improving utilization rates [7] - The platform supports a comprehensive product matrix for large model deployment, facilitating rapid implementation of AI applications in complex scenarios [7]

Group 1 - The 2025 ITValue Summit focused on the theme "The Truth of AI Scene Implementation," addressing ten core issues in AI application for enterprises, including strategy, reliability, data challenges, scenario selection, model selection, industry implementation, knowledge base construction, security compliance, human-machine collaboration, and talent bottlenecks [1] - During the summit, five closed-door meetings were held covering various topics and industries, allowing participants to discuss specific industry challenges in depth [1] Group 2 - Many small and medium-sized manufacturing enterprises face challenges in digital transformation, with 90% of their data remaining "asleep" due to a lack of unified data and business process standards [2][3] - The digitalization of supply chains is evolving from merely moving procurement online to achieving end-to-end collaboration and optimization through data integration [2] Group 3 - Companies like Shenzhen Genesis Machinery are integrating AI large model technology to break down data silos and enhance data sharing and value release [3] - The lack of standardization in business and data processes is a fundamental issue, particularly in non-standard manufacturing, where unique project characteristics complicate data integration [3] Group 4 - AI and data technologies are increasingly being applied to enhance supply chain transparency, responsiveness, and risk management [5] - Companies are utilizing AI to analyze historical sales and inventory data to predict risks, such as chip price increases, allowing proactive inventory management [6] Group 5 - The manufacturing sector's AI application differs significantly from the internet industry, focusing on "small data" and "scenario closure" rather than large models [6][7] - The core of successful digital transformation in manufacturing lies in standardization, followed by system implementation, data collection, and AI modeling [4] Group 6 - The financial sector is exploring AI infrastructure to address industry pain points, with companies like JD Cloud leveraging their diverse data advantages to enhance AI model training and application [10] - The successful application of AI in enterprises hinges on data quality, identifying suitable business scenarios, and establishing a supportive organizational structure [11][12] Group 7 - The retail industry is undergoing significant changes, with CIOs emphasizing the need to adapt to evolving consumer behaviors and market trends [19][20] - Successful retail operations require a focus on creating value for consumers and leveraging technology to enhance customer engagement [21] Group 8 - The hospitality and airline industries are integrating AI into their operations, with companies like East China Airlines deploying AI applications to improve efficiency and customer service [22][24] - The transition to AI-driven solutions in these sectors involves overcoming initial high costs and ensuring leadership commitment to AI initiatives [23][24] Group 9 - The CIOxCFO closed-door meetings highlighted the importance of collaboration between IT and finance leaders in driving AI implementation [25][26] - Key factors for successful AI application in enterprises include high-quality data accumulation, focusing on high-value business scenarios, and continuous operational improvement [27][30]

Core Insights - JD Cloud's JoyScale AI computing platform has been upgraded to support the most diverse domestic heterogeneous computing power scheduling in the industry, accommodating over 10 domestic AI computing cards and more than 20 training and inference frameworks, making it the only platform that supports remote invocation of both NVIDIA GPUs and Ascend NPUs [1][3]. Group 1: AI Native Computing Platform - The deep application of AI has created new technical challenges for infrastructure, necessitating an AI Native computing platform that is GPU-centric rather than CPU-centric [2]. - The increasing demand for inference and the need for efficient resource allocation are driving enterprises to rethink their investment in computing resources [2]. Group 2: JoyScale AI Platform Features - JoyScale AI platform is based on JD's internal unified GPU pooling practices, allowing for unified scheduling and resource sharing for training tasks and inference services [3]. - The platform boasts four core advantages: - Extreme computing performance with a 50% improvement in overall inference performance [5]. - Efficient heterogeneous computing power scheduling, achieving a 70% increase in overall resource utilization [5]. - Deep collaboration with domestic AI chip manufacturers to enhance the ecosystem [6]. - Support for over 20 AI training and inference frameworks, including PyTorch, TensorFlow, and Triton [7][8]. Group 3: Performance Optimization - JoyScale has overcome technical challenges for running mainstream models on domestic cards, achieving application and computing power separation for flexible resource allocation [9]. - The platform employs advanced scheduling algorithms to maximize task execution efficiency by optimizing resource allocation based on CPU NUMA and network topology [9]. - In model optimization, techniques such as GE graph compilation and ATB high-performance operator technology have been implemented to enhance inference speed in multi-modal scenarios [10].

Group 1 - The core viewpoint of the article highlights that JD Cloud ranks among the top five in the "2024 China Government Cloud Market Vendor Competitiveness Quadrant Analysis" due to its technological innovations and practical achievements in the government cloud sector [1][3] - The report indicates that the Chinese government cloud market is expected to maintain rapid growth, with a market size of 1214.8 billion yuan, reflecting a year-on-year increase of 12.6%. It is projected to reach 1689.9 billion yuan by 2027 [3] - JD Cloud's JDStack proprietary cloud platform is designed to meet the digital government needs, featuring flexible heterogeneous computing power management, comprehensive domestic adaptation, and large-scale scenario validation for enhanced security and stability [3][4] Group 2 - JD Cloud's JoyScale AI computing platform supports fine management of AI computing clusters, enabling intelligent scheduling of heterogeneous computing power to facilitate rapid local deployment of DeepSeek in government cloud applications [4][5] - JDStack achieves unified management across multiple clouds, chips, and active resources, supporting over ten million core resources with second-level scheduling, and is compatible with mainstream chips like X86, Kunpeng, and Feiteng [5] - JDStack has high-level security and stability guarantees, certified by various security standards, and has successfully supported high-concurrency events like JD's 618 and 11.11 sales, ensuring effective support for the digital transformation and innovation needs of government and enterprise sectors [5]

Core Insights - JD Cloud launched nine products including the JoyScale AI computing platform, JoyBuild large model development platform, and JoyAgent intelligent agent, aimed at helping enterprises reconstruct AI infrastructure and accelerate deep application adoption [1] - The company emphasized that the employment rate of digital employees will become a standard for measuring enterprise advancement, indicating that the extent of AI integration will determine future operational speed [1] - The new generation of agents, represented by JD Cloud's JoyAgent 2.0, is designed to assist enterprises in generating specialized digital employees, marking a significant step towards large-scale application and standardization of AI infrastructure [1]