Core Insights - The report highlights the emergence of AI infrastructure software (AI Infra) as a critical enabler for the deployment of generative AI applications, marking a golden development period for infrastructure software [1] - Unlike the model training phase dominated by tech giants, the inference and application deployment stages present new commercial opportunities for independent software vendors [1] - Key products in this space include computing scheduling software and data-related software, with computing scheduling capabilities directly impacting the profitability of model inference services [1][2] Computing Scheduling - AI Infra is designed to efficiently manage and optimize AI workloads, focusing on large-scale training and inference tasks [2] - Cost control is crucial in the context of a price war among domestic models, with Deepseek V3 pricing significantly lower than overseas counterparts [5] - Major companies like Huawei and Alibaba have developed advanced computing scheduling platforms that enhance resource utilization and reduce GPU requirements significantly [5][6] - For instance, Huawei's Flex:ai improves utilization by 30%, while Alibaba's Aegaeon reduces GPU usage by 82% through token-level dynamic scheduling [5][6] Profitability Analysis - The report indicates that optimizing computing scheduling can serve as a hidden lever for improving gross margins, with a potential increase from 52% to 80% in gross margin by enhancing single-card throughput [6] - The sensitivity analysis shows that a 10% improvement in throughput can lead to a gross margin increase of 2-7 percentage points [6] Vector Databases - The rise of RAG (Retrieval-Augmented Generation) technology has made vector databases a necessity for enterprises, with Gartner predicting a 68% adoption rate by 2025 [10] - Vector databases are essential for supporting high-speed retrieval of massive datasets, which is critical for RAG applications [10] - The demand for vector databases is expected to surge, driven by a tenfold increase in token consumption from API integrations with large models [11] Database Landscape - The data architecture is shifting from "analysis-first" to "real-time operations + analysis collaboration," emphasizing the need for low-latency processing [12][15] - MongoDB is positioned well in the market due to its low entry barriers and adaptability to unstructured data, with significant revenue growth projected [16] - Snowflake and Databricks are expanding their offerings to include full-stack tools, with both companies reporting substantial revenue growth and customer retention rates [17] Storage Architecture - The transition to real-time AI inference is reshaping storage architecture, with a focus on reducing IO latency [18] - NVIDIA's SCADA solution demonstrates significant improvements in IO scheduling efficiency, highlighting the importance of storage performance in AI applications [18][19]

Core Insights - Nvidia has acquired SchedMD and will continue to distribute its open-source Slurm software, which is widely used in high-performance computing (HPC) and artificial intelligence (AI) environments [1][2][3] Group 1: Acquisition Details - The acquisition allows Nvidia to enhance Slurm's development, ensuring it remains the leading open-source scheduler for HPC and AI [3] - Nvidia plans to provide open-source software support, training, and development for Slurm to SchedMD's customers [3] - The collaboration between Nvidia and SchedMD has been ongoing for over a decade, indicating a strong partnership [2] Group 2: Strategic Implications - Nvidia aims to accelerate SchedMD's access to new systems, optimizing workloads across its accelerated computing platform [4] - The acquisition supports a diverse hardware and software ecosystem, enabling customers to run heterogeneous clusters with the latest Slurm innovations [4] - SchedMD's CEO emphasized the importance of Slurm in demanding HPC and AI environments, highlighting its critical role [4][5] Group 3: Future Developments - Nvidia's investment in Slurm is expected to meet the demands of the next generation of AI and supercomputing while maintaining its open-source nature [5] - In addition to acquiring SchedMD, Nvidia also finalized the acquisition of Run:ai, a Kubernetes-based workload management provider, to enhance AI computing resource efficiency [6] - Nvidia's CEO noted that the company is navigating three significant platform shifts, including the transition to accelerated computing and generative AI [7]

Investment Rating - The industry investment rating is "Outperform the Market" and is maintained [1] Core Insights - The report highlights the launch of Huawei's AI container technology Flex:ai, which addresses the low utilization efficiency of computing power in the industry, currently averaging only 30% to 40%. Flex:ai enhances utilization by 30% through precise segmentation of GPU/NPU resources [4][5] - The report emphasizes the unique advantages of Flex:ai over Nvidia's Run:ai, particularly in virtualization and intelligent scheduling, which can optimize resource allocation for AI workloads [5][6] - The development of Flex:ai is seen as a significant step in strengthening domestic computing power capabilities, promoting a complete open-source ecosystem for AI tools [6][7] Summary by Sections Industry Overview - The closing index is at 5068.36, with a 52-week high of 5841.52 and a low of 3963.29 [1] Performance Analysis - The relative performance of the computer industry compared to the CSI 300 index shows fluctuations, with a notable decline of 13% from November 2024 to November 2025 [3] Key Developments - Huawei's Flex:ai is positioned to significantly improve AI cluster computing efficiency and reduce migration barriers for AI models, reinforcing the software capabilities in the domestic computing landscape [6][7] - The report suggests monitoring companies involved in AI containers and domestic computing power, including BoRui Data, Haohan Deep, and others [7]

Core Insights - The rapid development of the AI industry is creating a massive demand for computing power, but the low utilization rate of global computing resources is becoming a significant bottleneck for industry growth [1] - Huawei's new AI container technology, Flex:ai, aims to address the issue of computing resource waste by allowing a single GPU/NPU card to be divided into multiple virtual computing units, improving resource utilization by 30% [1][2] - Flex:ai is positioned to compete with Nvidia's Run:ai, focusing on software innovation to unify management and scheduling of various computing resources without hardware limitations [2] Group 1 - Flex:ai technology can split a single GPU/NPU card into virtual computing units with a precision of 10%, enabling multiple AI workloads to run simultaneously [1] - The technology has been validated in real-world applications, such as the RuiPath model developed in collaboration with Ruijin Hospital, which improved resource utilization from 40% to 70% [3] - Gartner predicts that by 2027, over 75% of AI workloads will be deployed and run using container technology, indicating a shift towards more efficient resource management [3] Group 2 - Flex:ai will be open-sourced in the Magic Engine community, contributing to Huawei's comprehensive ModelEngine open-source ecosystem for AI training and deployment [3] - Unlike Run:ai, which primarily serves the Nvidia GPU ecosystem, Flex:ai supports a broader range of computing resources, including both Nvidia GPUs and Huawei's Ascend NPUs [2]

Core Insights - Nvidia is experiencing challenges in its enterprise software sales as it attempts to onboard large clients in regulated industries while maintaining its growth trajectory amid the AI boom [1][2] Group 1: Software Sales Challenges - Internal communications reveal that Nvidia's sales team is struggling to present a unified message regarding its software offerings alongside its AI hardware [2][7] - The company is focusing on selling Nvidia AI Enterprise (NVAIE) and other software products, but there is a need for a comprehensive narrative to effectively communicate these offerings to clients [4][6] - A July email indicated that stand-alone software sales are projected to exceed targets at 110%, while software sold with hardware is only expected to reach 39% of its goal [6] Group 2: Client Education and Legal Concerns - There is a significant disconnect between Nvidia and its clients' legal and procurement teams, particularly in understanding the software sales processes during negotiations [8][9] - The company is planning workshops to educate clients on NVAIE and other products, addressing the need for better internal and external education [7][8] - Data security and indemnity obligations are highlighted as major negotiation sticking points, with clients requesting higher damages caps than Nvidia is comfortable with [9] Group 3: Market Position and Future Outlook - Despite the challenges, Nvidia is forecasting strong software sales, with NVAIE expected to hit 186% of its sales target for the quarter [6][5] - The company’s software segment, while smaller, is crucial for generating recurring revenue and increasing customer dependence on its AI products [5]