2025年算力调度平台行业：优化计算资源，支撑AI应用

Investment Rating - The report does not explicitly provide an investment rating for the computing power scheduling platform industry. Core Insights - The rapid development of artificial intelligence technology has led to an exponential increase in global demand for computing power, necessitating computing power scheduling for resource integration and optimization across regions and platforms [2]. Summary by Sections Overview of the Computing Power Scheduling Industry - Computing power is defined as the ability of computer devices or data centers to process information, categorized into general computing power, intelligent computing power, and supercomputing power [15][18]. - China's computing power scale has grown rapidly, reaching 280 EFLOPS by 2024, with intelligent computing power accounting for 32% [20][23]. Challenges in Heterogeneous Computing Power Scheduling - Heterogeneous computing power scheduling faces multiple core challenges, including increased scheduling complexity due to resource heterogeneity and software environment fragmentation, high migration costs for cross-architecture tasks, and a lack of unified scheduling standards leading to resource mismatch and low utilization [4][43]. Major Domestic Computing Power Scheduling Platforms - National-level computing power scheduling platforms are primarily government-led or constructed by major operators, emphasizing cross-regional collaboration and market-oriented transactions [5][48]. - Provincial platforms cover key regions like the Yangtze River Delta and Chengdu-Chongqing, while municipal platforms focus on local AI and smart manufacturing scenarios [48]. Mainstream Open Source Computing Power Scheduling Technologies - Domestic computing power scheduling platforms are often built on open-source technologies, with openFuyao emerging as a versatile scheduling platform with advantages in domestic adaptation, while Kubernetes and Slurm have strong foundations in cloud-native and HPC fields [6][51].