Core Insights - The 2025 China Computing Power Conference showcased significant breakthroughs in computing power, including the "Jiuzhou" computing power optical network developed by China Mobile, achieving 10 EFLOPS, capable of performing 100 trillion calculations in one second [1] - The conference highlighted ten major advancements in China's computing industry, focusing on both core underlying technologies and ecological solutions that enhance efficiency and sustainability [1][2] - These advancements are expected to drive GDP growth by over 12.6 billion yuan, indicating a strong economic impact from the computing power sector [1] Group 1: Major Breakthroughs - China Mobile's "Jiuzhou" optical network is the world's largest 400G all-optical inter-provincial backbone network [1] - Shanxi Qineng's integrated platform for computing and electricity saves 10 million yuan in electricity costs annually and reduces carbon emissions by 100,000 tons [1] - Super Fusion's FusionOne AI solution enables rapid deployment of AI applications, with over 500 projects already implemented [1] Group 2: Collaborative Technologies - China Telecom's "Wide-area Intelligent Computing Lossless Networking Technology" allows efficient collaboration between distant data centers [2] - China Unicom's "Distributed Training and Inference Key Technology" ensures efficient and secure AI training while protecting data privacy [2] - Lenovo's "Comprehensive Large Model Training and Inference Solution" enhances model inference performance by ten times and enables rapid fault recovery [2] Group 3: Ecosystem Development - Alibaba Cloud's heterogeneous GPU cloud platform facilitates collaboration among different brands of domestic GPUs, supporting over 100 major projects [2] - ZTE's intelligent computing supernode system enables large-scale high-speed interconnection of domestic GPU cards, supporting training of models with over one trillion parameters [2] - Shanghai AI Laboratory's DeepLink solution allows large-scale mixed training across provinces, optimizing national computing resources [2][3]