华为大动作！AI新技术

Core Insights - Huawei's Vice President Zhou Yuefeng announced the launch of AI container technology Flex:ai at the 2025 AI Container Application Implementation and Development Forum in Shanghai, aiming to address the challenges of computing resource utilization [1] - The technology is designed to package model code and runtime environments into lightweight images, facilitating seamless cross-platform migration and addressing deployment inconsistencies [1] - Gartner predicts that by 2027, over 75% of AI applications will be deployed using container technology, highlighting the growing importance of this technology in the AI industry [1] Group 1 - The Flex:ai technology allows for on-demand mounting of GPU and NPU resources, enhancing overall resource utilization in clusters [1] - The AI industry is experiencing rapid growth, leading to a significant demand for computing power, while global resource utilization remains low, resulting in substantial waste [1] - Issues such as small model tasks monopolizing resources and large model tasks lacking sufficient computing power contribute to the inefficiency in resource allocation [1] Group 2 - Flex:ai is built on the Kubernetes container orchestration platform, enabling precise management and intelligent scheduling of GPU and NPU resources [2] - The software aims to match AI workloads with computing resources effectively, significantly improving utilization rates [2] - Huawei plans to collaborate with academic institutions to continuously enhance the Flex:ai software, making AI technology more accessible to users and developers [2]