补上全栈AI能力底层板块，阿里自研AI芯片“真武”亮相

Core Viewpoint - Alibaba's subsidiary, Pingtouge, has launched a high-end AI chip named "Zhenwu 810E," marking the official debut of its self-developed PPU chip, which is part of Alibaba's comprehensive AI strategy involving models, cloud, and chips [1][2] Group 1: Product Launch and Features - The "Zhenwu" PPU utilizes a self-developed parallel computing architecture and inter-chip interconnection technology, achieving full self-research in both hardware and software [2] - The chip features 96G HBM2e memory and an inter-chip bandwidth of 700GB/s, making it suitable for AI training, inference, and autonomous driving applications [2] Group 2: Market Position and Performance - The "Zhenwu" PPU's overall performance surpasses that of mainstream domestic GPUs, and it has been deployed in multiple large-scale clusters on Alibaba Cloud, serving over 400 clients including State Grid, Chinese Academy of Sciences, Xiaopeng Motors, and Sina Weibo [2] - The integration of hardware and software is expected to enhance computational efficiency and model training effectiveness, facilitating the application of large models across various industries [2] Group 3: Strategic Development - Alibaba aims to create an AI supercomputer by synergizing chip architecture, cloud platform architecture, and model architecture to achieve maximum efficiency in training and deploying large models [1] - Pingtouge, established in 2018, has released several chips since its inception, including AI inference chips and server chips, and has been focusing on integrated cloud-based chip development [1]