Zhong Guo Jing Ying Bao·2026-01-29 03:34Core Viewpoint - Alibaba's self-developed high-end AI chip "Zhenwu 810E" has been officially launched, marking a significant step in the establishment of the "Tongyun Ge" AI supercomputing framework, which integrates Alibaba Cloud, Pingtouge, and Tongyi Lab [1][2]. Group 1: Product Launch and Features - The "Zhenwu" PPU chip 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 [1]. - The "Zhenwu" PPU features a self-developed parallel computing architecture and inter-chip network technology, with a memory capacity of 96GB HBM2e and an inter-chip bandwidth of 700 GB/s, suitable for AI training, inference, and autonomous driving [2]. - The performance of "Zhenwu" PPU surpasses NVIDIA's A800 and mainstream domestic GPUs, and is comparable to NVIDIA's H20, with an upgraded version reportedly outperforming NVIDIA's A100 [2]. Group 2: Market Impact and Strategic Positioning - Alibaba is recognized as one of the only two global tech companies, alongside Google, that possesses top-tier capabilities across large models, cloud services, and chip development [2]. - The launch of the "Zhenwu" PPU and the positive market reception have contributed to a significant increase in Alibaba's market value, with a reported rise of approximately HKD 100 billion in just two days following the news of Pingtouge's potential independent listing [3]. - The Tongyun Ge framework aims to achieve maximum efficiency for users training and deploying large models on Alibaba Cloud, showcasing a comprehensive AI stack developed over 17 years of strategic investment and vertical integration [1].