Alibaba’s T-Head Unveils Self-Developed AI Chip Zhenwu 810E
Alibaba Group’s semiconductor subsidiary, T-Head, has officially launched the Zhenwu 810E, a high-end artificial intelligence chip developed entirely using its in-house technology stack. The new hardware features 96GB of HBM2e memory and offers an inter-chip bandwidth of 700GB/s. Industry insiders indicate that its performance is comparable to NVIDIA’s H20, with certain upgraded versions potentially surpassing the capabilities of the A100 chip. The Zhenwu 810E has already been deployed in large-scale clusters comprising 10,000 cards on Alibaba Cloud, where it serves over 400 customers, including major entities like State Grid, the Chinese Academy of Sciences, and XPeng Motors. The chip is specifically optimized for the training and inference of Alibaba’s Qwen large language models. This launch signifies a strategic move by Alibaba to strengthen its full-stack capabilities across large models, cloud computing infrastructure, and proprietary semiconductor development, reducing reliance on external suppliers and enhancing its competitive position in the AI hardware market.
Wire timeline
Alibaba’s T-Head Unveils Self-Developed AI Chip Zhenwu 810E
Alibaba Group’s semiconductor subsidiary, T-Head, has officially launched the Zhenwu 810E, a high-end artificial intelligence chip developed entirely using its in-house technology stack. The new hardware features 96GB of HBM2e memory and offers an inter-chip bandwidth of 700GB/s. Industry insiders indicate that its performance is comparable to NVIDIA’s H20, with certain upgraded versions potentially surpassing the capabilities of the A100 chip. The Zhenwu 810E has already been deployed in large-scale clusters comprising 10,000 cards on Alibaba Cloud, where it serves over 400 customers, including major entities like State Grid, the Chinese Academy of Sciences, and XPeng Motors. The chip is specifically optimized for the training and inference of Alibaba’s Qwen large language models. This launch signifies a strategic move by Alibaba to strengthen its full-stack capabilities across large models, cloud computing infrastructure, and proprietary semiconductor development, reducing reliance on external suppliers and enhancing its competitive position in the AI hardware market.
TechNode