Alibaba Unveils New AI Chip for Training and Inferencing
Alibaba’s chipmaking unit T-Head introduced the Zhenwu M890 AI accelerator at an event in Hangzhou, designed for both training and inferencing tasks with 144GB of GPU memory. The company announced plans to release an upgraded Zhenwu chip annually, matching the pace of industry leader Nvidia.
Alibaba Group Holding Ltd. expanded its AI technology lineup with the launch of the Zhenwu M890, an AI accelerator developed by its chipmaking division, T-Head. The chip was unveiled during an event in Hangzhou, China, on Wednesday, featuring 144GB of GPU memory to support both AI model training and inferencing tasks. Alibaba highlighted its suitability for agentic AI applications, emphasizing performance and efficiency. The Zhenwu M890 marks a strategic move by Alibaba to strengthen its AI infrastructure, competing directly with industry leader Nvidia. The company announced plans to accelerate chip upgrades, releasing a new Zhenwu model annually to keep pace with advancements in the sector. This aligns with Alibaba’s broader goal of covering every stage of AI development, from hardware to software solutions. The chip’s design focuses on high-memory capacity, enabling complex AI workloads while optimizing power consumption. Alibaba’s push reflects growing competition in the AI chip market, where demand for specialized hardware is rising. The Zhenwu M890 is expected to support large-scale AI deployments, including cloud services and enterprise applications. T-Head’s innovation comes as tech giants increasingly invest in custom AI chips to reduce reliance on external suppliers. By upgrading its Zhenwu series yearly, Alibaba aims to ensure its hardware remains competitive and adaptable to evolving AI needs. The company’s announcement underscores its commitment to becoming a key player in the global AI ecosystem.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.