Alibaba's T-Head division unveiled the Zhenwu V900 AI accelerator chip, supporting both high-precision training and low-precision inference to optimize power efficiency. The chip scales up to 500,000 units per cluster and already has 650+ customers across automotive and AI sectors. It targets large language models with 5–10 trillion parameters, positioning Alibaba as a competitive alternative to Nvidia in AI infrastructure for Southeast Asian enterprises.
← Back to all articles