The Zhenwu V900 AI chip delivers 3 times the performance of its predecessor and supports clusters of up to 500,000 cards for training and inference. Unveiled by Alibaba at the Apsara Conference, the accelerator includes 216 GB of memory and 1,200 GB/s inter-chip bandwidth, with mass production scheduled for the first quarter of 2027. This hardware will power future Qwen model generations targeting between 5 trillion and 10 trillion parameters, compared to 2.4 trillion for the current Qwen 3.8 Max.

Alibaba Cloud aims to exceed 20 GW of global data center capacity by 2032, supported by a commitment of more than $53 billion over three years for AI infrastructure. Chief Executive Eddie Wu envisions machine-generated thinking reaching 1,000 times human capacity, stating that AI thinking currently comprises less than 3% of total human cognitive volume. The roadmap includes the launch of Yitian 720 and 730 CPUs for agentic AI workloads in 2027, while the Qwen 4 model family is currently in training.