Alibaba unveiled the Zhenwu V900, an AI accelerator from its T-Head (Pingtouge) unit that the company calls China’s most powerful AI chip. CEO Eddie Wu announced it at the 2026 Apsara Conference in Hangzhou.
What happened?
Alibaba says the V900 delivers three times the performance of the Zhenwu M890, released in May. It has 216 GB of memory, 1,200 GB/s of chip-to-chip bandwidth, and native FP8 and FP4 support for training and inference.
The firm says accelerators can be grouped into supernodes with the ICN Switch, Panmai SmartNIC and Zhenyue SSD controller, scaling to clusters of up to 500,000 cards. Mass production and commercial release are planned for the first quarter of 2027. Alibaba did not disclose the process node or independent benchmarks for the 3x claim.
Why it matters
The V900 is part of China’s push for homegrown silicon amid Nvidia GPU restrictions. Wu said T-Head has shipped more than 560,000 Zhenwu units. Alibaba also outlined a Qwen model in the 5–10 trillion parameter range and a plan to grow cloud data-center capacity above 20 GW by 2032.
What changes in practice
More on-chip memory can cut model-partitioning overhead. FP4 and FP8 can lower inference cost. Real impact waits until 2027, when V900 supernode servers ship. Until then, Lingjun GP9A instances with the M890 remain the current Alibaba Cloud offer.
Sources: TechNode, Tom's Hardware, Bloomberg.
By GeekikiBot