Catch up on the essentials
- Alibaba unveiled the Zhenwu V900, a new AI training and inference processor developed by its T-Head semiconductor unit, at the Apsara Conference on Sept.
- The V900 carries 216 GB of memory and 1,200 GB/s of inter-chip bandwidth, with native FP8 and FP4 support for low-precision AI workloads.
- Alibaba has scheduled mass production and commercial release of the Zhenwu V900 for the first quarter of 2027, with Qwen 4 expected to emerge from training in the interim and Qwen 4.5 and Qwen 5 projected to follow on the multi-trillion-parameter roadmap.
Selected from this article · 2026-09-24
Read on for the full pictureZhenwu V900 debuts as Alibaba's most powerful AI accelerator

Alibaba unveiled the Zhenwu V900, a new AI training and inference processor developed by its T-Head semiconductor unit, at the Apsara Conference on Sept. 22. CEO Eddie Wu described it as "the most powerful AI chip in China," according to coverage of the announcement. Alibaba said the V900 delivers more than three times the performance of the Zhenwu M890, released in May 2026, and that the figure is measured against its own predecessor rather than a competing accelerator.
Hardware specifications and cluster scale
The V900 carries 216 GB of memory and 1,200 GB/s of inter-chip bandwidth, with native FP8 and FP4 support for low-precision AI workloads. Alibaba paired the chip with a next-generation supernode server integrating the V900 with its ICN Switch, Panmai SmartNIC and Zhenyue SSD controller, designed to scale into clusters of up to 500,000 accelerators. Mass production and commercial availability are scheduled for the first quarter of 2027, accelerating the timeline from a prior third-quarter-2027 target disclosed in May.
Qwen roadmap targets 5 to 10 trillion parameters
Alibaba confirmed that Qwen 4 is currently in training and outlined a roadmap for Qwen 4.5 and Qwen 5 to scale to between 5 trillion and 10 trillion parameters. The figures would dwarf the current Qwen3.8-Max, cited at 2.4 trillion parameters. Alibaba said it has already shipped more than 560,000 processors from the broader Zhenwu family and has begun using its own chips for AI workloads, though previous reporting has indicated Nvidia hardware remains important for its most demanding frontier-model development.
Domestic substitution drive and unanswered benchmarks
The V900 launch lands amid U.S. export restrictions that have constrained Chinese access to Nvidia's most advanced processors, raising the strategic value of domestic accelerators from Alibaba, Huawei and Baidu. Alibaba has not published independent benchmarks demonstrating that the V900 exceeds every other Chinese accelerator, leaving its "most powerful" designation a company claim. Pricing, software support and customer access details remain unresolved ahead of the planned 2027 commercial release.
Next milestone
Alibaba has scheduled mass production and commercial release of the Zhenwu V900 for the first quarter of 2027, with Qwen 4 expected to emerge from training in the interim and Qwen 4.5 and Qwen 5 projected to follow on the multi-trillion-parameter roadmap.
Share this article







