- The new chip triples the performance of Alibaba’s previous generation and packs 216GB of memory
- Mass production is planned for early 2027 as Alibaba expands its homegrown AI infrastructure
Alibaba’s chip unit T-Head has unveiled a new AI processor that it says delivers three times the performance of its previous generation, as the Chinese technology giant expands its homegrown computing infrastructure for AI.
T-Head introduced the Zhenwu V900 on September 22 at the 2026 Apsara Conference in Hangzhou, calling it China’s most powerful domestically developed training and inference AI chip by computing performance.
The V900 delivers three times the performance of the Zhenwu M890, launched in May, with 216GB of memory and inter-chip bandwidth of up to 1,200GB per second, according to Alibaba.
As many as 500,000 chips
It supports FP8 and FP4, lower-precision formats that allow AI chips to process data faster while using less memory, and is designed to train and run models with trillions of parameters.
A single AI cluster using the V900 can scale to as many as 500,000 chips. T-Head’s ICN Switch technology links large numbers of chips so they can work together more efficiently.
Alibaba also unveiled a new Panjiu (磐久) AI server that combines the V900 with networking and storage chips, allowing the different parts of an AI system to work together as one.
Mass production in 2027
The V900 is scheduled to enter mass production in the first quarter of 2027 and will be deployed across Alibaba Cloud data centers.
Alibaba CEO Eddie Wu said the growing adoption of T-Head’s chips by customers should help drive a significant increase in annual shipments.
The V900 follows the M890, which launched in May with 144GB of memory and three times the performance of the previous Zhenwu 810E.
M890-based systems have been deployed at scale and can run AI models including Qwen3.8 and Kimi K3 with more than 2 trillion parameters, Alibaba said.
Another piece of the AI stack
As of June, the Zhenwu chip family had served more than 650 corporate customers in areas including autonomous driving, finance, AI models, embodied AI, energy and manufacturing.
The V900 gives Alibaba another piece of the AI infrastructure stack, from chips and servers to networking, storage and cloud computing, as demand for AI capacity continues to grow.
Header image credit: Alibaba T-Head
