Alibaba unveils Zhenwu M890 AI chip and teases 10-trillion-parameter model

Sep 22, 2026

At the Apsara Conference on September 22, 2026, Alibaba Group Holding unveiled its Zhenwu V900 AI chip, described as China's most powerful AI processor, and announced plans to train a next-generation Qwen model with up to 10 trillion parameters. Developed by T-Head, the V900 triples the performance of its predecessor to counter US export curbs on Nvidia chips. Alibaba also targeted expanding its data center capacity to 20 gigawatts by 2032, driving its shares up 5.1%.

Zhenwu V900 chip unveiling

▪Alibaba Group Holding unveiled the Zhenwu V900 artificial intelligence chip, developed by its T-Head semiconductor unit, at the Apsara Conference on September 22, 2026.
▪Alibaba Group Holding plans to begin mass production and commercial release of the Zhenwu V900 chip in the first quarter of 2027.
▪The Zhenwu V900 chip delivers three times the performance of its predecessor, the M890, and can be linked in clusters of up to 500,000 chips.
▪Alibaba Group Holding's semiconductor subsidiary T-Head previously launched the Zhenwu M890 chip in May 2026.

US export controls impact

▪Chinese technology companies are developing domestic alternatives to Nvidia's artificial intelligence processors due to tightening United States export controls.
▪Alibaba Cloud Chief Executive Officer Eddie Wu stated that self-developed chips improve operational efficiency and reduce supply-chain vulnerabilities.
▪Alibaba Cloud's expansion speed is limited by global supply chain constraints, even though customer demand for artificial intelligence remains exceptionally robust.

10-trillion-parameter model development

▪Alibaba Group Holding's current flagship artificial intelligence model, Qwen 3.8 Max, has 2.4 trillion parameters.
▪Alibaba Group Holding is training its Qwen 4 model, with future Qwen 4.5 and Qwen 5 models expected to scale up to 10 trillion parameters.
▪CEO Eddie Wu said Alibaba Group Holding's Qwen team plans to train a next-generation artificial intelligence model with 5 trillion to 10 trillion parameters to handle complex, long-horizon tasks
▪Alibaba Group Holding Chief Executive Officer Eddie Wu stated that Alibaba's Qwen team has enabled models to improve themselves by generating training data with limited human involvement
▪Alibaba Group Holding chairman Joe Tsai stated that the company is investing in building full-stack artificial intelligence to move from technological breakthrough to value creation.
▪Alibaba Group Holding Chief Executive Officer Eddie Wu stated that the truly groundbreaking products of the machine intelligence era have not yet arrived.

20GW data center expansion

▪Alibaba Cloud set a target to scale its global data center capacity to more than 20 gigawatts by 2032.
▪Alibaba Group Holding's Hong Kong-listed shares rose 5.1% on September 22, 2026, reaching their highest level in a month following its Zhenwu V900 chip and Qwen model announcements
▪Alibaba Cloud plans to begin bringing its artificial intelligence supernodes online at a commercial scale in the third quarter of 2026.

T-Head customer base growth

▪Approximately 60% of the chip production capacity of Alibaba Group Holding's semiconductor subsidiary T-Head serves external clients.
▪Alibaba Group Holding's semiconductor subsidiary T-Head has shipped more than 560,000 Zhenwu series chips to over 400 customers, including China Telecom and FAW Group.

Debatable claims

▪AI models can successfully achieve superintelligence by training on self-generated data
▪Alibaba's Zhenwu chips can successfully replace Nvidia processors in China