Mirror Review
September 22, 2026
Alibaba unveiled its new AI chip, Zhenwu V900, while announcing plans for future Qwen models with 5 trillion to 10 trillion parameters. The Alibaba new AI chip announcement also includes plans to expand Alibaba Cloud’s global data-center capacity to more than 20 gigawatts by 2032.
Alibaba CEO Eddie Wu announced the plans at the 2026 Apsara Conference in Hangzhou on September 22. The announcements cover three connected areas of Alibaba’s AI strategy: models, chips, and cloud infrastructure.
What Is Alibaba’s New Zhenwu V900 AI Chip?
The Zhenwu V900 is the latest AI processor developed by Alibaba T-Head, Alibaba’s semiconductor unit.
Alibaba said the V900 delivers three times the performance of its predecessor, the Zhenwu M890 chip. A V900-based cluster can scale to up to 500,000 cards for advanced AI model training, according to Alibaba.
Alibaba plans to begin mass production of the V900 in early 2027.
The launch expands the company’s portfolio of Alibaba AI chips as Alibaba builds more of the computing hardware needed for its growing AI services.
Why Is Alibaba Planning 5 to 10 Trillion-Parameter AI Models?
Alibaba also announced plans to train a new AI model with 5 trillion to 10 trillion parameters, a major increase from its current flagship model.
Alibaba’s Qwen3.8-Max, introduced in August, has 2.4 trillion parameters. Alibaba plans to scale future Qwen models to between 5 trillion and 10 trillion parameters, a major increase from Qwen3.8-Max’s 2.4 trillion parameters.
Parameters are variables that an AI model learns during training and are commonly used as a measure of model size. Alibaba says its larger models are intended to handle more complex and longer-horizon tasks and advance its research toward artificial superintelligence.
The planned Alibaba 10 trillion parameter AI model therefore represents a significant increase in scale compared with the company’s existing Qwen models.
Alibaba New AI Chip: How Does Zhenwu V900 Support AI Expansion?
Alibaba’s new AI chip is part of its broader effort to develop AI computing across its own hardware and cloud infrastructure. The Alibaba new AI chip, developed by Alibaba T-Head, is designed to support larger-scale AI workloads as the company expands its models, computing capacity, and cloud services.
Alibaba’s existing M890 AI supernode can already support inference for models with more than 2 trillion parameters, according to Eddie Wu. Alibaba is now developing the V900 for larger-scale AI workloads while continuing to expand its cloud infrastructure.
Alibaba has described this approach as a full-stack AI strategy covering silicon, infrastructure, platforms and applications. Its T-Head unit has also developed software infrastructure for Zhenwu chips, helping connect the company’s hardware with its broader AI ecosystem.
Why Is Alibaba Targeting More Than 20 GW of Data-Center Capacity?
Alibaba Cloud plans to operate more than 20 GW of global data-center capacity by 2032.
The target is important because larger AI models require substantial computing infrastructure for both training and inference. Alibaba is therefore expanding not only its AI chips but also the data centers, networking and computing capacity needed to operate large AI systems at scale.
The Alibaba Cloud 20 GW data-center expansion also fits the company’s wider investment in AI infrastructure as demand for cloud-based AI computing continues to grow.
What Happens Next for Alibaba New AI Chip?
The next major milestone for the Alibaba new AI chip strategy will be the planned mass production of the Zhenwu V900 in early 2027. Alibaba is also continuing its Qwen model roadmap while expanding cloud infrastructure to support increasingly large AI workloads.
Alibaba is also continuing work on Qwen 4 and preparing its longer-term Qwen 4.5 and Qwen 5 roadmap. At the same time, the company is targeting more than 20 GW of global data-center capacity by 2032.
The September 22 announcement therefore connects Alibaba AI chips, larger Qwen models and expanded cloud infrastructure in one AI development roadmap. The V900 is the hardware component of that plan, while the 5 trillion to 10 trillion-parameter model target points to the scale of AI computing Alibaba expects to support in the years ahead.
Gurushanth S Jatti
