Alibaba unveils Zhenwu V900 chip and larger Qwen model plans as China’s AI race intensifies

Alibaba used its annual flagship conference in Hangzhou to send a clear message about the next phase of China’s artificial intelligence race: the company wants to compete not only with models, but with the chips and cloud infrastructure needed to run them. AP reported that Alibaba introduced the Zhenwu V900, a new AI chip that CEO Eddie Wu described as China’s most powerful AI chip today, and said it can deliver three times the performance of the company’s previous Zhenwu M890 generation.

The announcement matters because AI competition is increasingly shaped by access to compute. Frontier models require specialized chips for training, then more compute for inference once they are used by customers. Alibaba’s Zhenwu chips are used in its own data centers and by cloud clients, giving the company a vertically integrated path from hardware to model development to hosted services. That structure resembles the playbook used by U.S. technology giants, where model progress is tied closely to proprietary infrastructure, large cloud contracts and steady access to advanced semiconductors.

Alibaba also laid out ambitious plans for the next generation of its Qwen model family. The company said it intends to train a model at the scale of five to 10 trillion parameters, a measure often used to describe the learning capacity of large AI systems. Its current Qwen3.8-Max model has 2.4 trillion parameters, according to the AP report. Parameter count alone does not determine quality, but the jump signals that Alibaba is still pursuing scale at a time when major labs are debating whether bigger models, better data, stronger reasoning systems or more efficient inference will define the next frontier.

The timing adds geopolitical weight. Alibaba’s announcement came days before a planned meeting between U.S. President Donald Trump and Chinese leader Xi Jinping, with AI competition expected to be one of the major background issues. Washington has tried to limit China’s access to the most advanced AI semiconductors, while Chinese companies have accelerated domestic chip work and model development. Alibaba’s chip update shows that the constraint has not removed China from the race; instead, it has made local compute capacity a strategic priority.

For cloud customers, the immediate question is not whether Alibaba has matched the strongest U.S. systems, but whether it can offer enough performance, availability and cost control for real deployments. Enterprises choosing AI infrastructure care about model quality, data rules, latency, reliability and long-term vendor support. A stronger internal chip roadmap could help Alibaba reduce dependence on outside suppliers and make its cloud services more predictable, especially for Chinese customers that need domestic infrastructure.

The broader industry signal is that AI is becoming a full-stack contest. Companies that want to lead cannot rely only on a popular chatbot or an open model release. They need chips, data centers, developer platforms, enterprise distribution and enough capital to keep scaling. Alibaba’s Zhenwu and Qwen announcements show how quickly that contest is hardening in China, and how closely model ambition is now tied to the physical infrastructure beneath it. The next test will be whether those claims translate into developer adoption and durable cloud revenue.