Alibaba has unveiled a new AI accelerator while setting out a much larger target for its Qwen model family, tying the two announcements to a broader expansion of its cloud and data-center business.
Reuters reported that the company introduced the Zhenwu V900 through its T-Head semiconductor unit at Alibaba Cloud’s Apsara conference in Hangzhou. Chief Executive Eddie Wu said the accelerator delivers three times the performance of the previous M890 chip and can be connected in clusters of up to 500,000 units for training and serving very large models. Alibaba said it expects mass production and commercial release in the first quarter of 2027.
The chip announcement arrived alongside an ambitious model roadmap. Wu said the Qwen team plans to train a model with 5 trillion to 10 trillion parameters for more complex, longer-horizon tasks. Reuters noted that Alibaba’s current Qwen 3.8 Max has 2.4 trillion parameters. Parameter counts are a rough measure of a model’s scale, not a standalone measure of usefulness or quality.
A full-stack AI bet
Alibaba is presenting the work as a connected effort rather than a single product launch. The company also set a goal for Alibaba Cloud’s global data-center capacity to exceed 20 gigawatts by 2032 and said it would begin bringing AI supernodes online at commercial scale this quarter. That buildout matters because training and operating larger models requires both accelerators and the power, networking and facilities around them.
The claims about the V900’s performance and the eventual model scale come from Alibaba and Wu. They show the company trying to build more of the AI stack domestically as U.S. export restrictions continue to limit Chinese access to some Nvidia processors; they do not yet establish how the new chip or future Qwen models will compare in independent tests.
