Alibaba has announced a major expansion of its artificial-intelligence ambitions, unveiling the Zhenwu V900 chip and outlining plans to train a model with as many as 10 trillion parameters. The announcements were made September 22, 2026 at Alibaba's annual cloud conference in Hangzhou.
Alibaba says the Zhenwu V900 is its most powerful AI chip so far, delivering about three times the performance of its predecessor. The company is developing the hardware for the rapidly growing demand for AI training and inference, areas that require large amounts of specialized computing.
A bigger model target
Alibaba also said it plans to train a new AI model reaching up to 10 trillion parameters. Its current top model, Qwen3.8-Max, is reported at 2.4 trillion parameters, while another major Chinese AI lab, Moonshot, has a 2.8 trillion-parameter open model.
Parameter count is not a complete measure of an AI system's capability, but the scale of the announcement illustrates how quickly the industry's compute requirements are growing. Training larger models requires not only powerful processors but also huge data-center infrastructure and efficient networking.
Data centers are part of the story
Alibaba says it plans to expand its data-center capacity to more than 20 gigawatts by 2032. That ambition shows the company is thinking about AI as an entire infrastructure stack rather than simply a software product. More computing capacity can support model training, cloud customers, inference workloads and other AI services.
The company is making these moves while access to the most advanced foreign AI processors remains constrained by export controls. Chinese firms including Alibaba and Huawei have therefore been investing heavily in domestic chip design and software ecosystems.
What the chip means for users
Consumers are unlikely to encounter the Zhenwu V900 as a component inside a phone or laptop. It is aimed at the data-center side of AI, where processors are used to train and run large models. The more direct impact for everyday users would come later through AI services that rely on the underlying cloud infrastructure.
That could include AI assistants, coding tools, search systems, content-generation services and business applications. The performance and efficiency of the hardware can influence how much compute is available and how quickly large models respond.
The broader AI race
Alibaba's announcement comes during a period of intense competition over both AI models and the hardware needed to operate them. The company's strategy combines domestic chip development, increasingly large models and a long-term data-center expansion plan.
The next important milestones will be real-world benchmarks, availability of the V900 for customers and evidence of how the planned larger model performs. For now, Alibaba has clearly signaled that it intends to compete across the full AI stack — from silicon to cloud infrastructure to very large models.
Source
Associated Press — Alibaba AI chip and model announcement
