Alibaba announced plans for a 5- to 10-trillion-parameter AI model and introduced a new chip designed to support China’s independent AI infrastructure.
Alibaba’s announcement
Alibaba announced plans to train a model with between 5 trillion and 10 trillion parameters and unveiled the Zhenwu V900 AI chip. The company said the model would target more complex and longer-horizon tasks as it develops its artificial-superintelligence strategy.
The V900 reportedly delivers three times the performance of its predecessor and can be linked in clusters of up to 500,000 chips.
Why model size matters
Parameter count is not a complete measure of model intelligence, but very large models require substantial computing resources and can support more complex training approaches.
A model at this scale would require:
– Large distributed-memory systems.
– High-speed chip interconnects.
– Advanced cooling.
– Large energy supplies.
– Fault-tolerant data centers.
– Sophisticated training software.
The announcement therefore concerns infrastructure as much as model design.
China’s full-stack strategy
Alibaba is building across several layers of the AI stack:
– Foundation models.
– Custom AI chips.
– Cloud infrastructure.
– Enterprise applications.
– Data-center capacity.
This approach is strategically important because China faces restrictions on access to some advanced foreign chips. Domestic hardware could reduce reliance on overseas suppliers and support continued AI development.
Commercial applications
Alibaba could deploy the technology across:
– E-commerce.
– Logistics.
– Cloud computing.
– Coding tools.
– Industrial automation.
– Financial services.
– Customer support.
The company’s existing business ecosystem gives it many internal use cases for large models and AI accelerators.
What to watch
The V900 is expected to enter mass production and commercial release in the first quarter of 2027. The important test will be whether Alibaba can deliver the promised performance at scale and operate the resulting models at commercially viable costs.
The September announcement reinforces a central industry trend: AI competition is increasingly about control of the entire technology stack, from chips and data centers to models and applications.
