Alibaba’s AI Bet Gets Bigger: 10 Trillion Parameters and a New Chip
Alibaba is expanding its artificial intelligence ambitions with plans for a new model containing between 5 trillion and 10 trillion parameters, while simultaneously developing new computing hardware designed to support increasingly demanding AI workloads.
The announcement illustrates how China's largest technology companies are trying to build more of the AI infrastructure they need themselves, from models and chips to cloud computing capacity.
A much larger AI model
Alibaba CEO Eddie Wu said the company plans to train a next-generation AI model with between 5 trillion and 10 trillion parameters.
That would represent a substantial increase from Alibaba's current flagship Qwen 3.8 Max model, which Reuters reported has around 2.4 trillion parameters.
Parameters are internal numerical values used by AI models to process information and generate outputs. A larger parameter count can provide additional capacity, but it does not automatically guarantee that a model will perform better.
Model architecture, training data, optimization techniques and computing resources also play major roles.
The chip is just as important
Alibaba's announcement was accompanied by the unveiling of the Zhenwu V900, an AI chip developed by its T-Head semiconductor division.
According to Alibaba, the new chip delivers roughly three times the performance of its predecessor.
The hardware is designed to work in large clusters, allowing the company to combine many processors when training or operating large AI models.
This matters because frontier AI models require enormous amounts of computing power. A company cannot realistically pursue increasingly large models without considering the availability, cost and efficiency of the chips and data centers needed to operate them.
China's push for an independent AI stack
The announcement comes as Chinese technology companies face restrictions on access to some advanced foreign semiconductor technology.
Those restrictions have encouraged Chinese companies to develop alternatives across several parts of the technology chain.
Alibaba's strategy is particularly broad. Rather than focusing only on an AI model, the company is investing in models, semiconductors and cloud infrastructure.
That approach could reduce dependence on external suppliers over time, although building competitive hardware at scale remains a major technical and financial challenge.
A massive data-center ambition
Alibaba has also set a target of expanding its global data-center capacity to more than 20 gigawatts by 2032.
The figure illustrates how closely AI development is becoming linked to physical infrastructure.
Training and operating large AI systems require huge amounts of electricity, cooling capacity, networking equipment and specialized computing hardware.
As models become larger and companies deploy AI to more users, the demand for data-center capacity is expected to remain an important part of the industry's expansion.
Why parameter counts can be misleading
The 10-trillion-parameter figure is attention-grabbing, but it should not be interpreted as proof that Alibaba's future model will outperform every competing system.
Parameter count is only one technical measurement.
Modern AI development increasingly focuses on efficiency, reasoning capabilities, training methods and the ability to complete complex tasks. A smaller model can sometimes outperform a larger one on particular tasks if it has better architecture or training.
Alibaba's announcement therefore demonstrates ambition and investment rather than a guaranteed future performance result.
The financial challenge
Building advanced AI infrastructure is expensive. Companies must spend on chips, data centers, electricity, engineers and model training long before the resulting systems generate enough revenue to cover those costs.
That creates a difficult balance for technology companies. Investing too little could leave them behind competitors, while investing too aggressively could pressure profits if AI revenue fails to grow quickly enough.
Alibaba's decision to expand simultaneously across hardware, models and cloud infrastructure indicates that the company is willing to make a long-term investment in the sector.
China's AI race is becoming an infrastructure race
The latest announcement also highlights a broader shift in artificial intelligence.
The competition is no longer limited to which company produces the most impressive chatbot. Increasingly, companies are competing over chips, electricity, data centers, cloud platforms and access to the supply chains required to build AI systems.
That makes AI development increasingly similar to an infrastructure industry as well as a software industry.
What happens next?
Alibaba's 5-to-10-trillion-parameter model remains a plan rather than a finished product. The company will still have to develop the architecture, train the system and demonstrate how it performs in real-world applications.
The Zhenwu V900 also has to move from announcement to production and large-scale deployment.
Those steps will determine whether the company's ambitious roadmap translates into practical AI capabilities.
For now, the announcement provides a clear signal about the direction of China's AI industry: bigger models, domestically developed hardware and massive investment in computing infrastructure are increasingly being pursued together.
The next question is not simply how large Alibaba's model becomes. It is whether the company can turn that scale into useful, reliable and commercially sustainable AI systems.