Alibaba Plans 10T-Parameter AI Model, Unveils New AI Chip

Alibaba is expanding its artificial intelligence ambitions with plans for a 10-trillion-parameter AI model and a newly unveiled AI chip, the Zhenwu V900. The announcement highlights China's growing focus on developing powerful AI models alongside the computing hardware needed to run them. If successfully developed and deployed, these technologies could influence the global AI race, AI infrastructure, and the future of artificial intelligence.

What Did Alibaba Announce About Its New AI Model and Chip?

At the 2026 Apsara Conference in Hangzhou, Alibaba CEO Eddie Wu announced plans to train a new AI model with 5 trillion to 10 trillion parameters. The company also unveiled the Zhenwu V900, a new AI chip developed by its semiconductor unit, T-Head.

Alibaba says the V900 delivers three times the performance of its predecessor. The company is also targeting more than 20 gigawatts of global data centre capacity by 2032.

These announcements form part of Alibaba's broader strategy to develop AI models, computing chips, and cloud infrastructure together. However, the proposed model is still a plan—not a publicly demonstrated, completed 10-trillion-parameter system.

1. Alibaba's 10 Trillion Parameter AI Model: What's the Plan?

Alibaba is planning a major expansion of its Qwen AI model family. Its proposed model is expected to contain between 5 trillion and 10 trillion parameters, making it substantially larger than its current flagship Qwen 3.8 Max, which Reuters reports has 2.4 trillion parameters.

But what does this mean in practical terms?

What Are AI Model Parameters?

Parameters are numerical values that an AI model learns during training. They help the model recognise patterns, process information, and generate responses.

A simple way to understand them is to think of parameters as adjustable settings inside an AI system. During training, these values are refined to help the model perform tasks such as answering questions, writing content, analysing information, or generating code.

A 10T parameter AI model would have an enormous number of these learned values. However, more parameters do not automatically mean better AI. Architecture, training data, computational efficiency, and real-world performance also matter.

Why Is Alibaba Building a Larger Qwen AI Model?

Alibaba's long-term AI strategy includes developing models capable of handling increasingly complex tasks. Its CEO has also discussed recursive self-improvement, a research direction in which AI systems help identify limitations and contribute to improving future systems.

The company's proposed model is part of this broader ambition. Still, its actual capabilities, release date, and performance benchmarks have not yet been established publicly.

2. Alibaba AI Chip 2026: Introducing Zhenwu V900

One of the most important parts of Alibaba's announcement is the Zhenwu V900 AI chip.

Developed by T-Head, Alibaba's semiconductor division, the V900 is designed to support the computing demands of advanced AI systems. Alibaba says it delivers three times the performance of its predecessor, the M890.

Key Details About the Zhenwu V900

Feature	Details
Chip name	Zhenwu V900
Developer	T-Head, Alibaba's semiconductor unit
Reported performance	3× its predecessor, M890
Intended use	AI training and inference
Mass production target	Early 2027

AI training is the process through which a model learns from data. Inference is when a trained model generates an answer, prediction, or output.

Both require significant computing resources, especially when AI models become extremely large. Alibaba's focus on developing its own AI chip technology could help it better coordinate hardware and software for its AI workloads.

The V900 is not yet proof that Alibaba has achieved a particular level of AI capability. Its real-world performance will depend on factors such as software optimisation, power efficiency, and how it performs in deployed systems.

3. Why Alibaba Is Investing in AI Models, Chips, and Cloud Infrastructure

Alibaba's AI strategy is not limited to developing a single large model. The company is investing across three connected areas: AI models, AI chips, and cloud infrastructure.

AI Models

Qwen is Alibaba's AI model family. More capable models could support coding, research, reasoning, and AI-powered business applications.

AI Chips

Zhenwu processors are intended to provide computing power for training and running AI models, helping Alibaba develop its own hardware ecosystem.

Cloud Infrastructure

Alibaba aims to expand its global data centre capacity beyond 20 GW by 2032, supporting growing demand for AI computing.

Reuters

This integrated approach could give Alibaba greater control over how AI systems are developed, deployed, and delivered to customers through the cloud.

It also reflects a wider industry trend: AI companies increasingly need more than powerful models. They need the chips, networking, energy, and infrastructure required to operate them at scale.

4. How Could Alibaba's AI Breakthrough Impact the Global AI Industry?

Alibaba's announcement could have implications beyond China. Its impact, however, will depend on whether the company can turn its plans into competitive, commercially useful technology.

A. More Competition in the AI Model Market

The proposed 10T parameter AI model could increase competition among companies developing advanced AI systems. Alibaba's Qwen ecosystem already plays a role in the global AI landscape, and a new generation of models could expand its reach.

B. Greater Focus on Domestic AI Chips

Developing proprietary chips could help Alibaba reduce its dependence on external hardware suppliers. This is particularly relevant as Chinese technology companies work to strengthen domestic AI computing capabilities amid restrictions on access to advanced foreign chips.

C. More AI Computing for Businesses

If Alibaba successfully expands its AI infrastructure, businesses using Alibaba Cloud could eventually gain access to more AI computing resources.

Potential applications include AI agents, software development, enterprise automation, data analysis, and customer support. The actual benefits will depend on availability, pricing, performance, and service deployment.

5. Could Alibaba's 10T AI Model Bring Us Closer to AGI or ASI?

Artificial General Intelligence (AGI) refers to AI with broad capabilities across many different intellectual tasks, while Artificial Superintelligence (ASI) describes AI that surpasses human capabilities across a wide range of domains.

Alibaba's announcement is relevant to these discussions because the company is pursuing larger models, advanced computing infrastructure, and research into recursive self-improvement.

However, a 10-trillion-parameter model would not automatically qualify as AGI or ASI.

Parameter count alone cannot establish whether a system can reason reliably, adapt to unfamiliar situations, learn independently, or outperform humans across domains.

The real test will be the model's demonstrated capabilities, reliability, and performance on demanding real-world tasks.

1. What is Alibaba's 10 trillion parameter AI model?

Alibaba's proposed AI model is a planned Qwen-family system with 5 trillion to 10 trillion parameters. CEO Eddie Wu announced the plan at the September 22, 2026, Apsara Conference. The model is intended to support more advanced AI capabilities, but it has not been publicly demonstrated as a completed system.

2. What AI chip did Alibaba unveil in 2026?

Alibaba unveiled the Zhenwu V900 AI chip, developed by its T-Head semiconductor unit. The company says it delivers three times the performance of its predecessor, the M890, and targets mass production in early 2027.

3. What are AI model parameters?

AI model parameters are numerical values learned during training that help determine how an AI system processes inputs and generates outputs. A model with more parameters is generally larger, but parameter count alone does not determine intelligence or performance.

4. How could Alibaba's AI model impact the AI industry?

Alibaba's planned model could intensify competition in advanced AI, encourage investment in AI chips, and expand cloud-based AI services. Its actual industry impact will depend on model performance, infrastructure readiness, commercial availability, and adoption.

5. Is Alibaba's 10T AI model available yet?

No public release or completed model has been confirmed in the September 22, 2026, announcement. Alibaba has announced plans to train a model in the 5-trillion-to-10-trillion-parameter range, but a confirmed launch date has not been disclosed.

6. Could Alibaba's new AI technology bring us closer to AGI?

It could contribute to AI research and the development of more capable systems. However, neither a large parameter count nor advanced AI chips prove that AGI has been achieved. Demonstrated capabilities, adaptability, and reliable performance across tasks are more meaningful indicators.

Conclusion

Alibaba's plan for a 10T parameter AI model, the unveiling of its Zhenwu V900 chip, and its expanding cloud infrastructure highlight the company's long-term AI ambitions.

The announcement also reinforces an important shift in the AI industry: the race is no longer just about building smarter models. It is also about developing the computing power and infrastructure needed to run them efficiently.

For businesses, developers, and AI enthusiasts, Alibaba's next steps will be worth watching. The key question is not simply how large its next AI model will be, but what it can actually do.

As the future of artificial intelligence unfolds, real-world performance—not just parameter counts—will determine the significance of this development.