Alibaba Introduces Its Own AI Chip for Neural Network Training and Inference

Alibaba Group has added a new chip to its growing AI stack, continuing its strategy of covering all aspects of artificial intelligence development.

About the New Processor

The company did not disclose the exact name, but indicated that the chip is intended for both training large language models and inference. This means the hardware is universal—it can be used in a data center for training and then repurposed for production tasks. The processor was developed with a focus on energy efficiency. Alibaba emphasized that the chip works optimally with its own cloud service Aliyun and integrates with the company's AI frameworks.

Why Alibaba Is Making Hardware

The strategy of the Chinese giant is clear: to control the entire AI development chain. Until recently, the company had to rely on Nvidia GPUs, but American chips are expensive and their export is limited by sanctions. Its own processor reduces this dependence.

• Reduction in operating costs by 15-20%
• Full control over the cloud stack
• Advantage over competitors with the same technology
• Fast optimization for Aliyun
• Backup strategy in case of further sanctions

Alibaba is already investing in other parts of the ecosystem—its own training framework, Qwen models, and hardware accelerators.

Competition Context

Competition in the field of proprietary AI chips is growing in the market. Google released TPUs, Meta is developing its own accelerators. Even startups like Groq are creating specialized processors. But chip development requires huge investments and experience in the semiconductor industry. Alibaba has both—its own experience in the cloud (Aliyun) and access to production supply chains through its contacts in China. This gives it an advantage over pure cloud companies in the West.

What This Means

The announcement shows that major cloud giants no longer want to depend on a single hardware supplier. Each is building its own stack—software, models, chips. For Alibaba, this is a step toward independence; for users, it's toward greater choice and possibly lower prices for AI infrastructure.

*Meta has been recognized as an extremist organization and is banned in the Russian Federation.