T-Head Group, a subsidiary of Alibaba, unveiled the open-source release of SAIL, the complete software package for its Zhenwu series AI chips, during the World Artificial Intelligence Conference (WAIC) in Shanghai.

This move aims to facilitate developers' transition from the Nvidia CUDA platform, which represents a major barrier to AI software development outside the Nvidia ecosystem.

SAIL allows programmers to adapt their projects to Zhenwu chips and run existing applications within popular frameworks such as PyTorch in less than a week, according to T-Head.

The majority of AI developers worldwide rely on CUDA to develop software for graphics processors, effectively locking them into the Nvidia ecosystem, which supports the company's market dominance and has boosted its market value to over $3.4 trillion.

China is pushing for self-reliance in AI by developing alternative software and infrastructure. Huawei previously open-sourced its CANN platform for Ascend processors in 2025, and Moore Threads is following the same approach with its own GPU software platform.

All these entities are competing to attract AI developers to Chinese solutions, as the biggest challenge currently lies in the widespread adoption of CUDA and the superiority of its software library, which has been developed over 17 years.

Huawei Unveils Atlas 850E: First AI Supernode with Air Cooling Technology

Alibaba's move comes at a sensitive time, as the company is involved in a legal dispute with the U.S. Department of Defense and faces allegations of violating U.S. regulations through its Qwen lab, while the Chinese giant seeks to solidify its presence in the open AI infrastructure ecosystem by making SAIL publicly available.

It is worth noting that the company has shipped over 560,000 Zhenwu chips to customers worldwide, strengthening the integration of its technical ecosystem and making it difficult to separate from the global AI landscape amid any new political restrictions.

Scroll to Top