Alibaba Group Holding’s chip design unit, T-Head, announced on Saturday, July 18, 2026, at the World AI Conference (WAIC) in Shanghai that it will open-source its proprietary software stack, SAIL. This foundational software architecture supports T-Head’s Zhenwu series of AI chips and is now freely available to international developers.

This initiative is part of a broader push by Chinese AI chipmakers, including Huawei Technologies and Moore Threads Technology, to foster open and collaborative software ecosystems as alternatives to Nvidia’s CUDA toolkit, which remains the industry standard for programming graphics processing units (GPUs). Currently, most AI programmers worldwide rely on Nvidia’s specialized software, which ties them closely to Nvidia’s hardware.

T-Head stated that its open-source effort aims to lower barriers for developers globally to adopt its hardware. The company added that programmers could adapt the SAIL stack to mainstream AI frameworks in less than seven days.

This move follows a similar strategy by Huawei in 2025, when it open-sourced its Compute Architecture for Neural Networks (CANN), the software platform for its Ascend AI processors. These efforts reflect a broader goal among Chinese technology firms to increase self-sufficiency amid ongoing US-China technology competition.

Sources