[DigitalToday reporter Chi-gyu Hwang] Alibaba Group Holding's chip design unit T-Head said it will release its in-house software stack as open source. The South China Morning Post reported on July 18 that the move is aimed at challenging Nvidia's dominance of the CUDA ecosystem.

T-Head announced at the World Artificial Intelligence Conference (WAIC) in Shanghai on July 18 that it will make the full technical stack of its software architecture, SAIL, available for free to overseas developers. It uses SAIL to develop its Zhenwu series of AI chips.

The move reflects a trend among Chinese AI chipmakers such as Huawei Technologies and Moore Threads Technology to build an open, collaborative software ecosystem as an alternative to Nvidia's CUDA toolkit, the industry standard for writing GPU software.

Most AI developers worldwide remain dependent on Nvidia-only software and in effect have no choice but to keep using Nvidia hardware. Chinese tech companies are seeking to boost self-reliance in the U.S.-China technology rivalry by offering alternative frameworks.

Huawei last year pursued a similar tactic by releasing as open source its CANN software platform for developing Ascend AI processors.

T-Head explained that open-sourcing SAIL is intended to lower entry barriers for overseas developers seeking to adopt its hardware. It added that developers can convert existing AI framework code to fit the SAIL stack within a week.

Gao Hui (가오후이), a T-Head vice president, said at WAIC, "SAIL has made developer experience its core philosophy from the first line of code." He stressed, "With the SAIL stack, developers can directly migrate and reuse existing code with minimal changes."