Chinese AI company DeepSeek is working with Huawei to develop programming tools for Huawei Ascend chips, Reuters reported on Tuesday.

The move is seen as targeting Nvidia's CUDA software ecosystem.

DeepSeek said on its official WeChat account it would open-source programming infrastructure for the Ascend platform, according to the report. What it disclosed includes computing libraries and communications libraries.

The announcement came 2 weeks after Huawei unveiled next-generation AI processors and a supernode computing system. Huawei expects its AI systems to be widely used for model training in 2027.

Huawei fully supported DeepSeek's development of the programming infrastructure.

DeepSeek and Huawei also developed a supernode solution made up of 128 Ascend 950 chips, boosting computing and communications performance, Reuters reported.

DeepSeek also highlighted the role of the open-source programming language for AI chips, TileLang. The company said TileLang improves development efficiency and simplifies code structure.

DeepSeek said, "To build an independent next-generation GPU software ecosystem, you first need an advanced language that is general-purpose and easy to program while drawing out hardware performance to the fullest." It added, "That is why we developed TileLang," and said TileLang has a simpler programming method than Nvidia's CUDA.