• Home
• Technology

Chinese artificial intelligence company DeepSeek has announced a strategic collaboration with Huawei Technologies to develop optimized software tools for Huawei’s Ascend AI chips. The partnership serves as the latest indicator of Chinese technology firms deepening their cooperation to build a robust domestic alternative to Nvidia’s dominant AI computing ecosystem.

In an official statement published on WeChat, DeepSeek revealed that it will release an open-source software architecture for the Ascend platform, which will include specialized libraries tailored for computation and communications.

By transitioning to an open-source model, DeepSeek is making the underlying source code accessible to the broader developer community, enabling engineers to inspect, modify, and build new applications directly on top of the framework.

Co-Developing the Ascend Software Architecture

DeepSeek noted that Huawei provided comprehensive support in developing the new software architecture. The two companies collaborated closely to engineer a "Super Node" solution powered by a cluster of 128 Ascend 950 chips.

This advanced solution is explicitly designed to maximize the efficiency of both computing processes and inter-chip communication, thereby allowing the massive workloads associated with training AI models to run far more efficiently.

This collaborative milestone arrives just two weeks after Huawei officially unveiled its next generation of AI processors and "Super Node" computing systems. Huawei anticipates that its proprietary hardware and computing systems will see widespread deployment for training advanced AI models over the coming year.