• Wion
  • /Technology
  • /DeepSeek’s new Huawei partnership could reshape China’s AI chip race with Nvidia

DeepSeek’s new Huawei partnership could reshape China’s AI chip race with Nvidia

DeepSeek’s new Huawei partnership could reshape China’s AI chip race with Nvidia

DeepSeek’s new Huawei partnership could reshape China’s AI chip race with Nvidia

Story highlights

DeepSeek is partnering with Huawei to develop open-source programming tools for Ascend AI chips, including a 128-chip Ascend 950 supernode. The move targets Nvidia’s software advantage and supports China’s push for a more independent AI chip ecosystem.

DeepSeek and Huawei are joining forces to build software for Huawei’s AI chips, in a move that could strengthen China’s effort to reduce its dependence on Nvidia’s technology ecosystem.

The Chinese AI company said it is open-sourcing programming infrastructure for Huawei’s Ascend platform, including computing and communication libraries. DeepSeek said Huawei provided full support for the work and that the two companies jointly developed a “supernode” solution based on 128 Ascend 950 chips. The partnership matters because AI chip competition is not only about hardware. Software is just as important. Nvidia’s CUDA ecosystem has become a major part of how developers build and optimise AI workloads for its processors. DeepSeek is now highlighting a different approach.

DeepSeek targets Nvidia’s CUDA software advantage

Add WION as a Preferred Source

DeepSeek said the partnership is focused on creating programming tools optimised for Huawei’s Ascend chips. A key part of that effort is TileLang, a high-level open-source programming language designed for AI chips. DeepSeek said TileLang can simplify code development while still allowing developers to reach the performance of the underlying hardware. The company compared its programming model with Nvidia’s CUDA, saying TileLang offers a simpler approach. That does not mean Nvidia’s software ecosystem has been replaced. But it highlights an important part of China’s broader effort to develop alternatives that can work independently of Nvidia technology.

Why the 128-chip Huawei system matters

DeepSeek said it and Huawei jointly advanced a “supernode” solution using 128 Ascend 950 chips, with both computation and communication optimised.

Trending Stories

Supernode systems are designed to connect large numbers of processors so they can work together on demanding AI workloads. The announcement comes about two weeks after Huawei unveiled its next generation of AI processors and supernode computing systems. Huawei said at the time that its AI systems were expected to see wider use for model training next year. For DeepSeek, compatibility with Huawei’s hardware could provide another route for running and developing AI systems without relying entirely on Nvidia’s platform.

China’s AI chip ecosystem is becoming more important

DeepSeek’s announcement shows how Chinese AI and semiconductor companies are increasingly working together around domestic hardware and software.

The company said that building a “new generation” of independent AI chip software ecosystems requires a high-level programming language that is universal, easier to use and capable of taking advantage of hardware performance. The partnership therefore goes beyond one AI model or one processor. It is about building the software layer needed to make alternative AI chips useful to developers. For Nvidia, the challenge is not simply whether competing chips can match its hardware. The bigger question is whether rival ecosystems can attract developers and make it easier to move AI workloads away from CUDA. For now, DeepSeek and Huawei are building that alternative one layer of software at a time.

About the Author

Abhinav Yadav

Abhinav is a versatile and adaptive journalist who covers defence, space, and technology for WION. He specialises in breaking down complex subjects into clear, engaging stories tha...Read More