DeepSeek And Huawei Collaborate For AI Chip Software To Reduce Reliance On NVIDIA

Chinese artificial intelligence company DeepSeek has partnered with Huawei Technologies to develop open-source programming tools optimized for Huawei’s Ascend AI chips, as Chinese technology companies build more alternatives to Nvidia’s widely used hardware and software ecosystem.

Reuters reported that the collaboration includes programming infrastructure, compute and communication libraries, and work on a large-scale computing system using Huawei’s latest Ascend processors.

TL;DR

 

• DeepSeek and Huawei are developing programming tools optimized for Huawei’s Ascend AI chips.
• DeepSeek is open-sourcing software infrastructure developed for the Ascend platform, including compute and communication libraries.
• The companies also worked on a supernode solution using 128 Ascend 950 chips.
• DeepSeek says its TileLang programming language offers a simpler programming model than Nvidia’s CUDA.

According to Reuters, DeepSeek announced the collaboration through its official WeChat account, saying Huawei provided extensive support in developing programming infrastructure for its Ascend AI platform.

DeepSeek is also making parts of that infrastructure open source, including libraries designed for computation and communication between AI processors.

Yahoo Tech, which carried Reuters’ reporting on the announcement, noted that the partnership represents another step by Chinese technology companies to develop alternatives to Nvidia’s ecosystem. Nvidia’s strength in AI computing comes not only from its GPUs but also from CUDA, its long-established software platform used by developers to program workloads for its processors.

The announcement comes around two weeks after Huawei introduced its next-generation AI processors and supernode computing systems. Huawei expects its new AI systems to see wider use for AI model training next year, Reuters reported.

Source

DeepSeek And Huawei Target The AI Software Stack

The companies have also jointly advanced what DeepSeek described as a “supernode” solution based on 128 Ascend 950 chips.

The work focuses on optimizing both computation and communication across the system, an important consideration when hundreds or thousands of processors must work together to train or operate increasingly demanding AI models.

DeepSeek also highlighted TileLang, an open-source high-level programming language designed for AI chips.

The company said the software is intended to make AI chip programming easier while still allowing developers to take advantage of the underlying hardware’s performance.

“To build a new generation of independent, self-controlled GPU software ecosystems, the first priority is establishing a high-level language that is universal, easy to program, and still capable of reaching the hardware's full performance potential,” DeepSeek said, according to Reuters.

DeepSeek added that TileLang was created to address this requirement and said it provides a simpler programming model than Nvidia’s CUDA.

ANI also reported the DeepSeek-Huawei collaboration as part of the companies’ effort to build AI chip software that can reduce dependence on Nvidia.

Why The DeepSeek-Huawei Partnership Matters

Developing competitive AI chips is only one part of challenging Nvidia.

Software tools, libraries and developer support are critical because companies building AI applications need efficient ways to train models and run workloads across large groups of accelerators.

 

By making its Ascend-focused development tools open source, DeepSeek is helping expand the software layer surrounding Huawei’s processors while potentially giving developers more options for running AI workloads outside Nvidia’s ecosystem.

The move also arrives as DeepSeek pursues more control over its AI computing stack. Reuters reported in July that the company was developing its own AI chip aimed at inference workloads, citing people familiar with the project.

For now, Nvidia’s CUDA ecosystem remains deeply established across AI development. However, the DeepSeek-Huawei partnership shows that competition in AI infrastructure is increasingly moving beyond processor performance and into the programming tools and software ecosystems that determine how easily developers can actually use those chips.