Chinese AI company DeepSeek says it is working with Huawei to develop open-source programming tools for Huawei’s Ascend artificial-intelligence processors, as China continues efforts to build more of the software ecosystem needed around domestic AI hardware.
Reuters reported on 30 September that DeepSeek announced the partnership through its official WeChat account. The company said it is open-sourcing programming infrastructure for the Ascend platform, including compute and communication libraries.
DeepSeek also highlighted TileLang, a high-level open-source programming language designed to make AI-chip programming easier while still allowing developers to target hardware performance.
Why the software layer matters
The global AI-chip market is often discussed as a contest over processor speed, memory and manufacturing capacity. But developers also depend heavily on the software surrounding those processors.
Nvidia’s position is reinforced by CUDA, the programming environment and library ecosystem that has become deeply embedded in AI research and commercial development. A competing chip can therefore face a major adoption barrier if developers need to rewrite software, learn unfamiliar tools or give up mature libraries.
DeepSeek and Huawei’s work is aimed at reducing that friction for the Ascend platform by providing higher-level programming tools and communication libraries.
A 128-chip system is part of the work
DeepSeek said Huawei provided extensive support while the two companies advanced a supernode system based on 128 Ascend 950 processors, with work focused on both computation and communication between the chips.
That matters because large AI models are generally trained across many accelerators working together. The speed of communication between processors can be as important as the capability of an individual chip.
No evidence yet of parity with Nvidia
The announcement should not be read as proof that the Huawei and DeepSeek software stack now matches Nvidia’s CUDA ecosystem in performance, developer adoption or maturity.
What it does show is that the competition is widening beyond hardware specifications. Programming languages, libraries, interconnect software, developer tools and model frameworks are becoming part of the same contest.
Why African AI developers should watch the shift
For developers and governments across Africa, a broader range of viable AI-computing platforms could eventually affect pricing, procurement and availability. But access will also depend on export controls, cloud-provider support, technical compatibility and whether the tools develop a large enough developer community.
Countries planning public AI infrastructure will therefore need to evaluate complete computing ecosystems rather than comparing processors only by headline specifications.
Source: Reuters, 30 September 2026. DeepSeek partners with Huawei to develop chip programming tools.