Chinese AI company DeepSeek has open-sourced six software tools built for Huawei’s Ascend AI chips, aiming to give Chinese developers a homegrown alternative to Nvidia’s software stack at a time when US export controls keep most advanced American processors out of reach.
The Hangzhou-based startup announced the release on its official WeChat account, according to Reuters. The package includes modules for computing and communication workloads, with DeepSeek saying the tools are based on open-source components it had previously developed for Nvidia hardware, which have now been adapted for Huawei’s processors. Alongside the software, the two firms co-developed what DeepSeek referred to as a “supernode” system running on 128 of Huawei’s Ascend 950 accelerators. The system is tuned to balance computation and data movement across the cluster.
DeepSeek open-sources six software tools for Huawei
The release, as reported by the South China Morning Post, is meant to seed an “independent and controllable” software ecosystem for GPUs instead of only a single product. The biggest part of the release is an Ascend-compatible version of TileLang, DeepSeek’s high-level programming language for writing the critical kernels that drive model performance. Nvidia remains its main back end, but the language now officially supports Huawei’s Ascend 950, adding features such as native code generation, automatic scheduling and synchronization, according to SCMP.
DeepSeek also positioned TileLang as an alternative to Nvidia’s proprietary CUDA platform, saying it simplifies programming while still allowing developers to get the most out of the underlying hardware. The company described it as an early step toward a more independent GPU software ecosystem. CUDA’s dominance is what makes the challenge so difficult. Nvidia’s nearly two-decade lead in CUDA tools and libraries has made its chips a core part of AI workloads, and no Chinese rival has managed to break that hold.
As US export controls limit Chinese companies’ access to Nvidia’s most advanced chips, Huawei has been pushing its Ascend processors even more. The DeepSeek tools arrived about two weeks after Huawei unveiled its next-generation AI processors and supernode systems, which the company expects to begin training models by next year. A week earlier, Huawei said it would ship its next-generation Ascend 960DT chip in the first quarter of 2027, moving the launch up by three quarters, and drew a chip roadmap stretching up to 2029.
Rotating Chairman Eric Xu claimed Huawei’s Ascend chips already hold a bigger slice of China’s AI chip market than Nvidia does, although he gave no figures to back this claim. Wednesday’s release builds on a partnership that has grown closer this year. DeepSeek had already previewed its V4 model running on Huawei’s Ascend processors, moving away from its earlier reliance on Nvidia. Huawei also stated at the time that it had worked with DeepSeek to make V4 compatible across the entire Ascend lineup.
The partnership also shows how China’s biggest AI companies are adapting to Nvidia’s absence and are not just waiting for export restrictions to ease at some point in the future. The key question remains how much these tools can narrow the performance gap that has helped keep developers tied to Nvidia’s CUDA, which is one that we still have no answer to, for now.
