- Huawei makes its Cann Ai GPU Toolkit Open Source to challenge Nvidia’s proprietary Cuda platform
- Cuda’s nearly 20-year-old dominance has excluded developers in Nvidia’s hardware ecosystem exclusively
- Cann supplies multilayer programming interfaces to AI applications on Huaweis Ascend AI GPUs
Huawei has announced plans to make his Cann Software Toolkit for Ascend Ai Gpus Open Source, a step aimed at challenging Nvidia’s long-time cuda dominance.
Cuda, often described as a closed “Vollgrav” or “swamp”, has been considered a barrier to developers seeking compatibility across platforms of some for years.
Its tight integration with NVIDIA hardware has locked developers in a single supplier ecosystem for almost two decades, with all efforts to bring CUDA functionality to other GPU architectures through translation layers blocked by the company.
Opening Cann for Developers
Cann, abbreviated for calculating architecture for neural networks, Huawei’s heterogeneous computer framework is designed to help developers create AI applications for its ascend AI GPUs.
Architecture offers several programming layers, giving developers opportunities to build both high level and performance -intensive applications.
In many ways, it is Huawei’s equivalent to Cuda, but the decision to open its source codes signals an intention to grow an alternative ecosystem without the limitations of a proprietary model.
Huawei has reportedly already begun discussions with major Chinese AI players, universities, research institutions and business partners to contribute to an open sourced ascend development community.
This search could help accelerate the creation of optimized tools, libraries and AI frames for Huawei’s GPUs, potentially making them more attractive to developers who are currently dependent on NVIDIA hardware.
Huaweis AI -hardware -performance has improved steadily with the claims that certain ascend chips can surpass NVIDIA processors under specific conditions.
Reports such as CloudMatrix 384’s benchmark results against Nvidia running Deepseek R1 suggest that Huawei’s performance course closes the gap.
However, RAW Performance alone will not guarantee developer migration without similar software stability and support.
While Open Sourcing Cann could be exciting for developers, its ecosystem is in its early stages and may not be something close to Cuda, which has been refined for almost 20 years.
Even with open source status, adoption may depend on how well cannon supports existing AI frames, especially for new workloads in large language models (LLM) and AI author Tools.
Huawei’s decision could have wider consequences beyond the convenience of the developer as the open-sourcing channel adapts to China’s wider push for technological self-sufficiency in AI-computing, reducing the dependence on Western chipmakers.
In the current environment where US limitations are targeted at Huawei’s hardware exports, it becomes equally critical to build a robust domestic software stack for AI tools such as improving the chip performance.
If Huawei can successfully promote a vibrant open source community around Cann, it could present the first serious alternative to cuda for years.
Nevertheless, the challenge is not only in coding, but in building trust, documentation and compatibility on the scale, Nvidia has achieved.
Via Toms hardware
