Huawei Challenges Nvidia by Launching an Open-Source AI Ecosystem

Tech
By 24matins.uk,  published 6 August 2025 at 17h19, updated on 6 August 2025 at 17h19.
Tech

Huawei is stepping up competition against Nvidia by launching its own open-source AI ecosystem. This strategic move aims to boost Huawei’s position in the rapidly evolving artificial intelligence sector, offering developers new alternatives and fostering innovation in AI technology.

Tl;dr

  • Huawei open sources CANN to rival Nvidia’s CUDA.
  • CANN aims for AI software independence in China.
  • Ascend ecosystem seeks long-term competitiveness.

Strategic Shift: Huawei’s Open Source Gamble

Huawei Technologies has taken a bold step in the ongoing competition over AI chips. During its developer conference in Pékin, rotating chairman Eric Xu Zhijun announced the company would open source its crucial software toolkit, CANN (Compute Architecture for Neural Networks). The move is seen as a clear signal of intent to challenge the entrenched dominance of Nvidia—a name practically synonymous with the field due to its powerful hardware and proprietary software.

The Closed World of Nvidia CUDA

For years, developers have found themselves locked into Nvidia‘s tightly controlled ecosystem. The backbone of this regime is CUDA, a proprietary programming framework that leverages the full potential of Nvidia GPUs but offers little flexibility elsewhere. Even promising attempts at bypassing these restrictions—like ZLUDA, which aimed to run CUDA code on non-Nvidia GPUs through translation layers—were abruptly blocked when such workarounds were prohibited in early 2024. For many, this reinforced the sense that innovation outside Nvidia’s walls was stifled.

CANN: Opening Doors for Developers

In contrast, Huawei‘s decision to open source CANN marks an invitation to developers and industry partners to break free from such constraints. As Eric Xu put it, « accelerate developer innovation and simplify usage of Ascend chips ». Notably, the company has already reached out to major players in China’s artificial intelligence landscape—including university research labs and industrial partners—to foster an open and collaborative environment around its Ascend chips.

The practical implications for developers are significant:

  • CANN provides multiple programmable interfaces tailored for Ascend GPUs.
  • It aims to establish a credible Chinese alternative to CUDA.
  • The initiative encourages local AI software development targeting Ascend hardware.

    • A Long-Term Bet Against a Formidable Rival

    Of course, challenges remain. While the performance of Huawei’s Ascend processors is inching closer to Nvidia’s latest Blackwell B-series chips, there is no denying that CUDA’s nearly two-decade legacy presents a formidable benchmark. However, as trade tensions persist and technological sovereignty becomes increasingly valued in China, Huawei’s strategy may yet reshape the landscape. Only time will reveal whether the Ascend ecosystem can grow into an indispensable pillar for homegrown AI applications—potentially rewriting the rules of global AI chip competition.