Cambricon, NVIDIA’s New AI Chip Competitor in China, Aims to Raise $560 Million Amid Beijing’s Push for Domestic AI Chips in Datacenters

China’s artificial intelligence (AI) sector is currently undergoing significant transformation, driven by recent governmental initiatives aimed at bolstering domestic chip production. These developments are largely influenced by the escalating trade tensions between the United States and China, both of which are reassessing their technological strategies.

Domestic AI Chips in China: Emphasis on Self-Reliance Amid NVIDIA Dominance

In light of the ongoing geopolitical climate, Beijing is focusing on curbing its dependency on NVIDIA’s H20 AI chip. Reports indicate that the Chinese government is advising local technology firms against procuring foreign chips, citing concerns over potential security vulnerabilities. This push for autonomy has resulted in new regulations mandating that over half of the AI chips utilized in data centers must be sourced from domestic manufacturers, providing a substantial boost to companies like Huawei and Cambricon.

Specifically, the government is striving to decrease reliance on NVIDIA amid fears that the U. S.might implement security measures in chips exported to China. As a result, there is speculation that chips such as Huawei’s Ascend may gain broader acceptance. However, the challenge remains that current domestic AI chip offerings are reportedly insufficient to meet the high performance needs required for training leading-edge AI models. This technology gap has delayed the launch of DeepSeek’s anticipated R2 model, highlighting ongoing hurdles in the industry.

With the heightened interest in developing local AI chips, companies like Cambricon are seizing the opportunity to boost their financing for ambitious projects. The firm is projected to raise approximately 4 billion yuan to enhance its offerings positioning them as a viable alternative to industry stalwarts like AMD and NVIDIA. Cambricon’s product lineup includes the Siyuan Series tailored for data centers and cloud computing, while also advancing cutting-edge solutions for large language models (LLMs), enabling local businesses to develop next-generation AI technologies. Nonetheless, significant breakthroughs remain elusive for the time being.

Huawei Preps Ascend 910C To Tackle NVIDIA's H100 In China's Domestic AI Market 1 — Image Credit: Huawei – Ascend AI chip

Huawei currently stands as a leading supplier of AI chips within China, with its Ascend lineup featuring several models, including the highly anticipated Ascend 910B and 910C. The latter is reportedly designed to outperform NVIDIA’s H100 chip in training performance. In addition, Huawei offers a rack-scale solution known as CloudMatrix 384, which aims to compete with NVIDIA’s Blackwell NVL72 system. Yet, there’s a notable challenge: the lack of software compatibility with NVIDIA’s widely utilized CUDA Toolkit, making the transition to domestic solutions difficult for Chinese firms.

While China is actively seeking substitutes for NVIDIA’s AI chips, it appears that, at least for the foreseeable future, the nation will continue to rely on American technological innovations. The need for developed software frameworks alongside advanced hardware remains a crucial aspect of building a self-sufficient AI industry.

Source & Images