Chinese company seeks to reduce dependence on Nvidia.

The startup Qingcheng.AI and a development team led by computer science professor Di Guidong at Tsinghua University (China) recently announced the development of the Chitu AI framework, which is said to reduce reliance on Nvidia chips for AI model inference, according to the South China Morning Post on March 16.

Nvidia's logo at a technology exhibition in Spain on March 5th.

An AI framework is a toolkit for inference tools for large language models (LLMs), providing libraries and tools to help developers design, train, and test models efficiently. Chitu is an open-source tool that supports popular models such as Meta's Llama and DeepSeek-R1, a Chinese "homegrown" AI model that has been making waves worldwide in recent months for its perceived capabilities comparable to Western models but at a much lower cost.

In testing, when equipped with Nvidia's A800 GPU, Chitu increased the inference speed of the most powerful version of DeepSeek-R1 by 315% while reducing GPU usage by 50% compared to open-source frameworks from abroad, according to the company. This means the model can produce results much faster, saving time and computational resources.

Qingcheng.AI's development is part of an effort by Chinese AI companies to reduce their reliance on Nvidia's next-generation high-performance GPUs, which are currently subject to export restrictions. The US government has banned Nvidia from selling its Hopper H100 and H800 series chips to customers in China. On the other hand, according to AIBase.com, open-sourcing Chitu allows developers and researchers in China to freely use, modify, and optimize the tool, thereby promoting the development and improvement of domestic AI technologies.

China strives for self-reliance.

While Qingcheng.AI collaborates with leading domestic GPU manufacturers such as Moore Threads, Enflame, and Iluvatar CoreX, other technology companies in China are also pushing to reduce their dependence on foreign technology, based on the success of DeepSeek. In February, Infinigence AI, a provider of computing infrastructure platforms, announced it was working to promote cooperation among seven leading Chinese AI chip manufacturers: Biren Technology, Hygon Information Technology, Moore Threads, MetaX, Enflame, Iluvatar CoreX, and Huawei's Ascend.

According to China Daily on March 7, Liu Qingfeng, a member of the Chinese National People's Congress and chairman of AI company iFlytek, said that China urgently needs to research and develop domestically produced LLM models based on chips to build a strong AI ecosystem, ensuring sustainable and high-quality development. He argued that failing to develop a domestically produced AI industrial ecosystem is like building a tower on someone else's foundation. He stated that, aside from iFlytek's Spark model, all currently publicly downloadable LLMs are trained on Nvidia chips, highlighting China's shortcomings in chip development.

Currently, many large Chinese technology companies such as Alibaba, Tencent, Baidu, ByteDance, iFlytek, and Huawei, along with thousands of other startups, are racing to develop AI models. Most recently, Baidu announced two models, Ernie 4.5 and X1, last weekend to compete with models from DeepSeek and OpenAI.

Source: https://thanhnien.vn/cong-ty-trung-quoc-tim-cach-giam-phu-thuoc-nvidia-185250317205207263.htm