Nvidia Faces New AI Challenge As Chinese Start-Up Unveils Framework To Cut Reliance On Its GPUs
A new AI framework called "Chitu" developed by Chinese start-up Qingcheng.AI aims to reduce reliance on Nvidia chips for AI model inference. The framework, designed by Tsinghua University professor Zhai Jidong, is tailored for large language models and can run on Chinese-made chips. This move is part of Chinese tech companies' efforts to achieve technological self-sufficiency amidst U.S. export restrictions. Chitu directly challenges Nvidia's dominance in supporting advanced AI models like DeepSeek-R1. In testing, Chitu showed a significant increase in model inference speed and reduced GPU usage compared to foreign frameworks.
AI Framework Development in China
A new artificial intelligence framework developed by a Chinese start-up could reduce reliance on Nvidia Corp.'s NVDA chips for AI model inference, marking the latest effort by Chinese tech companies to achieve technological self-sufficiency amid U.S. export restrictions.
Overview of the Framework
Qingcheng.AI, founded by Tsinghua University professor Zhai Jidong, unveiled “Chitu” – a high-performance inference framework for large language models that can operate on Chinese-made chips, reported South China Morning Post. The open-source framework directly challenges Nvidia’s dominance in supporting advanced AI models like DeepSeek-R1, according to a joint statement released Friday.
When tested with DeepSeek-R1 using Nvidia’s A800 GPUs, Chitu achieved a 315% increase in model inference speed while reducing GPU usage by 50% compared to foreign frameworks, the company claimed.
Implications and Competition
The development comes amid heightened AI competition in China following DeepSeek’s breakthrough in January, which demonstrated high-performance models at significantly lower costs than U.S. competitors. This success triggered a selloff in U.S. tech stocks, with Nvidia losing $593 billion in market value as investors feared increased competition.
Chinese tech giants are rapidly entering the race, with Baidu Inc. unveiling ERNIE X1 and claiming performance comparable to DeepSeek-R1 at half the price. Alibaba Group Holding Ltd. recently launched R1-Omni, an emotion-reading AI model available for free. These initiatives reflect China’s broader push for technological independence after Washington banned sales of Nvidia’s advanced H100 and H800 chips to China-based clients.
Backed by Beijing’s municipal AI fund, Qingcheng.AI has partnered with top Chinese GPU manufacturers including Moore Threads, Enflame, and Iluvatar CoreX. The AI surge has reignited investor interest in Chinese tech, potentially reversing three years of venture capital decline.
Future Prospects
As the Chinese government prepares a fund expected to mobilize 1 trillion yuan ($137.7 billion) for tech investment, the industry is witnessing a rush to discover the next breakthrough akin to DeepSeek.
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Stock Score Locked: Want to See it? Benzinga Rankings give you vital metrics on any stock – anytime. © 2025 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.