Digital Transformation

China narrows AI gap with US despite chip constraints

Phan Van Hoa DNUM_ABZAIZCACE 17:58

China is rapidly closing the gap and competing head-to-head with the US in artificial intelligence (AI), despite a shortage of advanced chips.

China is rapidly closing the gap and is competing head-to-head with the United States in the field of artificial intelligence, according to experts and analysts in the field of AI. The boom in AI applications, fueled by strong government support, has helped China make surprising strides despite a shortage of advanced chips.

The race to develop AI among the tech superpowers is fierce, and China is no exception. The country's leading tech companies are focusing on creating large language models (LLMs) - the foundation of innovative AI applications like ChatGPT.

Notably, many Chinese companies have announced LLMs that are said to be on par or even superior to their US counterparts, which is all the more impressive given that US restrictions on the export of advanced chips are causing major difficulties for China’s AI industry.

Anh minh hoa1
Illustration photo.

“Faced with a shortage of advanced graphics processing units (GPUs) due to US sanctions, Chinese tech companies are actively looking for breakthrough solutions to improve the efficiency of AI models. This urgency has become the driving force for continuous innovation and creativity in the field of AI in China,” said Winston Ma, author of “Digital Warfare: How Chinese Tech Companies Are Shaping the Future of AI, Blockchain, and Cyberspace.”

In the race to develop AI technology, a little-known startup from Beijing, Shengshu AI, has just surprised everyone by launching a text-to-video tool, Vidu. With the ability to create vivid clips from simple commands in both Chinese and English, Vidu has joined the fierce competition with giants like Kuaishou and Zhipu AI, promising to bring a new breeze to the technology market.

While Sora is considered a pioneer in the field of text-to-video, the technology race has seen a strong rise of Chinese startups. While OpenAI, the father of this idea, is still in the testing phase, Chinese companies such as Kuaishou, Zhipu AI and now Shengshu AI have quickly brought their tools to global users, creating a fierce competition.

Chinese tech companies are taking a leading role in expanding the global AI development community. By releasing open-source LLMs like Alibaba’s Qwen2, they are giving anyone from researchers to independent developers the opportunity to build and customize their own AI systems. The fact that Alibaba, which owns the South China Morning Post, is leading this trend underscores the importance of open source in the future of AI.

According to Clement Delangue, co-founder and CEO of the US-based machine learning and AI startup Hugging Face, Qwen 72B has outperformed its competitors and China's open-source models are completely dominating those of other countries.

China's rapid progress in AI is partly due to its ability to overcome chip constraints to develop the intelligent computing power needed to train local LLMs, analysts say.

In the face of export restrictions on Nvidia’s A100, H100, A800, and H800 chips imposed by the United States, China has shown considerable initiative in developing domestic solutions. By investing heavily in chip research and production, Beijing and its leading technology companies have built a solid foundation for intelligent computing that is competitive in the international arena.

“Looking at the numbers, domestic computing power has increased rapidly as many state-owned enterprises and local governments have been tasked with developing intelligent computing power,” said Li Yangwei, a technical consultant working in the field in Beijing.

Chips developed by domestic companies such as Huawei Technologies have become popular, and Huawei's Ascend AI chip solution is China's best solution in developing domestic AI infrastructure, Li Yangwei added.

Wang Tao, CEO of the Kunpeng Ecosystem Innovation Center in Jiangsu, said on the sidelines of the Nanjing World Semiconductor Conference in June that Huawei's Ascend 910B AI chip has been proven in some tests to have performance reaching 80% to 120% compared to Nvidia's A100 chip when training LLMs.

The state's build-up of computing resource infrastructure has also helped ease concerns about a shortage of advanced chips, said Zhang Yi, founder and chief analyst at technology consultancy iiMedia.

Compared to other countries, China's large market size and public sector demand for AI and its applications are advantages driving the development of AI in the country.

Furthermore, with a large-scale industrial production system and the need to improve production efficiency, China is opening up a huge market for AI to maximize its potential.

Phan Van Hoa