China launches 100 times faster AI model, challenging US AI

China's launch of its SpikingBrain 1.0 artificial intelligence (AI) model - 100 times faster than conventional models and independent of Nvidia chips - makes the AI ​​race between the US and China even more dramatic and exciting.

Chips produced by Chinese technology company MetaX are displayed at the World AI Conference in Shanghai on July 29, 2025 - Photo: VCG

Chinese state media recently proudly announced the world's first large-scale brain-inspired AI system, SpikingBrain 1.0, developed by the country. They called it a breakthrough that could "change the landscape of global technology development and energy sustainability," making the AI ​​race between the two superpowers even more intriguing.

Breakthrough AI technology

Current large language models (LLMs) like ChatGPT or Gemini are powerful but power-hungry and require huge computational resources. In China, this problem is exacerbated by limited access to Nvidia chips – the type used for most advanced AI today.

Scientists at the Chinese Academy of Sciences (CAS) have found another way with SpikingBrain - a system that simulates the way neurons use electrical pulses to activate, helping to cut energy and increase processing speed.

The team built two versions: a 7 billion parameter model and a 76 billion parameter model. In tests, the smaller version outperformed traditional AIs. SpikingBrain was 100 times faster with a 4 million token input text. In another test, the model generated the first token from a 1 million token context 26.5 times faster.

What’s most impressive is that this performance is achieved with less than 2% of the training data compared to current LLMs, which could solve the problem of increasingly depleted online data sources to “feed” massive AI models.

The model opens up a new direction for AI development, providing an optimal framework for domestic chip platforms in China, said Li Guoqi, principal researcher at CAS. SpikingBrain can be useful in processing long data chains such as legal documents and medical records.

The race is fierce

According to Stephen S. Roach, a lecturer at Yale University, although no one has officially declared victory in the US-China AI race, the market is betting on the US. Nvidia has become the world's first $4 trillion company, and Microsoft has also reached a valuation of $3.7 trillion.

But being early doesn’t mean winning, especially when it comes to innovation. Almost every day there are new reports of Chinese AI advancements. The US may have ushered in a new era with ChatGPT, but China’s DeepSeek and many subsequent models have also amazed the world.

In recent years, since President Donald Trump's first term, the US has increased restrictions on technology exports to China, preventing the country from accessing chip manufacturing tools for servers, AI and advanced applications.

In April, Chinese President Xi Jinping said Beijing must “respond to the rapid development of new-generation AI, leverage the power of the national system to achieve self-reliance.” Beijing has also put pressure on Washington, recently banning major domestic tech companies from buying AI chips from Nvidia.

Major companies like Alibaba and Huawei have jumped into developing their own AI chips. According to research firm Central China Securities, Nvidia’s market share of AI chips in China is expected to fall from 66% in 2024 to 54% in 2025, while Huawei’s will increase from 23% to 28%.

Chinese tech giant Tencent has also announced a shift to domestically produced AI chips. However, Charlie Dai, vice president and principal analyst at Forrester Research, noted that the shift is “accelerating but not yet complete.” China is aiming to “accelerate the maturation of its semiconductor ecosystem.”

Two separate ecosystems

According to Forbes magazine, the US's restrictions on the export of advanced chips have triggered a two-pronged response from China. On the one hand, a black market network has emerged, smuggling banned chips into the country. On the other hand, Beijing has strengthened its industrial policy, promoting technological innovation.

The result is the emergence of two distinct AI ecosystems: China has built its ecosystem on an open-source model and domestically optimized hardware. The US has pushed a closed, proprietary model, tied to hardware advantages. The race is not just about powerful chips, but also about talent, software, and strategic direction.

Back to topic

PEACEFUL

Source: https://tuoitre.vn/trung-quoc-tung-mo-hinh-ai-nhanh-gap-100-lan-thach-thuc-ai-my-20250928233649368.htm