How China’s AI Has Caught Up to the US: The Secret Behind Its Rapid Rise

LLM Scores of DeepSeek and Others Nearly Match the US
Accelerating Localization of AI Chips by Huawei and Cambricon
Government Efficiently Controls Data and Computing
Talent Development... China Has Three Times More Top 100 AI Researchers Than the US

The Chinese government has prohibited domestic companies from purchasing Nvidia's latest chips. After making its presence felt in the large language model (LLM) sector earlier this year with the shock of "DeepSeek," China is now actively fostering domestic chipmakers such as Huawei and Cambricon to compete against Nvidia. What is the secret behind China's rapid catch-up with the United States?

China Narrowing the Gap with the US in LLMs... Accelerating Localization of Domestic Chips

According to Frost & Sullivan, the global artificial intelligence (AI) market was worth $615.7 billion (approximately 86.2 trillion won) last year. Of this, the Chinese market accounted for about 700 billion yuan ($133 billion), representing 15.4% of the total market. This makes China the second-largest market in the world after the United States. China is quickly catching up to the United States in both AI software (LLMs) and hardware (chips).

Looking at the leaderboard trends from LMSYS Chatbot Arena, the score gap between US and Chinese models was about 60 to 80 points until mid-last year. However, as of August this year, the highest national scores are 1,470 points for the US and 1,427 points for China-a difference of only 43 points. The average score for China has also narrowed to about a 20-point gap with the US, standing at 1,373 points. While OpenAI (ChatGPT), Anthropic (Claude), and Google (Gemini) still dominate the top ranks, Chinese models such as DeepSeek (V3) and Alibaba (Qwen) have quickly risen to the upper tiers.

The United States still controls the key resource of GPUs. However, China, leveraging government-led localization strategies and massive financial support, is working to improve the performance of domestic chips, focusing on companies like Huawei and Cambricon. With abundant electricity production and the world's second-largest domestic market, demand for domestic chips in China is expected to grow rapidly, even if they are less efficient than US chips.

Centralized Policy Control Proves Effective... Efficient Use of Data and Talent

What is the core of China's AI competitiveness? According to the "China AI Investment Note" report by Shinhan Investment Corp., the key lies in China's strong centralized policy implementation. Shin Seungwoong, a strategist, explained, "China's unique strengths include extensive access to data, proactive acquisition of AI infrastructure, and a systematic environment designed to allow researchers to focus on R&D."

In countries like the United States and South Korea, issues such as personal data protection make it difficult to utilize private sector data. However, in China, where the government can quickly reprioritize national resources, a system is in place that allows even private data to be mobilized when necessary. In addition, the government has rapidly fostered LLM and chip companies through a strategy of intensive support for leading firms.

Another key factor is "talent." Of the 96,000 major AI research papers published over the past decade, the United States accounted for 35,000 (36%), while China produced 31,000 (33%). However, since 2023, China has surpassed the US in publication volume and widened the gap last year. According to data from the United Nations Industrial Development Organization, 57 of the world's top 100 AI researchers are affiliated with China, while only 20 are with the United States.

This content was produced with the assistance of AI translation services.