China Is Rapidly Catching Up with the US in Artificial Intelligence: A Gap of 3–6 Months
In its extensive 2025 analysis titled "Measuring the US-China AI Gap," Recorded Future provides a comprehensive overview of the current state of competition between the United States and China in artificial intelligence. The report, prepared by Insikt Group, shows a dramatic narrowing of the technological gap between the two superpowers and reveals not only the rapid progress of Chinese AI models but also the sophisticated methods China uses to achieve these results. According to the study's conclusions, Chinese generative artificial intelligence models currently lag behind their American counterparts by only three to six months in terms of performance, representing a dramatic improvement over previous years, when the gap was measured in years.
An analysis of model performance based on standardized benchmarks reveals the unprecedented speed at which China is catching up. While the overall performance gap between the best American and Chinese models stood at 9.26% at the beginning of 2024, by February 2025 it had fallen dramatically to just 1.70%. This convergence is even more pronounced in specific tests: on the MMLU benchmark, which measures broad knowledge, the US lead fell from 17.5% to just 0.3%; on the MATH benchmark focused on problem-solving, from 24.3% to 1.6%; and on the HumanEval code-generation test, from 31.6% to 3.7%. The AI field as a whole is thus converging at the technological frontier, with the gap between the best and tenth-best models on the Chatbot Arena leaderboard falling from 11.9% to 5.4%, while the gap between the top two models narrowed from 4.9% to just 0.7%.
Investment and financial support for artificial intelligence research are significant factors in this competition. In terms of government funding, China was likely exceeding total US federal and state investment in AI technologies by early 2025. The US budget measure for fiscal year 2025 allocated $11.2 billion to AI and IT research and development, representing an increase of nearly 27% over the previous year. Nevertheless, private investment in AI in the US reached an impressive $109.1 billion in 2024, almost twelve times China's $9.3 billion. This massive lead in private investment gives US AI companies significant access to venture capital, supporting rapid scaling and innovation, while Chinese companies rely more heavily on state-led initiatives. The differing approaches to financing reflect fundamental differences between the two countries' economic systems and their impact on technological development.
China's artificial intelligence ecosystem is maturing rapidly thanks to strong cooperation among government, industry, and academia, supported by steady progress in semiconductor manufacturing. However, China continues to face difficulties in meeting demand due to ongoing export controls. In this context, a major turning point came in September 2024, when TSMC (Taiwan Semiconductor Manufacturing Company) violated US export controls by producing advanced AI chips for Huawei through a Chinese intermediary company. The scale of this violation was unprecedented—TSMC manufactured approximately 3 million chip dies for Huawei using its 7nm process, enabling the production of China's Ascend 910B and upcoming 910C AI chips. While previous smuggling cases typically involved only tens of thousands of units, this incident involved millions of units, representing a qualitatively different scale of violation of the control measures.
Although these chips lag approximately four years behind the most advanced US technology, their collective computing power is substantial—equivalent to approximately 1 million Nvidia H100 chips, Nvidia's previous-generation chips from 2023. This incident significantly boosted China's AI computing capacity despite the fact that China lacks domestic capabilities to manufacture the most advanced semiconductors. The United States still maintains a strategic advantage in overall computing capacity thanks to its possession of more advanced AI chips, which remains a key factor in economic transformation and maintaining technological leadership in the global AI ecosystem.
In terms of talent, the United States still maintains a leading position thanks to its international AI talent pool, supported by elite educational institutions and its historical advantage in immigration. However, this lead is gradually weakening as China's domestic talent pool grows and universities strengthen their AI programs. The practical impact of the US talent advantage is also eroding as China retains more of its top graduates and accelerates domestic education. At the same time, it is becoming clear that the decisive factor in the AI race will not only be frontier innovation but also the ability to diffuse AI technologies across industry and society. While it is unclear which country currently leads in overall diffusion, China leads in AI patents across several industries, indicating strong momentum in the application of AI in manufacturing, finance, and other sectors.
The report also analyzes the regulatory environment and its impact on AI development in detail. China's regulatory environment may slow the development and deployment of consumer-facing AI products, but these restrictions do not significantly impede progress in cutting-edge AI research, particularly in government- or industry-focused solutions. Thus, while regulation affects commercial products, China's fundamental advances in AI remain robust. By contrast, the US regulatory environment is less restrictive, potentially accelerating product launches and research. This divergence in regulatory approaches has significant implications for the pace and nature of technological development in both countries.
The report pays particular attention to Chinese state-sponsored threats and methods of acquiring technology. China uses a variety of approaches to advance its AI capabilities, including state-led and tacitly tolerated economic espionage. In February 2025, the US Department of Justice charged a former Google software engineer with attempting to steal proprietary information about hardware and software enabling supercomputers for large AI models. China's AI sector also benefits from coordinated efforts to attract and recruit foreign talent. In addition, there is evidence that Chinese companies are circumventing the terms of service of foreign platforms to improve their own models through model distillation techniques. These activities represent a sophisticated approach to acquiring technological knowledge and expertise that combines both legal and problematic methods.
In the context of manufacturing capabilities and digitalization, the report emphasizes that US companies lead in generative AI applications across the manufacturing sector and are pioneering new paradigms such as software-defined manufacturing. In China, however, the adoption of advanced digital manufacturing remains limited—as of 2022, only 37% of Chinese manufacturers had achieved a basic level of digitalization, and just 4% had cutting-edge capabilities. This gap in the application of AI technologies in manufacturing represents a significant opportunity for US companies, which can leverage their technological lead to gain a competitive advantage in the global manufacturing sector.
Insikt Group analysts emphasize that maintaining an accurate and up-to-date understanding of China's AI capabilities is "almost certainly" critical to US and allied corporate and national security decision-making in the coming years. The report refutes the notion of a sudden "Sputnik moment" for Chinese AI and instead attributes China's progress to sustained, long-term investment and iterative innovation, particularly in response to external pressures such as chip controls. This perspective is essential for understanding the true nature of the technological competition between the two countries.
Overall, the report "Measuring the US-China AI Gap" by Recorded Future's Insikt Group presents a comprehensive analysis of the rapidly evolving competition between the US and China in AI. While the US maintains critical advantages in private investment, elite talent, and advanced chip manufacturing, China's government coordination, industrial policy, and improving model performance point to a dynamic landscape with significant implications for both the economy and national security. The fact that Chinese models have achieved near parity with their American counterparts after previously trailing by dozens of percentage points demonstrates the effectiveness of China's approach and the need for US policymakers to maintain strategic thinking and investment in the future of AI technologies. This analysis is essential for understanding the future geopolitical balance of power in the age of artificial intelligence and for formulating appropriate strategic responses on both sides of the Pacific.
