China Narrows AI Gap, Challenges US Dominance

As the technological gap in artificial intelligence (AI) between China and the United States narrows, Beijing is vigorously promoting a narrative centred on “open source” and “inclusive AI” as the foundation of its global governance vision.

In his keynote speech at the World Artificial Intelligence Conference (WAIC) in Shanghai on Friday (17/07), Chinese President Xi Jinping affirmed his country’s ambition to take a leading role in setting global AI standards. He criticised what he called the “excessive expansion of the concept of national security.”

Xi Jinping did not explicitly mention the United States. However, Beijing’s efforts to position itself as the champion of a new AI order and to challenge Washington’s dominance were plainly evident.

Speaking to DW, Rogier Creemers, an assistant professor at Leiden University in the Netherlands who focuses on Chinese digital technology policy, said the 2026 Shanghai summit sent several important signals. In his view, WAIC demonstrates that AI has now become one of Beijing’s political priorities.

Creemers said China intends to challenge the US through its support for open source AI models and by driving AI-based industrialisation. He also noted that China hopes to win backing from Global South countries for these efforts.

Beijing narrows the AI gap

China is closing in on the United States in the development of AI models. Meanwhile, the application of AI in industrial processes is gradually maturing.

Paul Triolo, senior vice-president for China and head of technology policy at DGA-Albright Stonebridge Group, told DW that Chinese AI model developers continue to compete on the global stage, chiefly by training open-weight models using computer hardware that is limited yet still sophisticated.

Triolo believes that AI adoption across Chinese companies may be proceeding faster than in the United States.

Moreover, public fear of AI in China is lower than in the US, accelerating the uptake of AI applications at the consumer level.

In early 2026, Chinese state media declared that the performance gap between Chinese and American AI models had “largely disappeared.”

A number of experts broadly agree with this assessment, noting that several benchmark tests of Chinese Large Language Models (LLMs), such as GLM-5.2 and Kimi K3, show the gap narrowing or even closing entirely.

For most applications, Triolo said, China’s leading open-weight AI models are cheaper to run and more than capable of handling the majority of tasks.

At the same time, advanced proprietary models from leading American firms such as Anthropic, OpenAI and Google still hold an edge in safety and scalability.

Are US AI blockades effective?

In January 2025, the US Department of Commerce expanded export restrictions related to AI through the “Framework for Artificial Intelligence Diffusion”, covering advanced computing chips, certain AI model weights, and access to large-scale AI computing infrastructure.

The US also tightened export controls on advanced chips and semiconductor manufacturing equipment destined for China.

Laila Khawaja, research director at Gavekal Research, told DW that these US efforts have been partly successful.

Export controls have indeed affected China’s ability to produce chips domestically, cheaply and at scale.

However, regarding export controls on advanced AI chips, Khawaja noted that large quantities of chips still enter China through the black market. At the same time, Chinese companies retain access to computing resources via cloud services abroad.

According to Triolo, the evidence so far shows that the US strategy to slow China’s advanced AI development has failed on many levels. At the same time, it has produced significant side effects for American companies, particularly on their research and development budgets and supply chains, such as those supplying rare earths.

China’s push for open source AI

In his speech, Xi Jinping emphasised the importance of “promoting open source development.”

Chinese state media have also promoted the message that China “does not impose technology blockades.”

Analysts believe Beijing is increasingly positioning “open source” and “inclusive AI” as a counter-narrative to the largely closed US AI ecosystem.

Most Chinese AI models are open-weight, with flagship products such as DeepSeek, GLM 5.2 from Zhipu, and Kimi K3 released under the MIT licence, allowing users to download, deploy, modify, distribute and commercialise them.

Creemers noted that, for now, Chinese AI models are quite open in terms of weights, although access to their source code remains limited.

In Creemers’ view, China hopes to encourage other countries to adopt Chinese AI models, thereby creating overseas markets for its technology and ensuring Beijing remains part of the next wave of technological development.

China’s WAICO challenges Pax Silica

In Shanghai, Xi Jinping also announced that the World AI Cooperation Organization (WAICO), an initiative launched by China, has attracted 29 member states.

He described the organisation as “an important milestone in the history of AI development” and said it responds to Global South countries’ calls for a greater role in AI governance.

WAICO is seen as an initiative to rival the US-led international Pax Silica initiative, which aims to secure global AI supply chains and critical minerals.

The real competition, according to Triolo, lies in which parts of the US and Chinese AI technology stacks are ultimately adopted by Global South countries.

He believes China holds a significant advantage thanks to lower costs and the digital infrastructure already widely built by Chinese companies such as Huawei in many developing countries.

The future of the AI rivalry

Overall, experts believe the United States will continue to lead in AI hardware design and computing capacity, while China has the potential to seize the advantage in model innovation and will lead in AI deployment.

Creemers argues that China’s biggest structural weakness remains the lack of trust from Global North countries.

At the same time, US export restrictions continue to leave Chinese companies short of advanced AI computing hardware, limiting their ability to rapidly expand inference capacity to meet growing demand.

Looking ahead, Creemers contends that the rise of China’s AI sector is part of a broader historical shift.

“We need to understand that, for the first time in modern history, we are facing a non-Western technology superpower,” he said.

In his view, the future will not be determined by a single global technological sphere, but rather by “a complex set of overlapping technology networks and processes.”