Chinese AI models close capability gap despite U.S. chip controls

Chinese artificial intelligence companies are rapidly closing the performance gap with U.S. frontier labs, even as American export controls restrict their access to cutting-edge computing hardware.

Clément Delangue, CEO of AI platform Hugging Face, told CNBC this week that China is "clearly dominating on open models right now" and could lead at the frontier level by year-end or 2028. His comments highlight a shift in the global AI landscape that has caught the attention of policymakers and industry leaders.

Beijing-based Moonshot released its Kimi K3 model in July, achieving benchmark scores that approached or exceeded leading systems from Anthropic and OpenAI in certain areas. The performance gains have translated into real-world adoption, with Western companies and developing economies increasingly turning to Chinese models as capable, cost-effective alternatives to American options.

Why it matters

The competitive dynamics in AI carry implications beyond technology. If Chinese AI systems become the default choice for developing nations, those countries may align more closely with Beijing politically while giving Chinese firms market footholds abroad. The question is no longer whether China can compete at the frontier, but whether the U.S. can maintain its lead across multiple dimensions including cost, deployment speed, and global reach.

Where China leads

China has established clear advantages in open-source AI models—systems that can be downloaded, modified, and self-hosted. The world's most capable open models currently come from Chinese labs. The country is also considered ahead in robotics and autonomous vehicle development.

"Based on current trends it seems more likely than not that Chinese AI will become the default for developing countries," Daniel Remler, senior fellow at the Center for a New American Security, told CNBC.

Chinese models have gained traction partly because most AI applications don't require frontier-level capabilities. Lower prices and strong performance in common use cases have driven adoption in American markets and reportedly across African nations.

The compute bottleneck

U.S. export controls have severely limited Chinese access to advanced AI chips, creating what may be the most significant barrier to China's AI ambitions. The restrictions affect both model training and inference—the process of running AI systems at scale.

Moonshot was forced to pause new subscriptions after demand for Kimi K3 exceeded its computing capacity, illustrating the practical constraints Chinese firms face.

Yet China is working around these limitations through multiple channels. Companies have been accused of accessing compute resources overseas, using techniques to extract knowledge from U.S. models, and smuggling Nvidia chips into the country. China's domestic semiconductor industry is also advancing, though it remains substantially behind American capabilities.

U.S. advantages persist

The United States maintains "the most capable models in the world, strong tech alliances and an overwhelming compute advantage," according to Keegan McBride, director of science and technology policy at the Tony Blair Institute for Global Change.

America's private capital ecosystem has enabled companies like Anthropic and OpenAI to raise unprecedented funding and scale rapidly. The country continues to attract top AI talent globally.

"If the United States can sustain these strengths, it will retain its AI advantage, serious geopolitical leverage, and ability to shape global AI rules," Remler said.

The details were first reported by CNBC.