China’s AI industry looks unstoppable in the race to best US rivals. But is it?

China’s AI Leaders Assess Prospects in Global Competition

In January, China’s leading artificial intelligence entities convened in Beijing for a pivotal gathering, sparking debate over the likelihood of a Chinese AI firm surpassing U.S. pioneers within the next three to five years. A top AI scientist at the event offered a candid perspective, stating:

“Below 20 percent,” said Justin Lin, technical lead for Alibaba’s Qwen AI models. “And I think 20 percent is already very optimistic.”

This cautious outlook contrasted sharply with a year marked by headlines extolling China’s AI surge. Following the unexpected rise of DeepSeek, a lesser-known startup, with its cost-effective AI model, Chinese companies dominated global downloads for open-source models and achieved significant fundraising in initial public offerings. Yet, amid the celebration, some experts cautioned that China might have widened its gap in developing cutting-edge AI models.

Restrictions on access to advanced chips and limited capital have persisted as key challenges. Tang Jie, co-founder of Z.ai (Zhipu), echoed this concern during the same meeting, asserting:

“The performance gap between Chinese and U.S. models may be widening. In some areas we may be doing fairly well, but we also need to acknowledge the challenges and gaps we still face,”

he noted.

Despite these hurdles, China’s AI sector continues to evolve. Constraints on high-performance chips and capital have driven a distinct approach, emphasizing open-source models to foster public engagement and competition with U.S. counterparts. This strategy has enabled Chinese firms to make measurable strides, with applications emerging across manufacturing, e-commerce, and robotics.

Chinese leader Xi Jinping highlighted the nation’s rapid innovation in a televised New Year address, praising AI models that “are racing ahead” and noting “breakthroughs” in domestically developed chips as part of Beijing’s push for technological self-reliance. Qwen, for instance, surpassed Meta’s Llama as the most-downloaded open model on Hugging Face last September. Even U.S. companies like Airbnb have adopted these models for AI-driven customer service.

An unprecedented wave of AI listings unfolded in Hong Kong, with unicorns Z.ai and MiniMax, developers of competitive open models, going public in January. The companies raised $560 million and $620 million respectively, seeing sharp share price increases. Global attention is growing, as evidenced by Meta’s December announcement to acquire Manus, a Chinese AI agent firm now based in Singapore. While the deal has triggered…