China’s top artificial intelligence researchers are saying the country is unlikely to overtake the United States anytime soon because of limited computing power, chip shortages, and gaps in foundational research. While long-term ambition remains strong, leading voices from Alibaba, Tencent, and Zhipu AI agree that the next three to five years will be especially challenging.
Lin Junyang, technical lead of Alibaba Group Holding’s Qwen AI team, recently stated that the probability of any Chinese company surpassing U.S. AI leaders such as OpenAI or Google DeepMind within the next three to five years is below 20 percent. He described this estimate as “highly optimistic,” pointing to the massive advantage the United States holds in computing infrastructure.
According to Lin, U.S. companies operate with computing capacity that is one to two orders of magnitude larger than what is available to most Chinese firms. “Most critically, OpenAI and others are pouring enormous computational resources into next-generation research,” he explained. “Meanwhile, in China, we are stretched to the limit just meeting daily demand, which already consumes most of our available compute.”
This imbalance affects not only performance but also the ability to experiment, take risks, and explore entirely new AI architectures.
Tang Jie, co-founder and chief AI scientist at Zhipu AI said that the gap between China and the U.S. may actually be widening. He said the U.S. has a major advantage because many of its most advanced large language models are not publicly released.
“We just released some open-source models, and some people feel excited, thinking Chinese models have surpassed the U.S.,” Tang said. “But the real answer is that the gap may actually be growing.”
Tang believes the strength of U.S. firms lies not only in hardware but also in their ability to quietly test and refine advanced models behind closed doors, giving them a strategic edge that is difficult to measure from public benchmarks alone.
Yao Shunyu, Tencent Holdings’ newly appointed chief AI scientist and a former OpenAI researcher, said the world’s leading AI company in three to five years could still be Chinese but only if several major obstacles are addressed.
Yao pointed to China’s success in scaling technologies such as electric vehicles and advanced manufacturing as evidence of its potential. However, he warned that AI leadership will require overcoming deep structural challenges.
These include the lack of domestically produced extreme ultraviolet lithography machines for advanced chip manufacturing, slower enterprise adoption of AI, and limited investment in foundational AI research.
“We are very good at optimizing within existing frameworks, extracting as much as possible from a small number of GPUs,” Yao said. “What is still missing is the risk-taking spirit to define the next paradigm.”
Although Washington approved the sale of Nvidia’s H200 chips to China late last year, reversing earlier export restrictions, the situation remains complex. Beijing later urged some Chinese technology companies to pause orders for U.S. chips as part of a broader push to replace foreign hardware with domestic alternatives.
Experts say this policy tension creates uncertainty for AI developers, who must balance short-term performance needs with long-term self-reliance goals. Limited access to chips and lithography tools continues to slow progress in training large-scale models.
