Who Will Win the US-China AI Battle? Tang Hsiang-lung Reveals 'America Facing Difficulties': Chinese Models Surpassing US Is Real

Data from OpenRouter, an AI model aggregation platform, shows that Chinese AI models have surpassed their US counterparts in usage volume (measured in tokens), with the gap continuing to widen. Since March, Chinese models such as Qwen, DeepSeek, GLM, MiniMax, and Moonshot have overtaken US models on the platform, indicating a rising competitiveness of China's open-source AI ecosystem.

Senior media commentator Tang Hsiang-lung analyzed this trend on the program 'East, West, North, South: Dragon and Phoenix Match,' stating that prior to March, many US enterprises and investment institutions dismissed the idea of Chinese models surpassing US ones as mere market 'rumors' or noise. However, after March, they realized this shift was real.

Tang emphasized that in the 'AI war' between the US and China, China treats AI as a public good for the common good. By enabling domestic enterprises—especially those lacking development capabilities or financial resources—to access homegrown AI, China reduces its dependence on US-controlled technologies. This strategy effectively counters the US capitalist AI industry's ability to impose price hikes or sanctions.

He pointed out that the US government's inconsistent AI policies reflect its difficulties in this competition. Some US AI companies are lobbying the White House to ban Chinese open-source models. However, Tang questioned the effectiveness of such bans, citing Huawei's resurgence in international markets after US sanctions.

Tang predicted that high-level AI talks between the US and China could occur in September, possibly involving US Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng. These discussions may address AI model regulations and global governance.

He believes AI will soon end the era of traditional search engines, dismantling the US's historical dominance in digital penetration. Chinese AI models are already demonstrating rapid and remarkable market penetration. Despite earlier skepticism—given that less than 5% of global data is in Chinese—Tang stressed that AI models improve with use, making market share critical. With Chinese models being open-source and more affordable, they offer lower token costs, directly enhancing enterprise competitiveness in the long run.