On December 5, John Thornhill, the innovative content editor of the Financial Times, wrote on Thursday that,Compared with the lightweight, low-cost open model developed by China, the United States is pursuing a large-scale, closed-source model, which may be a wrong direction.


Popularity of China's open model exceeds that of the US model

The following is the full text of the article:

Last month, at a high-tech expo in Shenzhen, some technology influencers were surrounded by humanoid robots that could free-fight and play the piano. They couldn't help but wonder: Can the West still catch up with China?

This question might have sounded absurd twenty years ago, but it is anything but today. This week, the Australian Strategic Policy Institute (ASPI) released its latest key technology tracking report, covering research with globally important influence in 74 fields. The report found that China currently leads in 66 technologies in fields as diverse as computer vision, quantum sensors and nuclear energy, while the United States leads in another eight.

ASPI researchers found a familiar story across multiple technology fields: The United States achieved an early overwhelming advantage in scientific research output in the first decade of this century, but this advantage has been surpassed by China's long-term and sustained investment in basic research. In 2005, China accounted for just 6% of the world's most cited research papers, but this year this share has risen to 48%. The proportion of American papers fell from 43% to 9%. Jenny Wong-Leung, one of the study's authors, said the United States is defunding many federal science programs while China is conversely "building an entire technology ecosystem."

ASPI's findings are consistent with Nature's latest ranking of research institutions. The ranking tracks articles from 145 scientific journals. In terms of scientific research output, nine of the top ten research institutions in the world are from China, and only Harvard University in the United States ranks among the top. China is now "producing" scientific research results on a large scale, and it has truly become a scientific superpower.

However, published research results do not automatically translate into technical capabilities. Furthermore, the location of scientific expertise does not always correspond to technology commercialization success. Many long-frustrated British scientists know this all too well.

However, another report released earlier this year by the U.S. Special Competition Studies Program (SCSP) also highlighted that China has made significant progress in adopting many cutting-edge technologies. According to SCSP staff assessments, the United States remains the leader in semiconductors, synthetic biology, and quantum computing, while China dominates advanced batteries, 5G, and commercial drones.

AI battle

But the most competitive, and perhaps most important, area is AI.

U.S. President Trump has stated that the United States will lead the world in the field of AI “at all costs.” Large US technology companies such as OpenAI, Alphabet, Microsoft, Meta and Amazon are making huge investments to achieve this goal. OpenAI alone plans to invest $400 billion in the next few years to build its "Stargate" data centers across the United States.

Last month, the Trump administration launched the Genesis Mission to advance the private AI industry by sharing public data sets and computing resources from 17 U.S. national laboratories. "We are essentially pitting our private capitalists against China's state power," said SCSP senior adviser David Lin. "Both sides have two very different sets of resources, traits, strengths and weaknesses."

However, the United States and China have adopted completely different strategies in popularizing AI. Large American technology companies tend to develop large, private, "closed weight" models, such as ChatGPT and Gemini, which may be best suited to achieve the goal of general artificial intelligence (AGI). In contrast, Chinese AI companies prefer smaller, lower-cost "open weight" models, such as DeepSeek and Alibaba's Qianwen, which are easier for developers to quickly adapt. Part of this disparity stems from China's lack of access to advanced U.S. chips, turning a limitation into an advantage. This also reflects China’s development philosophy of prioritizing the rapid popularization of technology.

U.S. strategic mistakes

Michael Power, the former global strategist of Ninety One, an investment institution, believes that the United States’ heavy bet on giant closed AI models is a “catastrophic strategic mistake.”

Ball noted that “China’s model is proving to be much more effective in terms of real-world computing power available,” especially given China’s lower energy costs. Even OpenAI CEO Sam Altman expressed personal concerns, believing that “we may be on the wrong side of history.”

The latest research from MIT and Hugging Face shows that China’s open weight model has now surpassed similar American models in global application popularity. Many U.S. companies, such as Airbnb, have become fans of the “fast and cheap” Qianwen model. In this critical area, the same question resurfaces: Can the West catch up with China?