The history of Chinese exports has gone through two stages. The first were the 'old three'—goods such as furniture, clothing, and household appliances, whose production relied on cheap labor. This was followed by the emergence of the 'new three': electric vehicles, lithium batteries, and photovoltaic products, whose development was based on supply chain density.
Now the third stage is beginning, and its key elements are not raw materials, but so-called 'new new three': industrial and service robots, large language models of artificial intelligence, and innovative pharmaceutical preparations.
Recent export data and usage examples confirm this trend. Chinese robots for cleaning facades, capable of climbing walls, are replacing human spiders on high-rise buildings in Australia. Furthermore, large language models developed in China are used to service utility networks in Brazil. Fruitvatinib, discovered in Shanghai and sold worldwide under the name FRUZAQLA, has been included in authoritative global treatment guidelines.
It is important to note that none of these exported goods pass through container ports. The units of exchange are not tons, but model weights, API calls, and licensed molecules.
The export pattern has three noticeable characteristics. Firstly, it is a high value density per kilogram. According to the latest weekly data from OpenRouter on token volume, where four leading positions were held by Chinese models, it can be assumed that Chinese AI infrastructure is now measured by global traffic, not cargo manifests.
Secondly, it is the gravity of intellectual property. The 'new new three' are built on patents, model weights, drug development, and closed data loops. An example is the 'Xiaotian' robot dog from Hengzhi Future, used for industrial inspection. Another example is the portfolio of small molecules from Runsheng Pharmaceutical, which is based in Chongqing. Thus, the unit of export has transformed from equipment to know-how.
Thirdly, it is standard setting. China is no longer just waiting for international bodies to define standards and then adopting them. Now China is actively participating in AI working groups, pharmacopeia committees, and robotics safety standards bodies. This shift from passive follower to active definer is the most significant change in the narrative.
Nevertheless, there are serious obstacles. The US and Europe maintain control over the initial stages of production: the architectures of base models, the core components of industrial robots, and the primary targets for drugs, which cannot yet be fully replaced by domestic developments. Policies of friendly and reshoring trade create friction, and national drug regulatory systems still do not recognize each other's approvals.
Signals from OpenRouter traffic rankings and from examples of implementation abroad in Australia, Brazil, and Africa indicate progress. The speed of this growth will depend not so much on China's industrial policy as on the willingness of the rest of the world to accept Chinese innovations as a global standard. If the 'old three' asked the world to buy cheap goods, and the 'new three' asked for green ones, then the 'new new three' require buying underlying intelligence, which presents a much more complex sales challenge.
China's export strategy is undergoing significant changes. While the country previously focused on producing electric vehicles, batteries, and solar panels, its focus is now shifting to meeting global demand in the fields of artificial intelligence, robotics, and pharmaceuticals.
Chinese artificial intelligence models have been the most sought-after globally for twelve consecutive weeks. Furthermore, the number of countries using surgical robots has increased from 23 to 49. The pharmaceutical industry worldwide paid $110 billion for licenses to Chinese medicines over the last six months.
Thus, China is not limited to simply exporting goods; the country is beginning to export its own ideas.
