Traffic monopolisation by internet giants impedes China’s AI development

I have mentioned in my previous articles in this series that China’s current wave of artificial intelligence (AI) startups is facing a chip chokehold in the underlying technology of large language models and a lack of capital at the financial level. On the application side of large language models, there is another major difficulty in the short term: the lack of traffic.

China’s internet traffic is mainly concentrated in the application end, with internet giants WeChat and Douyin occupying the lion’s share. It can even be said with no exaggeration that China’s internet ecosystem is basically made up of these two applications, making it almost impossible for entrepreneurs to have a share in this market.

Independent app developers are challenged by a lack of traffic, coupled with the fact that these two platforms block external links. Once traffic enters either of these two platforms, it is as if the users have fallen into a black hole and become difficult to convert.

For example, say a startup has developed an application used to generate educational powerpoints. In the US, through search engine optimisation on Google, users searching for such applications will directly see these newly developed products. A paid traffic campaign can be launched, or the product could even be recommended to a huge number of organic consumers because of its good reputation.

But in China, as Douyin and WeChat block external links, applications would have to resort to paid advertising to gain users. However, this promotional model often fails to attract users organically, burdening the startup team with huge user acquisition costs. In the current landscape, internet giants continue to limit the viability of software applications.

The reason why the US market has been able to produce numerous successful large language model applications is because of the abundance and freedom of web traffic.

Businesses are able to continuously generate free web traffic just by developing web applications and using search engine optimisation. At the same time, enterprises can also direct traffic to their own websites through external links on foreign social media platforms such as X and Facebook. Thus, various startups are able to thrive in a relatively relaxed market environment. In particular, the rapid launch of innovative AI products on the web has also further propelled web traffic growth.

In the process of regulating the industry, the Chinese government had intervened in traffic monopolisation by internet giants multiple times to open them up. Consequently, there is speculation about whether Chinese internet giants will voluntarily open up in the near future or if the government will compel them to do so. Many industry practitioners anticipate that these changes will occur shortly.

In addition to US law and its internet traditions, a more important reason why US internet giants are actively opening up traffic to AI startups is that the former is not facing a chip chokehold. Thus, they are focusing on building their own underlying large language model ecosystems and do not need to direct their energy to developing applications. In fact, the thriving applications market in turn drives the development of US internet giants.

Meanwhile, China’s internet giants and the entire information sector are still developing their own AI chips and related ecosystems. At this juncture, the prosperity of large language model applications will essentially only boost the AI ecosystem led by Nvidia and others.

Thus, the Chinese government and China’s internet giants are not placing a priority on fostering the startup ecosystem for AI applications by liberating traffic. This will only follow once China’s information industry successfully addresses the underlying chip ecosystem problem.

Related: Risk averse: China’s private capital unmotivated to fund disruptive innovation applications | China’s AI race is not about expanding users | With Sora, is China’s AI falling further behind?