OpenAI recently surprised users by announcing that, starting July 9, 2024, it will cease providing API services to China and other regions, including Hong Kong.
This abrupt decision spells trouble for Chinese startups reliant on OpenAI's API, forcing them to seek alternative solutions for product development. As a result, Chinese AI companies now have a prime opportunity to capture market share.
Bloomberg and South China Morning Post report that this move comes as the US ramps up tech sanctions on China. The US Treasury Department recently released a draft to regulate American investments in Chinese sectors like semiconductors, AI, quantum computing, and microelectronics, which could threaten national security, requiring future investments to be notified or prohibited.
Chinese media Securities Times speculates that OpenAI's decision may be linked to the recent addition of NSA Director Paul M. Nakasone to OpenAI's board and security committee.
Xiaohu (Neil) Zhu, founder and chief scientist of University AI, which provides training for Chinese enterprises, sees this as a direct consequence of the US-China tech rivalry. Previous US regulations limiting Chinese access to semiconductor and AI technologies hinted at this development, making OpenAI's API suspension unsurprising.
Singapore-based Dorje AI underscores the need to monitor whether other open-source large language models (LLMs) like Meta Llama will also cut off access for Chinese developers. In July 2023, China introduced interim measures for managing generative AI services, requiring regulatory approval before consumer-oriented generative AI LLMs can be sold in China to control public opinion and social mobilization.
By March 2024, over 200 domestic LLMs had applied for approval, with 117 receiving the green light from Chinese authorities, and no foreign products among them. Consequently, ChatGPT remains inaccessible in China.
For Chinese developers starting from scratch with limited funds, OpenAI's technology is a crucial tool for training their own AI systems and building applications, integrating its API into their products for sale.
In China, there are two ways to use OpenAI's API: purchasing Microsoft's Azure cloud services or using overseas servers to relay requests, though the latter suffers from stability issues and requires additional servers. For OpenAI, losing the Chinese market means missing out on API access revenue, which is charged based on token consumption.
Chinese internet company Qihoo 360 suggests that OpenAI's exit from the Chinese market will push users towards local models, accelerating the growth of China's LLM industry. The Center for Regulation and Global Governance at the Chinese University of Hong Kong (Shenzhen) also indicates that leading Chinese LLM providers will benefit, helping eliminate small, inefficient players.
Companies like Zhipu, Tencent, Baidu, Alibaba, Kai-Fu Lee's 01.AI, SenseTime, and Baichuan are seizing this opportunity with "migration" plans and model discount schemes. Zhipu, for instance, has announced a special migration plan, offering training, consulting services, and tokens equivalent to what developers used with OpenAI. Baichuan Intelligence promises that migrating from OpenAI to their platform will take only five minutes. According to iiMedia Research, the quick response from Chinese AI firms reflects the increasingly fierce competition in the local industry.
Since ChatGPT's debut at the end of 2022, Chinese companies have launched several LLMs amid the AI boom. Dorje AI notes that OpenAI's move may eliminate many small AI startups that emerged during the frenzy.
In the long run, lacking access to global tools will hinder the development of Chinese AI companies. Alibaba has stated that Chinese AI models will need at least two years to catch up with the US. Furthermore, Chinese tech companies and startups might relocate overseas for a more stable and growth-friendly business environment.
