DayOne secures $237m bank loan for Hong Kong data center

A significant stockpile of approximately 600,000 Nvidia H200 chips is currently stored at a Hong Kong port, awaiting approval from Chinese authorities before being transported to the mainland. Despite strong demand from Chinese technology firms, Beijing has yet to grant permission for large-scale imports, citing its goal to reduce reliance on U.S. technology.

While Chinese companies still require Nvidia's advanced computing power to cater to the growing need for AI services, this has created a tension between the government's long-term technological aspirations and the industry's immediate requirements. In recent months, China has permitted limited quantities of H200 processors into the mainland, with several major Chinese internet firms each acquiring around 10,000 chips.

Smaller quantities have also entered through intermediaries and indirect methods. The H200, part of Nvidia's Hopper generation, offers capabilities that can't be matched by Chinese alternatives, especially for AI model training. However, Chinese officials reportedly prefer companies to prioritize domestic processors for inference tasks while allowing limited H200 imports for AI training.

Despite the availability of Huawei's Ascend series of AI processors, which are aimed at competing with Nvidia's data center chips, Chinese companies continue to seek Nvidia hardware due to the software ecosystem's value for developers building and training advanced AI models. The U.S. export controls further complicate the situation, as they restrict China's access to advanced semiconductor technology.

The stockpile in Hong Kong underscores China's reluctance to deepen its dependence on U.S. technology, while also meeting the immediate requirements of the industry.