Alibaba, the Chinese e-commerce and cloud computing conglomerate, has opened a data centre in southern China powered entirely by its own artificial intelligence chips, marking a significant step in Beijing's drive to build homegrown AI infrastructure independent of Western technology.
The facility, located in Shaoguan in Guangdong province, was launched in partnership with China Telecom, the state-owned telecoms operator, which will own and run the site.
It houses 10,000 of Alibaba's Zhenwu semiconductors, which the company developed through its T-head chip design unit and which are built to handle both AI training, the process of building AI models, and inferencing, the process of running them once built.
The chips are capable of supporting AI models containing hundreds of billions of parameters, a measure of model complexity, placing them among the most demanding workloads in the industry.
The two companies said the facility is expected to eventually scale to 100,000 chips and could serve industries ranging from healthcare to advanced materials research.
The announcement underscores how China's largest technology companies have accelerated efforts to develop domestic chip alternatives following sustained US export restrictions that have progressively cut Chinese firms off from Nvidia's advanced AI semiconductors and other key Western components.
Last month, a separate computing cluster built using Huawei's Ascend 910C chips went online, pointing to growing momentum across the Chinese technology sector to establish sovereign AI hardware capacity.
Alibaba separately announced internal organisational changes designed to accelerate its AI development, with chief executive Eddie Wu forming a new technology committee that he will personally chair.
The committee will include the company's chief AI architect Zhou Jingren, the chief technology officer of Alibaba Cloud Li Feifei, and Alibaba Group's chief technology officer Wu Zeming.
The moves come as Chinese technology firms take a broadly different approach to AI investment compared with their US counterparts.
Related reading
- Google Finance expands to 100+ countries
- Meta urges Europe to scale artificial intelligence leadership
- Intel joins Musk's Terafab project to build one terawatt of AI compute capacity
While American hyperscalers, the largest cloud computing providers, are expected to collectively spend around $700 billion on AI infrastructure this year, Chinese companies have adopted a more targeted strategy, directing spending toward industry applications they believe will generate near-term returns rather than pursuing scale for its own sake.
Alibaba's cloud computing division, through which it sells AI services and models, has been one of its fastest-growing businesses in recent quarters.