BEIJING: Chinese tech giant Baidu Inc. said Friday it plans to spin off its artificial intelligence chip subsidiary, Kunlunxin, and list it in Hong Kong, as domestic chipmakers seek capital amid Beijing’s drive for semiconductor independence.
The company announced it confidentially submitted a listing application to the Hong Kong Stock Exchange. Details of the offering, including its size and structure, have not been finalized. The proposed spin-off would require regulatory approvals, and Baidu cautioned there is no guarantee it will proceed.
Baidu, a leader in China’s AI sector, both uses and designs specialized chips. Kunlunxin, founded in 2012, is central to Baidu’s “full stack” AI strategy, covering hardware, software and services. While Baidu remains a major buyer of Nvidia chips, it increasingly uses its own Kunlunxin semiconductors in data centers powering its Ernie AI models.
The move aligns with a broader trend. Several Chinese chipmakers, including Moore Threads and Biren Technology, have recently announced listing plans. The push comes as U.S.-China tech tensions lead to restrictions on accessing advanced foreign chips and Beijing mobilizes public funds to develop domestic alternatives.
“The proposed spin-off would enable Kunlunxin to better attract specialist investors and expand financing options, highlighting its standalone market potential,” Baidu stated. The company owns approximately 59% of the unit, which would remain a subsidiary after listing.
Kunlunxin has evolved into a more independent entity, expanding sales to external customers. Reuters reported the unit’s revenue was projected to exceed 3.5 billion yuan ($500 million) last year, nearing break-even, with external sales expected to top half its revenue in 2025.
Analysts see Kunlunxin as a key player in China’s emerging AI chip ecosystem.
“In the market, Kunlunxin is seen as one of the most practical and widely used AI chips in China,” said Brady Wang, associate director at Counterpoint Research. He noted its software compatibility with common AI frameworks makes transitioning workloads from Nvidia easier.
Despite growth, analysts caution domestic chips cannot fully replace foreign ones yet due to manufacturing constraints. Kunlunxin’s strengths are seen in inference workloads and sectors like government and telecom, where supply stability and cost can outweigh peak performance needs.
“Beijing is not relying on a single company. Instead, Kunlunxin works together with Huawei Ascend, Cambricon, Alibaba and others to build a domestic AI computing ecosystem,” Wang said.
In its announcement, Baidu said the listing would help align management incentives with performance and elevate the unit’s market profile.