A general view of the Baidu logo is seen at the Shanghai New Expo Center during the World Artificial Intelligence Conference 2025 in Shanghai, China, on July 28, 2025.
Ying Tang | Nurphoto | Getty Images
Chinese tech giant Baidu plans to spin off its artificial intelligence chip subsidiary, Kunlunxin, and list it in Hong Kong, as more domestic chipmakers seek funds amid Beijing’s push for semiconductor self-sufficiency.
The company said in an announcement Friday that it had confidentially filed a listing application on the Hong Kong Stock Exchange, though details of the offering, including size and structure, remain undecided.
The move would still require regulatory approvals, including from China’s securities watchdog. Baidu emphasized there is no guarantee the spin-off will proceed. The company reportedly owns about 59% of Kunlunxin.
Baidu, a major player in China’s growing AI space, is both a buyer of specialized AI chips for data centers and cloud computing, as well as a designer of them through Kunlunxin.
The firm said that the spin-off would align with its strategy to highlight Kunlunxin’s standalone potential, attract sector-specific investors, and expand financing options. Kunlunxin would remain a Baidu subsidiary, it added.
The move comes against a backdrop of intensifying U.S.-China tech tensions. Both Washington and Beijing have imposed various restrictions on Chinese AI companies’ access to leading-edge AI chips from California-based Nvidia.
Meanwhile, Beijing has increasingly encouraged domestic chip purchases and mobilized billions in public funds towards development.
In recent months, several Chinese chipmakers have announced plans to list, including Moore Threads and Biren Technology.
A Growing Business Focus
Founded in 2012, Kunlunxin is central to Baidu’s ambition to become a “full stack” AI company, spanning hardware, servers and data centers, as well as AI models and applications.
While Baidu still relies heavily on Nvidia’s chips for AI computing power, Kunlunxin has enabled the company to increasingly use a mix of its self-developed chips in data centers running its Ernie AI models.
Kunlunxin has also shifted to operate as a separate entity, expanding its sales to third-party customers outside Baidu.
“In the market, Kunlunxin is seen as one of the most practical and widely used AI chips in China,” Brady Wang, associate director at Counterpoint Research, told CNBC.
He added that one of the chipmaker’s main strengths is in software. “Instead of forcing users to adopt a closed system, Kunlunxin works well with common AI frameworks and makes it easier to move workloads from [Nvidia].”
Reuters previously reported that Kunlunxin’s revenue is projected to exceed 3.5 billion yuan ($500 million) last year, reaching break-even. External sales were expected to account for more than half of its revenue in 2025, the report added.
In another sign of strength last year, Kunlunxin won orders worth over 1 billion yuan from suppliers to China Mobile, one of the country’s biggest mobile carriers.
China Mobile also participated in the entity’s latest funding, which had raised over 2 billion yuan and valued the unit at about 21 billion yuan, according to Reuters.
In its announcement, Baidu said its plans to spin off and list Kunlunxin would better tie management incentives with performance and elevate the unit’s market presence.
Late last year, JPMorgan analysts forecast that Kunlunxin’s chip sales would increase sixfold to 8 billion Chinese yuan in 2026.
However, while Kunlunxin may help reduce China’s reliance on chips from Nvidia, it cannot fully replace them, Counterpoint’s Wang said, citing Beijing’s ongoing constraints in advanced chip manufacturing.
“[Kunlunxin’s chips] work best for inference and other workloads that are easier to move, especially for government, telecom, and state-owned cloud users, where stable supply and lower cost matter more than top performance,” he said.
“Because of this, 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.”
