China’s artificial intelligence chip sector has attracted significant investor interest following a series of successful initial public offerings. But analysts say that while these newly listed companies have led the market rally, they are not leading China’s efforts to reduce dependence on Nvidia. Instead, most analysts point out that tech giant Huawei and its top-secret chip division, HiSilicon, stand out among the domestic competition not just for its technological advances, but also for its size and its extensive supply chain. However, Huawei has no plans to go public. The company, best known for its smartphone and communications businesses, has long maintained its intention to remain privately held, unlike many smaller chipmakers that have recently gone public. Huawei did not respond to a request for comment from CNBC. That stance limits options for mainland investors seeking exposure to China’s efforts to build a domestic alternative to Nvidia, currently the world’s most valuable company by market capitalization. Instead, investors are turning to smaller companies such as Biren Technology, MetaX, Moore Threads and Shanghai Iluvatar CoreX Semiconductor, which recently went public. Moore Threads, a graphics processing unit developer founded in 2020, raised more than $1.1 billion in its Dec. 5 public offering. Fellow GPU developer MetaX followed suit on December 17th, raising nearly $600 million. Chinese technology giant Baidu also recently announced plans to spin off its AI chip subsidiary Kunlun Xin and list it in Hong Kong. Huawei is likely to continue to dominate China’s AI processor and GPU market over other Nvidia alternatives. The reason is not just a single chip, but the stack. Principal analyst at Counterpoint Wei Sun Analysts said that despite the strong market performance of these companies, Huawei continues to dominate China’s AI processor market in terms of market share and product breadth. Bernstein analysts estimated in a report last month that Huawei and Nvidia are neck and neck in China, each accounting for about 40% of processor sales in dollar terms. Bernstein predicted that Nvidia’s share would shrink to 8% under export restrictions, but Huawei’s share could rise to 50% depending on policy changes. Cambricon is expected to hold about 9% of the market, with smaller players such as Hygon and MetaX trailing further behind. Analysts say the market is in a period of consolidation. “Semiconductor companies tend to consolidate with larger companies to invest in more advanced research and development,” said Qingyuan Lin, senior China semiconductor analyst at Bernstein. Full-stack Huawei has expanded its Ascend series of AI chips in recent years, and plans to release the next generation Ascend 950 in 2026. While these chips are still less powerful than Nvidia’s most advanced chips on a one-to-one performance basis, analysts said other innovations set the Shenzhen-based company apart in China. “Huawei is likely to continue to lead other Nvidia alternatives in China’s AI processor and GPU market,” said Wei Sun, principal analyst at Counterpoint. “The reason is not just the single chip, but the stack. That’s more important than the headline specs.” “Stack” refers to the multiple layers of the AI computing supply chain, spanning chip design, data center construction, and software development. Last year, Huawei released the “AI CloudMatrix 384” system, which links numerous cutting-edge Ascend chips to maximize performance. Analysts found that this cluster outperformed Nvidia’s leading systems at the time on some metrics. In addition to AI infrastructure, Huawei is also developing a CANN software layer to compete with Nvidia’s CUDA, which will directly compete across multiple layers of the AI ecosystem. Sun said these factors help give Huawei a clear lead, but it doesn’t mean Huawei is invincible. Other Chinese technology giants, including Baidu, are pursuing similar full-stack strategies. Production Constraints Many newly public Chinese AI chip companies, such as Moore Threads, are designers rather than manufacturers, so they rely on outside foundries to produce their chips. For example, NVIDIA relies on Taiwan Semiconductor Manufacturing Company. For Chinese chip designers, the closest equivalent to TSMC is Semiconductor Manufacturing International Corporation (SMIC). Analysts said SMIC’s limited production capacity was holding back output, and Huawei had received preferential access so far. “The SMIC landscape, where Huawei is being prioritized, perhaps under some direction from industrial planning authorities, leaves little room for other large GPU manufacturers such as Biren, MetaX, Enflame, Illuvatar and Sophgo,” said Paul Triolo, partner and senior vice president for China at DGA Group. “If some startups burn through their IPO proceeds, they could be held liable if they cannot produce enough at SMIC and slow market penetration of their products.” Remain Private Despite Huawei’s dominance in China’s AI chip space, analysts do not expect the company to go public anytime soon. But it’s not due to a lack of investor interest. “As a publicly traded company, Huawei will attract significant investor interest given its broad product line and significant lead in AI hardware and software development,” DGA Group’s Triolo said. But Huawei founder Ren Zhengfei has long preferred to maintain internal control over all aspects of the business rather than answering to shareholders, Counterpoint’s Sun said. Analysts say Huawei has less need to raise capital than smaller competitors given the profitability of its other businesses. The company has also benefited from the Chinese government’s support for the domestic semiconductor industry. In a 2021 letter to employees, Ren reportedly wrote that Huawei “may enter the market in stages in the future.” Huawei’s operating margin declined from 14.8% ($14.7 billion) in 2023 to 9.2% ($10.9 billion) in 2024. “In the post-ren era, it is not inconceivable for the company to consider going public,” Toriolo added. “However, given the success of Huawei’s communications and smartphone businesses, there have been no problems with revenue and R&D funding so far.”
