DeepSeek said last month that its latest artificial intelligence model is compatible with domestic AI chips that would be released soon, unleashing a surge of interest in a handful of Chinese chipmakers.
Startup chip designer Cambricon became China’s most expensive stock by price-earnings ratio as investors bet that U.S. export controls would force China to fast-track development of its own chip ecosystem to reduce dependence on Nvidia. A group of Chinese chip and AI model companies recently announced a “chip-model” alliance to push for greater self-sufficiency, and several AI chip designers are preparing initial public offerings to tap investor interest.
DeepSeek’s announcement confirmed that Chinese AI developers are able to engineer models in a way that may not require Nvidia’s most advanced graphics processing units (GPUs). Nvidia’s latest earnings forecast assumed no H20 chip sales to China. For small chip designers in China, the uncertainty over access to Nvidia chip supplies creates a major opportunity.
But which Chinese company, or companies, will fill the Nvidia void is still unclear. HuaweiiHuaweiHuawei is a Chinese technology company focused on mobile phones and telecommunications, and is seen as a poster child for China’s global tech ambitions.READ MORE is the current market leader. Many other chip designers are struggling with high research and development costs, a small customer base of mostly state-owned enterprises, U.S. blacklisting, and limited chip fabrication capacities.
“The success of the domestic AI chipmakers will hinge on their success in terms of securing commercial partners, aside from state-backed companies,” Ray Wang, a semiconductor analyst at research company The Futurum Group, told Rest of World.
“Globally for GPUs, you really just have two biggest winners, and all the other startups are struggling or have limited market shares. Something similar will happen in China — maybe the top three companies can survive,” he said.
Here are some of China’s hottest Nvidia wannabes:
Cambricon was founded in 2016 by two brothers, Chen Yunji and Chen Tianshi. They both studied at a prestigious youth program at the University of Science and Technology of China, known for producing science prodigies.
The older Chen, now 42, left Cambricon to work as a researcher at the state-run research institute Chinese Academy of Sciences, while the younger Chen, 40, stayed on as chairman of the company.
After Cambricon listed in Shanghai in 2020, it earned a large part of its revenue from a few major government-backed data centers, while recording large losses. In 2024, it launched its 7-nanometer AI chip Siyuan 590, modeled after Nvidia’s A100, and turned profitable for the first time.
In 2022, Cambricon was added to the U.S. Entity List, which names persons or companies believed to pose national security risks. This effectively blocked it from placing orders with the Taiwan Semiconductor Manufacturing Company (TSMC).
Beijing-based Moore Threads was founded in 2020 by James Zhang Jianzhong, who previously served as Nvidia’s global vice president and China general manager. Two other co-founders also worked at Nvidia earlier.
Moore Threads was added to the U.S. Entity List in 2023. Despite the restrictions, the company said it had built GPUs that matched the performance of Nvidia’s older-generation gaming chips RTX 3090 and RTX 4090. Moore Threads said this year it could run DeepSeek and Qwen models on its own GPUs.
The loss-making company filed for an IPO in Shanghai in July and said it planned to raise about 8 billion yuan ($1.1 billion).
Shanghai-based Biren was founded in 2019 by Zhang Wen, a Harvard-educated former president of AI company SenseTime. Two other co-founders, with experience at Qualcomm, Nvidia, and Advanced Micro Devices (AMD), have since left the company.
In 2022, Biren unveiled its BR100 GPU, which it said matched the performance of Nvidia’s H100 AI chip. It was added to the U.S. Entity List in 2023.
Biren was valued at approximately 14 billion yuan ($2 billion) before a funding round in June. The loss-making company plans to list in Hong Kong.
MetaX, headquartered in Shanghai, was founded in 2020 by former AMD executive Chen Weiliang. It recently unveiled a new C600 chip, designed to support AI training and inference.
MetaX has suffered from production shortages in the past. After running out of its most advanced GPU, MetaX, along with chip startup Enflame, had to downgrade its designs so they could still get chips produced by TSMC.
MetaX is still losing money and has a small number of customers. It plans to raise 3.9 billion yuan ($545 million) in a Shanghai IPO.
Enflame was founded in 2018 by former AMD executives Zhao Lidong and Zhang Yalin. The Tencent-backed company has supplied chips to several local data centers. It is not on the U.S. trade blacklist, so has access to TSMC. It also plans to go public.
Hygon’s chip business started in 2016, when its Chinese parent company formed a joint venture with AMD and licensed the American company’s key x86 chip design architecture.
In 2019, Hygon was added to the U.S. Entity List. AMD halted its support, and Hygon said it was able to continue with its own technology. Hygon now makes central processing units (CPUs) and general-purpose GPUs used in AI.
While Huawei is the leader, Chinese companies don’t want to rely entirely on the company. Other tech giants have their own chip divisions.
BaiduiBaiduBaidu is a Chinese technology company that operates the country’s biggest search engine and video-streaming service iQiyi.READ MORE in April switched on a computing cluster with 30,000 of its own Kunlun chips, which it said could train DeepSeek-like models. Baidu’s chip business recently won a $139 million order for AI chips from state-owned telecoms giant China Mobile.
AlibabaiAlibabaAlibaba, founded in 1999 by Chinese entrepreneur Jack Ma, is one of the most prominent global e-commerce companies that operates platforms like AliExpress, Taobao, and Tmall.READ MORE’s chip unit, called T-Head, has been developing chips with the open-source RISC-V architecture. In March it launched a new XuanTie C930 chip, designed for AI and other high-performance computing use.