Some companies to watch as China looks for a domestic alternative to Nvidia.

• Nvidia's graphics processing units (GPUs) have been the focus of attention, as they are essential for training large AI models, such as those developed by OpenAI.
• Beijing has intensified efforts to develop its domestic semiconductor industry in response to the U.S.'s attempts to limit China's access to advanced semiconductors.
• Nvidia alternatives are being developed by some of China's tech giants, including Huawei, Alibaba, and Baidu.

Beijing has intensified its efforts to develop its domestic semiconductor industry due to U.S. sanctions.

The growth of artificial intelligence and fundamental models has further motivated China's ambition to be a dominant player in the semiconductor industry.

The American company that has been making headlines is the one that produces graphics processing units (GPUs), which are essential for training large AI models, such as those used by OpenAI to create ChatGPT.

Washington has demonstrated its readiness to cut off Nvidia from the most advanced semiconductor chips and tools, prompting renewed interest in China's efforts to develop its own semiconductors to support its AI industry.

Two analysts identified some of Nvidia's leading Chinese competitors, as spoken to by CNBC.

Huawei

HiSilicon is Huawei's chip design unit, which is part of its business that encompasses telecommunications infrastructure, consumer electronics, and cloud computing, making it one of China's leading tech companies.

Huawei, a Shenzhen-based company, produces the Ascend series of data center processors, which it then sells as part of servers for use in data centers to train AI models under the brand name Atlas.

The Ascend 910C, the next generation of chip from the firm, is expected to be as powerful as Nvidia's H100 product, according to a Wall Street Journal report in August.

Nvidia named Huawei as a competitor in various fields such as chips, AI software, and networking products in their annual report this year.

The success of Huawei's Ascend series of datacenter processors is not solely dependent on hardware, but also on the overall ecosystem, developer tools, and the ability to continuously evolve this ecosystem as technology advances. According to Paul Triolo, an associate partner at consulting firm Albright Stonebridge, Huawei holds several advantages in this regard and is actively working to establish a software ecosystem around its Ascend processors.

Alibaba and Baidu

Nvidia chips are bought by both Baidu and for AI processes, but they also design their own semiconductors.

Kunlun, a brand owned by Baidu, is used for servers and autonomous cars that are designed by the Chinese internet giant.

T-Head, Alibaba's semiconductor design unit, has created an AI inference chip called the Hanguang 800. This chip is used in the real-world application of AI, such as a chatbot responding to user queries.

Alibaba's AI inference chip has been deployed to speed up its recommendation system on its e-commerce platform. Baidu has integrated its Kunlun chip into its data centers and autonomous driving sector, according to Wei Sun, a senior analyst at Counterpoint Research.

Biren Technology

Biren Technology, like Nvidia, creates a versatile GPU and offers a software development platform for creating applications on top of the hardware.

Biren's Bili series of products, including these chips, are intended for use in data centers for AI training.

Last year, Biren was placed on the U.S. Entity List, limiting its access to specific American technology.

Cambricon Technologies

Cambricon Technologies specializes in creating semiconductors for AI training and running AI applications on devices, rather than in data centers.

According to the South China Morning Post, the company has reported significant losses and allegedly laid off workers last year.

Cambricon Technologies is also on the U.S. Entity List.

Moore Threads

In 2020, Moore Threads was established with the goal of creating GPUs that can effectively train large AI models.

The company's data center product, MTT KUAE, is equipped with GPUs to help it achieve its goal of becoming a "global GPU leader," as stated on its website.

ByteDance, the owner of TikTok, is an investor alongside major venture capital firms such as Sequoia and GGV Capital.

Moore Threads is also on the U.S. Entity List.

Enflame Technology

In China, Enflame Technology is competing with other start-ups to become a domestic substitute for Nvidia. The company specializes in creating chips for data centers that are optimized for AI training and processing.

Enflame has Tencent, a major Chinese tech company, as an investor.