Anamika Dey, editor
Brief news
- Chinese companies are striving to create alternatives to Nvidia’s AI processors amid U.S. sanctions, which hinder their technological advancements and access to essential manufacturing resources.
- Major players like Huawei, Alibaba, and Baidu are competing but lag behind Nvidia, facing challenges in GPU production and software ecosystem development due to restrictions on advanced chip manufacturing.
- Startups like Biren and Enflame are optimistic, seeking IPOs to secure funding for growth, despite experiencing layoffs and financial constraints compared to larger competitors like Huawei.
Detailed news
Chinese enterprises are intensifying initiatives to develop a feasible substitute for Nvidia’s processors that drive artificial intelligence, as Beijing persists in its endeavors to reduce reliance on American technology.
Recent U.S. sanctions imposed on China, along with Nvidia’s preeminence in the sector, have posed significant obstacles to Beijing’s initiatives, at least in the medium term, experts informed CNBC.
Nvidia’s notable growth has been propelled by significant cloud computing enterprises purchasing its server equipment equipped with its graphics processing units, or GPUs. These chips help organizations, such OpenAI, the creator of ChatGPT, in training their extensive AI models on vast datasets.
These AI models are essential for applications such as chatbots and other developing AI technologies.
Since 2022, the U.S. government has imposed limits on the sale of Nvidia’s most sophisticated processors to China, with these restrictions becoming more stringent last year.
These chips are crucial to China’s aspirations to emerge as a dominant force in artificial intelligence.
CNBC consulted analysts who identified some of China’s prominent challengers to Nvidia, including technological giants Huawei, Alibaba, and Baidu, as well as startups such as Biren technological and Enflame.
The prevailing perspective is that they are now trailing Nvidia.
Wei Sun, a senior analyst at Counterpoint Research, said CNBC that “these companies have achieved significant advancements in the development of application-specific integrated circuits (ASICs) for artificial intelligence.”
“Nonetheless, competing with Nvidia continues to pose significant challenges due to technological disparities, particularly in general-purpose GPUs.” It is improbable to match Nvidia in the short run.
China’s principal challenges
According to Sun, Chinese enterprises have a deficiency in technological skills, underscoring a significant difficulty.
Nonetheless, it is the U.S. sanctions and their subsequent repercussions that present the most significant obstacles to China’s aspirations.
Several prominent Nvidia competitors in China have been included to the U.S. Entity List, a designation that limits them access to American technology. Simultaneously, several U.S. restrictions have prohibited the sale of essential AI-related chips and gear to China.
All GPU manufacturers in China design their chips and depend on a fabrication business for production. Initially, this would have referred to Taiwan Semiconductor Manufacturing Company, or TSMC. However, U.S. laws prevent many of these companies from accessing chips produced by TSMC.
Consequently, they must rely on SMIC, China’s largest chip manufacturer, whose technology lags several generations behind TSMC. One reason for its lag is Washington’s restriction on SMIC’s access to essential gear from the Dutch business ASML, necessary for producing the most advanced chips.
Simultaneously, Huawei has been advancing the creation of more sophisticated processors for its smartphones and AI applications, therefore occupying capacity at SMIC, as stated by Paul Triolo, a partner at the consulting company Albright Stonebridge.
“The primary constraint will be domestic foundry leader SMIC, which faces the intricate challenge of allocating finite resources for its advanced node production among Huawei, currently the predominant beneficiary, GPU startups, and numerous other Chinese design firms that have been or may be restricted from utilizing global foundry leader TSMC for the fabrication of their advanced designs,” Triolo informed CNBC.
Nvidia encompasses more than only graphics processing units.
Nvidia has achieved success owing to its sophisticated semiconductors and its CUDA software platform, which enables developers to design programs compatible with the U.S. chipmaker’s technology. This has resulted in the creation of an ecosystem surrounding Nvidia’s goods that others may find challenging to duplicate.
“This is crucial; it encompasses not only the hardware but also the comprehensive ecosystem, developer tools, and the capacity for ongoing evolution of this ecosystem as technology progresses,” Triolo stated.
Huawei is in the forefront.
Triolo recognized Huawei as a prominent player in China with its Ascend line of data center CPUs.
The company’s latest processor is the Ascend 910B, and it is preparing to release the Ascend 910C, which may rival Nvidia’s H100 device, as reported by the Wall Street Journal in August.
Nvidia’s annual report earlier this year expressly designated Huawei, with other firms, as a rival in sectors including chips, AI software, and networking devices.
According to Triolo, Huawei possesses several advantages in the realm of software development and cultivating a developer community. However, it has comparable obstacles to the remainder of the industry in its efforts to compete with Nvidia.
“The GPU software support ecosystem is predominantly centered on Nvidia, with AMD playing a secondary role. Huawei encounters significant obstacles in both the production of advanced GPUs, such as the Ascend 910C, and in the ongoing innovation and enhancement of hardware performance, particularly due to U.S. export restrictions that hinder SMIC’s capacity to manufacture advanced semiconductors,” Triolo stated.
Upcoming Chip Initial Public Offerings?
The difficulties confronting China’s Nvidia rivals have been apparent over the last two years. In 2022, Biren Technology implemented a series of layoffs, subsequently followed by Moore Threads the following year, with both firms attributing the actions to U.S. sanctions.
Startups remain optimistic, seeking to secure funding to achieve their objectives. Bloomberg reported last week that Enflame and Biren are both seeking to conduct initial public offerings to attract capital.
Triolo stated, “Biren and other GPU startups employ seasoned professionals from Nvidia, AMD, and other prominent western semiconductor firms; however, they face the additional obstacle of insufficient financial resources compared to Huawei.”
Biren and Enflame are pursuing IPOs in Hong Kong to provide funds for further recruitment and growth.
Source : CNBC News
