Chinese AI is at risk of falling behind due to a lack of American chips.
A US ban on Nvidia from shipping GPU chips to China could seriously damage the country’s most powerful artificial intelligence system.
In recent years, the field of artificial intelligence in China has achieved great success. Volsenjin’s ByteDance Cloud Service reduced the training time for an AI-based image recognition model from five to three days. Sinian, Alibaba’s cloud computing platform, has surpassed Google’s record by being able to identify 1,078 million people offline. Frames per second. The NF5488A5 from Inspur, China’s largest server manufacturer, is recognized as a world-class artificial intelligence product for medical imaging, speech recognition and natural language processing.
The above achievements are mainly thanks to the graphics chip provided by Nvidia. Based in Santa Clara, USA, the company plays an important role in the development of artificial intelligence in China, especially in data analysis and computing power.
AI wishes are affected
Asking Nvidia and AMD to stop exporting their best AI chips to China is seen as a major setback to Beijing’s AI ambitions. Nvidia can’t offer A100 and H100 without a license, making it difficult to work in areas like cloud computing and smart cars in China because there is no direct replacement.
Nvidia’s chips are currently used in AI training models for self-driving cars, semantic analysis, image recognition, weather detection, and big data analysis. follow up SCMPWhile it’s difficult to fully assess the ban’s impact on China’s tech industry, most experts believe the problem is serious.
follow up Reuters, the US also plans to extend the ban on AI chip supplies to China next month. In the face of the new ban, some Chinese GPU companies are offering to end contracts with former Nvidia customers in hopes they can produce their own products. However, the direct effects of the ban are easy to see.
“All Chinese customers will start stockpiling the affected chips before the ban officially takes effect,” said a sales manager at China’s second-largest cloud services and telecoms equipment provider. National VSTECS Holdings commented. Nvidia will probably not be able to deliver AI chips to China until 2023. September.
The director added, “Most cloud service providers in China use Nvidia chips in systems that require GPUs. If that hardware is limited, there will be a shortage of GPU-based cloud services. This will have a significant impact on the entire market.”
In fact, Chinese cloud providers like Alibaba’s Alibaba Cloud, Tencent’s Tencent Cloud, and Baidu’s Baidu Cloud still use some AMD or Intel chips. But Nvidia’s A100 and H100 are preferred for their ability to handle parallel streams of computation with large amounts of data. Compared to Intel and AMD processors with limited processing cores, NVIDIA GPUs use thousands of cores, making them ideal for machine learning and deep learning AI models.
China is now considered an AI powerhouse. Statistics show that by the middle of this year there will be almost a million companies in the country claiming to be involved in AI. “It’s a very big market,” NVIDIA CEO Jensen Huang said mid-year.
Some experts say companies can combine multiple processors to boost AI processing power. In this area, however, the GPU proves to be excellent and is therefore preferred. Both Alibaba Cloud and Baidu Cloud have integrated the A100 chip into most of their cloud infrastructure to provide GPU-based cloud services. Some of China’s leading strategic research institutes, such as Tsinghua University and the Institute of Computer Technology of the Chinese Academy of Sciences (CAS), use the chip.
If it is a total ban, Chinese companies can choose an alternative. However, Mr. Huang believes that these products can make the situation worse if they contain large amounts of data.
Another Nvidia executive leading the Chinese market said companies need to spend more time training AI models, which could slow the progress of their AI ambitions.
Lu Jianping, chief technology officer of Elevator, one of China’s leading GPU makers, admits that domestic GPU technology lags far behind American brands like Nvidia. “China’s domestic GPU maker has not been able to match Nvidia with similar products,” Jianping said. economic observer early last week.
On the other hand, many believe that the ban could accelerate the development of GPUs in the country, which has been one of Beijing’s technological autonomy strategies for years. Zhang Wen, Biren Technology – a GPU manufacturer founded in Shanghai in 2019, the founder is of this belief. follow up pitch bookThe company collected 633 million in the first 18 months. USD and BR100 GPU were introduced last month. “The BR100 can break the world record for computing power,” Wen said.
Wang Yu, head of Hengqin Financial Investments, which focuses on semiconductors and hardware, believes the US ban opens up opportunities for China’s self-government. But it’s still a “big risk that needs serious consideration.”
“Chinese GPU startups are no match for Nvidia and AMD. They even compete fiercely with each other,” said Sravan Kundojjala, senior analyst at Strategy Analytics.
Bao Lam (Consequences SCMP)