Over the years, China has gradually emerged as a new power in the field of artificial intelligence with a series of amazing achievements. ByteDance’s enterprise cloud service, Volcengine, has cut training time for its artificial intelligence (AI) image recognition model from 5 days to 3 days. Alibaba Cloud’s Sinian computing platform beat the record set by Google by recognizing 1,078 million images per second in offline scenarios. And the NF5488A5 server model from China’s largest server manufacturer Inspur has been hailed as a world-class product in the medical imaging, speech recognition and natural language processing segments.
But, none of China’s most powerful tech companies would have been possible without the powerful graphics processors provided by Nvidia Corp. Headquartered in Santa Clara, California, USA. These processors have played an important role in driving China’s advancement in AI, data analytics, and computing power.
And the US government’s abrupt decision last month to restrict Nvidia from selling its two most advanced chips, the A100 and H100, to customers in China caused a stir in the entire community. By all the fields of AI, cloud computing and smart cars in China will not be able to find a product to replace them immediately. And according to industry experts and technology analysts, Nvidia’s GPUs have long trained AI models for autonomous driving, semantic analysis, image recognition, weather variables, and more. big data analysis.
Nvidia won’t be able to sell some of its GPUs to Chinese customers in the future.
It is difficult to fully assess the impact that the latest Washington ban on Chinese industries. Nvidia says it’s trying to help minimize the impact on Chinese customers, such as replacing advanced GPUs with products that aren’t subject to US government licensing requirements.
Elsewhere, some Chinese GPU companies have expressed hope that the ban could give them a chance to win over old Nvidia customers as US-made chips are now out of stock. .
But it is not difficult to detect the psychological impact of this targeted US measure on Chinese policymakers and investors. At an important meeting this month, the Chinese President emphasized that it is necessary to accelerate the development of technologies important to national security and seek breakthroughs.
He Xiaopeng, founder of Xpeng, one of China’s largest electric vehicle manufacturers, commented on his social media account that a chip ban from Nvidia would be bad news for the driver training system. cloud-based automation. But that won’t affect Xpeng too much, as they already buy the cutting-edge chips they need in sufficient quantities for the next few years.
Industry dealers say a scramble for Nvidia chips from Chinese customers is expected. A Shanghai-based sales manager at VSTECS Holdings, China’s second-largest distributor of telecommunications equipment and cloud services, said all Chinese customers will start stockpiling the chips. affected before the ban officially took effect. Nvidia has said that it can continue to ship AI chips from its Hong Kong facility through September 2023.
“Most of the cloud service providers in China use Nvidia chips in their GPU-based products,” said the manager, who requested anonymity because he was not authorized to speak to the media. pine. “If this hardware is restricted from shipping to China, the supply of GPU-based cloud services will be in short supply and the impact will be huge.”
Alibaba Yitian 710 internal server processor uses ARM architecture.
Most of China’s cloud providers, including Alibaba Cloud, Tencent Cloud, and Baidu Cloud already use chips from other vendors, such as Intel and AMD, to power their infrastructure. their cloud. Many cloud servers rely on central processing units (CPUs), using Intel’s x86 or Arm-based architectures, to do the toughest jobs.
But Nvidia’s A100 chip, due out in 2020, and TSMC-made H100 – the chip is expected to be ready to ship in the second half of this year – has a particular advantage in handling parallel compute streams. . Compared to multi-core CPUs from Intel and AMD, Nvidia’s GPUs use thousands of cores, making them ideal for machine learning and deep learning AI models.
Nvidia’s data center business, which includes sales of two cutting-edge chips, was one of the fastest growing with $3.8 billion in revenue for the fiscal quarter ending July 2022. , with a strong annual growth rate of 61%.
As China has become a global AI powerhouse with nearly a million businesses claiming to be AI-related, the country has become a “huge market” for Nvidia, according to the founder and CEO at a conference call. the company’s important meeting this month.
China’s demand for chips is one reason Nvidia’s market capitalization is now about three times that of Intel’s. And this market also accounts for about a quarter of all sales at Nvidia, and the maximum third-quarter revenue loss from the latest US government ban is estimated at about $ 400 million, the company said. know.
While technically advanced GPUs can be replaced by a combination of less powerful GPUs for certain functions, the difference can be huge and the A100’s capabilities are already a boon. attractive selling position for Chinese customers. For example, both Alibaba Cloud and Baidu Cloud have incorporated the A100 chip into a large amount of their cloud infrastructure to power GPU-based cloud services.
According to a recent Reuters report based on government tender documents, Nvidia’s A100 chip has also been purchased by some of China’s most important strategic research institutes, such as Tsinghua University and Tsinghua University. Institute of Computer Technology at the Chinese Academy of Sciences.
Inspur, China’s leading server manufacturer, has also highlighted the inclusion of Nvidia’s A100 GPU in its products.
Nvidia’s H100 GPU is very sought after by Chinese businesses.
In a statement, the Nvidia CEO said that the company is looking for alternatives for Chinese customers. But Mr Huang also said the alternatives could be “worse for customer workloads that need the full performance of the A100”.
An Nvidia manager in China, who requested anonymity due to the sensitivity of the matter, said that Chinese AI companies will likely spend more time training their AI models, and this could slow the progress of the country’s AI sector. China has a lot of data coming from its 1.4 billion residents and it is still relying on powerful chips to improve its algorithms.
The US administration’s latest ban, coupled with previous restrictions, is expected to hit the industry’s weak point as domestic Chinese GPU makers have yet to be able to make any comparable products. Nvidia.
Lu Jianping, chief technology officer of Iluvator, one of China’s top GPU manufacturers, shared last week that the company’s GPUs are still far behind those of Nvidia’s. Lu, who worked for Nvidia and Samsung Electronics, said his company’s plan is to cover 5% of the market while Nvidia has the remaining 95%.
However, some domestic investors hope that the ban could accelerate China’s development of making its own GPUs, in line with the self-sufficiency plan in the fields. Beijing’s strategic technology.
Another Chinese GPU company, Biren Technology, founded in 2019, is now a prominent investor in GPU development. According to PitchBook, it raised about $633 million in its first 18 months and last month launched a GPU called the BR100. The company claims the product can break world records for computing power, but it has yet to be officially tested.
Wang Yu, a manager at Hengqin Financial Investment, a fund focused on semiconductors and hardware, said that the US ban could provide new reasons for investors to pour money into the GPU sector. But he said his company still sees some investment risk in the process. Because that would be a long step, even for Biren, to replace Nvidia.
“Chinese GPU startups are no match for Nvidia and AMD,” said Sravan Kundojjala, a senior analyst at Strategy Analytics. “Because even a seasoned player like AMD is struggling to compete with Nvidia.”
One of Nvidia’s main advantages is its unified device architecture, a parallel programming model, which the company began developing in 2006. It has been widely adopted by developers as an easy way easy to accelerate computer-intensive applications, according to Kundojjala. As such, those who want to compete with Nvidia have to create an entire software ecosystem, which is extremely difficult.
However, there are some who are willing to bet on China’s long-term ability to remove the influence of companies like Nvidia. Two days after Nvidia confirmed the US government’s plan to ban exports, an executive at a Shanghai-based GPU startup received a call from an investor who said he wanted more money. into the joint venture, according to a person familiar with the situation.
Refer to SCMP, BI