Nvidia, AMD exported high -end GPUs to China to be restricted to the United States request for a new license to apply for a new license

21st Century Business Herald Reporter Ni Yuqing Shenzhen Report

Due to the new orders of US export permits, Nvidia and AMD's high -end GPUs will be restricted to China.

On August 31, Nvidia issued an announcement saying that on August 26, the United States had notified companies to export A100 and H100 chips to China and Russia to require new license requirements. At the same time, DGX (AI system product portfolios) or any other products including A100 or H100 chips, as well as chips with higher future performance than A100, will be controlled by new regulations.

This means that there is no American permission, Nvidia will suspend the providing A100 and H100 related products for the Chinese market. These two products are professional CPUs, data centers and high -performance computing scenarios, not PC and other consumer terminal equipment. Essence

Nvidia's announcement also stated that the new license requirement may have an impact on the company's timely completion of H100 development and support existing customers in A100, and may require the company to transfer certain businesses to China. At present, Nvidia is seeking exemption from the US government on the one hand, and on the other hand, it is communicating with Chinese customers to seek products other than the use of new rules to meet customer needs.

Nvidia predicts that in the third quarter, the potential sales of Chinese business restricted by the new license are about $ 400 million. In addition to Nvidia, a spokesman for AMD said that AMD also received a relevant permit ban.

At a press conference on the Ministry of Commerce on September 1, a spokesman for the Ministry of Commerce, Shu Jingsing, said that for a period of time, the United States has continuously abused export control measures to restrict the export of semiconductor -related items to China, and China has resolutely opposed it. The relevant practices of the United States depart from the principle of fair competition, violate the rules of international economic and trade, not only harm the legitimate rights and interests of Chinese enterprises, but also seriously affect the interests of American companies, hinder international scientific and technological exchanges and economic and trade cooperation, and cause the global industrial chain supply chain to stabilize and the world economy recovery Intellectual.

Nvidia's computing power ecology

Specifically, the A100 is the GPU released by Nvidia in 2020. It adopts 7 nanometer process and consists of 54 billion crystal tubes to support AI reasoning and training. H100 is a new product released in March this year. The crystal tube, and the new architecture Hopper is launched in the third quarter of this year.

In recent years, GPUs, especially in the field of AI applications, are riding dust. In the process of deep learning, the characteristics of convolutional operations and parallel operations provided by GPU can meet the needs of AI computing. Under the trend of global cloudization, the company needs GPU to provide computing power support behind the cloud and calculations.

Nvidia even seized the key time for the needs of computing equipment. From graphic rendering, to artificial intelligence, and even the blockchain field, the calculation of the blockchain was very prominent, and the application scenarios also covered PC, data centers, automobiles and other markets.

At present, in the global GPU market, the main players are Nvidia and AMD, and Intel has also developed GPU product lines in recent years. Nvidia still occupies the main market share, and as early as 2007, Nvidia proposed the concept of GPGPU (GPU for general computing), which extensively applied GPUs from the traditional impression of image processors to computing training. More importantly, Nvidda launched its own CUDA platform to build a strong AI computing ecology.

CUDA is Nvidia's parallel computing platform and programming model. CUDA can greatly improve the calculation performance by using the GPU's processing capacity to allow GPUs to have the ability to solve complex computing problems. It can be said that the CUDA platform is a complete set of software and hardware ecological standards established by Nvidia. Most AI chips are compatible with it, especially the training end chip.

Therefore, although the GPU or AI startups are emerging endlessly, most of them have the CUDA platform of Rong Yingwei, and they have to create their own ecological barriers and costs. The obstruction of the GPU supply will also affect the computing power applications such as servers and data centers. Regardless of the industry leader or SMEs, GPU products are inseparable from the trend of digitalization.

GPU companies in China

On the other hand, under the transformation of the external environment in recent years, domestic manufacturers are also accelerating the research and development of the GPU field. At the product level, the emerging GPU companies also frequently target Ying Weida, but still facing the challenges of giants.

From the corporate side, related companies that have already been listed include Cambrian, Sibyuan shares, Longxinzhongke, Jingjiawei, etc. Among them, Xinyuan shares are IP companies, and Longxin Zhongke has a self -developed architecture. Jingjiawei was founded in 2006. At the same time, it was also founded by the core movement technology in 2007. They are all representatives of the domestic GPU industry chain.

同时，近几年新一波的GPU和AI热潮中也涌现了一批初创企业，包括壁仞科技、沐曦、摩尔线程、天数智芯、燧原、瀚博、后摩智能、芯瞳、 Debut technology and so on. Among these emerging companies, the core teams are basically the top experts in the industry. Many technologies come from science and technology companies such as Nvidia, AMD, Huawei, Shangtang.

Start -up companies have completed new financing one after another, and have launched new products to accelerate commercialization.In August, Gyeonga Technology released the first general GPU chip BR100, which has a 7 -nanometer process. Up to now, Takuya Technology has completed Series B financing, with a total financing amount exceeded 5 billion yuan; in July, Mu Xi announced the completion of 1 billion yuanYuan RMB Pre-B round financing, the first chip has been filled in January this year; in the same month, the number of wisdom cores announced the completion of the C+round and C ++ round of financing over 1 billion yuan.The chip is 垓 100. As of the end of March 2022, the Sky 垓 100 achieved a sales order of nearly 200 million yuan.EssenceSeveral investors pointed out to reporters that whether it is GPU or AI, although financing is constant, the most important thing depends on the situation of commercial landing and scale. There are many companies related to AI chips on the market, and they will be integrated in the future.

- END -