NVIDIA is the world's largest GPU manufacturer, and its chips have a wide range of applications in artificial intelligence, cloud computing, gaming, and other fields. However, due to export restrictions by the U.S. government, Nvidia had to launch a "special" chip for the Chinese market with reduced performance - GeForce RTX 4090 D. The chip, which will hit the market in January 2024, has already been snubbed by some of China's biggest cloud computing companies. Alibaba Group, Tencent, Baidu and ByteDance have all said they plan to significantly reduce their purchases of Nvidia's GPUs this year and instead seek alternatives to domestic chips.
1. Why did Nvidia launch "special" chips?
Nvidia's "special" chips were born in response to the U.S. government's technology blockade. In October 2022 and October 2023, the U.S. Department of Commerce's Bureau of Industry and Security (BIS) updated export controls on advanced semiconductors and computing equipment to China, including some high-performance AI chips. These chips include NVIDIA's A100, H100, A800, H800, etc., as well as AMD's MI300X, MI300A, MI250X, MI250, etc. The common feature of these chips is that they have a data transfer rate of more than 600 GB per second, or a computing power of more than 4800 TOPS, which is considered a chip of strategic importance and must not be exported and sold to China.
In order to circumvent U.S. export restrictions, Nvidia has launched a "special" chip that reduces performance - GeForce RTX 4090 D. Nvidia says the chip, which consumes nearly 75% of the energy of the RTX 3090 Ti, was launched to commemorate the Chinese Year of the Dragon, but was actually designed to comply with U.S. regulatory requirements. Nvidia's CEO Jensen Huang said that the company will continue to fully comply with U.S. trade regulations while providing compliant chip products for the Chinese market.
2. Why was NVIDIA's "special" chip snubbed by Chinese buyers?
Although Nvidia's "special" chips have been reduced in performance, there has been no corresponding adjustment in the price. It is reported that the price of the RTX 4090 D chip starts at 12999 yuan, which is the same as the original RTX 4090 chip. This means that Chinese buyers need to buy chips with reduced performance at the same price. This is obviously not a good deal, and it is the main reason why Chinese buyers are not interested in Nvidia's "special" chips.
In addition, NVIDIA's "special" chips also have hidden dangers in terms of safety and reliability. On the one hand, as the U.S. government's technology blockade policy is likely to continue to tighten, it is still unknown whether Nvidia will be able to supply Chinese buyers consistently and stably. If the U.S. government imposes tougher restrictions again, Nvidia's "special" chips may also be banned from export, which will bring huge risks and losses to Chinese buyers. On the other hand, since NVIDIA's "special" chips are designed to circumvent U.S. regulations, their technology and quality may be inferior to the original chips, and there may also be some potential flaws and vulnerabilities. This will affect the user experience and trust of Chinese buyers, and may also threaten the data security and business interests of Chinese buyers.
3. How can domestic chip companies such as Huawei seize the opportunity?
NVIDIA's "special" chips are cold in China, reflecting the demand and expectations of Chinese buyers for domestic chips. In fact, as the U.S. technology blockade of China continues to escalate, local Chinese companies have made great strides in the design and manufacture of AI chips, and although they have not yet fully caught up with the international top level, they are growing faster than many expected. Among them, Huawei is the leader of domestic chip companies, and its Ascend series AI chips have demonstrated strong competitiveness and innovation in cloud computing, edge computing, and terminal equipment.
For example, the Ascend 910 chip released by Huawei in 2019 is the world's first AI chip with a 7nm process, and its computing power is as high as 256TOPS, which is twice that of NVIDIA's V100 chip. In addition to the Ascend series of AI chips, Huawei has also launched the Kunpeng series of general-purpose chips, which are based on the ARM architecture and are suitable for servers, desktops, and notebooks. It is reported that Huawei's Kunpeng 920 chip is the world's first 7nm ARM server chip, and its performance is higher than Intel's Xeon chip.
Therefore, Huawei's Ascend and Kunpeng chips not only have the capital to compete with international giants in terms of performance, but also have obvious advantages in terms of price, security, reliability, compatibility, and openness. As a result, Huawei's Ascend and Kunpeng chips have become more and more popular and recognized in the Chinese market, and have also attracted the favor of cloud computing giants such as Alibaba, Tencent, Baidu, and ByteDance.
epilogue
Therefore, NVIDIA's "special" chips are cold in China, reflecting the demand and expectations of Chinese buyers for domestic chips. As a leader in domestic chips, Huawei has seized this opportunity with the strong performance and advantages of its Ascend and Kunpeng chips, and has won the trust and support of Chinese buyers. Huawei's Ascend and Kunpeng chips will not only help improve China's cloud computing and AI capabilities, but also promote the development and innovation of China's chip industry, making important contributions to China's digital transformation and independent and controllable technology system. The above is all the content of this article, if you like this article, welcome to like and collect and follow~