Recently, Bill Dally, the chief scientist of NVIDIA, said that the AI export control ban imposed by the United States on China has given China a great deal of development space. Researchers in China who previously wrote programs for NVIDIA are now all helping Huawei write programs.

Bill Dally, Chief Scientist of NVIDIA, NVIDIA official website

Previously, Jensen Huang, CEO of NVIDIA, mentioned in the conference call for the first fiscal quarter earnings report that China is one of the largest AI markets in the world and also a stepping stone to global success. There are half of the world's AI researchers there, and winning the platform market in China will lead globally. It was reported that in 2019, China had about one-third of the world's high-end AI researchers, and this number has grown to nearly half of the global total today.

Dally pointed out that although China does not have as large a share in the global AI hardware market as the United States, it has successfully attracted and cultivated a large number of AI R&D talents. He mentioned that many of these talents were previously involved in NVIDIA's R&D work. Dally believes that if it were not for U.S. regulations, Huawei would not have grown so rapidly, and now those researchers are writing software for Huawei.

Dally stated that Huawei has already provided a fairly reliable solution and is growing rapidly.

Dally noted that whether AI learning can make significant leaps in the future depends on both software and hardware, and pointed out that NVIDIA has greatly improved the energy efficiency of AI learning over the past decade, with only a small part coming from advancements in semiconductor technology, while most of it comes from optimization in hardware design; Dally believes that optimization in hardware design will remain the main driver of future AI growth.

In April this year, Zhang Ping'an, standing director of Huawei and CEO of Huawei Cloud, announced a breakthrough in AI infrastructure architecture and launched the CloudMatrix 384 super node. The release of NVIDIA NVL72 last year provided a 30 times real-time speed increase for inference of trillion-parameter large language models, attracting widespread attention. As the first commercially available large-scale super node in the country, Huawei CloudMatrix 384 surpasses NVIDIA NVL72 comprehensively in terms of scale, performance, and reliability.

At the same time, in terms of chips, the Modern International Relations Institute of China posted in May that although U.S. export controls are painful, they have instead sparked a wave of self-innovation in advanced AI chips in China. Huawei's Ascend chip is the best example. Chinese companies and institutions have begun to purchase and use Ascend chips on a large scale. The Financial Times of the UK predicts that more Chinese tech companies will compete for this Ascend chip. Meanwhile, chips from Chinese chip companies such as Hygon Information and Cambricon are currently being tested by major Chinese tech giants.

According to NVIDIA's first fiscal quarter financial report for 2025, NVIDIA's revenue for the first fiscal quarter was $44.1 billion (approximately RMB 317.5 billion), an increase of 69% year-over-year, with a net profit of $18.78 billion (approximately RMB 135.2 billion), an increase of 26%. At the same time, as of April 27, the share of the Chinese market in NVIDIA's revenue was 12.5%, further declining from 14% and 15% in the previous two months.

This article is an exclusive contribution from Guancha Observer, unauthorized reproduction is prohibited.

Source: https://www.toutiao.com/article/7512251046313935396/

Disclaimer: The views expressed in this article are solely those of the author. Feel free to express your opinion by clicking the 'thumbs up' or 'thumbs down' button below.