Author of this article: Li Xiaoyin
Source: Hard AI
After the US stock market on Wednesday, May 22, Hwang In-hoon revealed in the latest earnings call that Nvidia will speed up the chip architecture update speed, from a biennial update to an annual update:
“I can announce that Blackwell will launch another chip later. Our pace is one generation at a time.”
At the same time, Hwang In-hoon said that Nvidia will not only accelerate the development of AI chips, but will also accelerate the production of all other types of chips:
“We are going to fully advance at a very fast pace.”
“A new CPU, a new GPU, a new network card, a new switch... lots of chips are on the way.”
According to information, Nvidia's previous chip architecture update cycle was about every two years: Ampere was launched in 2020, Hopper was launched in 2022, and Blackwell was launched in 2024.
The reason for the decision to speed up the shift to “yearly replacement” of AI chips is mainly due to strong market demand.
On the conference call, Hwang In-hoon predicted that AI chips will continue to be in short supply for some time.
“We expect demand to exceed supply for some time as we transition to H200 and Blackwell. Because everyone is anxious to get their infrastructure online, they want to both save money and make money, and they want to achieve this goal as soon as possible.”
Meanwhile, demand for GPUs in the automotive market continues to soar. Hwang In-hoon said that car companies will become “the largest vertical enterprises in data centers this year.”
According to Tianfeng Securities analyst Guo Mingyi's previous forecast, Nvidia's next-generation architecture “Rubin” will be launched in the fourth quarter of 2025. At that time, a new R100 AI chip will be introduced, which is in line with Huang Renxun's plan announced this time.
