Nvidia says new AI models capable of arriving at more complex answers will only increase demand for computing infrastructure. CEO Jen-Hsun Huang said concerns that DeekSeek's artificial intelligence model R1 would require fewer chips and more powerful servers were misplaced.
Huang Renxun told investors at the GTC conference on Wednesday that "the outside world's understanding of R1 is completely wrong. In fact, this model requires much higher computing power."
Nvidia is trying to convince a growing number of industries to invest in artificial intelligence equipment, promising that the economic benefits of the technology are just around the corner. As the world's most valuable chipmaker, Nvidia will face more scrutiny from investors in 2025. The market's main concern is whether Nvidia customers will maintain high spending on artificial intelligence infrastructure.
Nvidia's stock price rose 2.4% on Wednesday and has fallen 14% so far this year.
During the analyst conference, Huang was asked about efforts by customers to develop their own chips that could replace Nvidia's AI accelerators used in data centers. Companies such as Google have been working with Broadcom to develop their own application-specific integrated circuits (ASICs). But Huang Renxun countered that although many ASICs are well designed, they are not necessarily actually used in data centers.
He said that large customers need better chips to get more revenue from their infrastructure and should not use cheaper chips just to save on finished products.
Huang Renxun said that there is no comparison between competitor chips and Nvidia's Hopper chips, and the performance of the current Blackwell platform is 39 times higher than the previous generation.
If the U.S. economy falls into recession, Huang said that companies will shift investment to AI equipment because this is their business growth point.
Huang Renxun said that the impact of tariffs on imports should be small in the short term and will not have a major impact on Nvidia, because Nvidia will continue to transfer the manufacturing base of its most important product lines to the United States. NVIDIA is already using TSMC's production equipment in Arizona. As TSMC's production capacity increases, NVIDIA will increase production in the United States.