NVIDIA CEO Jen-Hsun Huang's first public response to DeepSeek: investors misunderstand, AI development still needs strong computing power

Feb. 21 - At an online event on Thurs.NvidiaChief Executive Officer of the CompanyJen-Hsun Huang(Jensen Huang) said that investors are interested in the DeepSeek Misconceptions about the progress made in the field of artificial intelligence have contributed to the market's misguided reaction to NVIDIA stock.

NVIDIA CEO Jen-Hsun Huang's first public response to DeepSeek: investors misunderstand, AI development still needs strong computing power

In January, DeepSeek released an open-source inference model called R1. DeepSeek claimed that the big language model behind R1 was developed based on a weaker chip and far less capital investment than mainstream AI models in the West. The news triggered a sell-off of Nvidia shares, causing Nvidia's market capitalization to evaporate by $600 billion (note: currently around RMB 4.35 trillion) and Jen-Hsun Huang's personal fortune to shrink by nearly 20%. However, Nvidia's shares have since recovered most of the lost ground.

After the launch of DeepSeek-R1, investors began to question the need for big tech companies to spend trillions of dollars on AI infrastructure, arguing that the computing power needed to train models could be far less than expected.

In a pre-recorded interview on Thursday, Jen-Hsun Huang pointed out that the market's sharp reaction stems from investors' misinterpretation. Jen-Hsun Huang emphasized thatAlthough the development of R1 seems to have reduced the reliance on computing power, the AI industry still needs strong computing power to support model training post-processing methods, these methods enable AI models to reason or predict after training.

As inference methods continue to evolve and diversify, the demand for the powerful arithmetic provided by NVIDIA's chips will grow as well. Jen-Hsun Huang said at the event, "From an investor's perspective, they believe that the world is divided into two phases: pre-training and inference, and that inference is about asking the AI questions and getting immediate answers. I don't know who is responsible for this misunderstanding, but clearly this notion is wrong." He further noted that pre-training is still important, but post-processing is "the most important part of intelligence" and "a key part of learning to solve problems."

Jen-Hsun Huang also mentioned that DeepSeek's innovations have energized the field of artificial intelligence. He said, "The incredible enthusiasm that has been shown globally since R1 was made open source is an extremely exciting thing to see."

Previously, a spokesman for NVIDIA had responded to the market's reaction with a written statement, but this is the first time Jen-Hsun Huang himself has publicly commented on the issue.

statement:The content of the source of public various media platforms, if the inclusion of the content violates your rights and interests, please contact the mailbox, this site will be the first time to deal with.
Information

OpenAI Launches Operator, an AI Agent Tool for Multinational ChatGPT Pro Users: Automates Daily Tasks

2025-2-21 18:00:50

Information

Meta Defends Using Pirated Material to Train AI: Downloading Without Sharing Is Legal

2025-2-22 11:16:00

Search