Why the Depseek model needs ‘100 times more computer science’

CEO Jensen Huang discussed the impact of the new artificial intelligence model of the China Depseek startup in a Wednesday interview with CNBC’s Jim Cramer in NvidiaThe annual GTC conference. Huang said that the innovative model will really require More calculation that many thought of the industry.

According to Huang, Deepseek’s R1 model is “fantastic” because it is “the first open source reasoning model.” He explained that the model breaks down the problems step by step, you can find different answers and can verify if your answer is correct.

“This reasoning of AI consumes 100 times more computing than an unreasonable AI,” he said. “It was exactly the opposite, it was exactly the opposite conclusion that everyone had.”

At the end of January, the Deepseek model triggered a massive sale in AI shares, since investors feared that the model could work as well as the main competitors that use less energy and money. Nvidia collapsed 17% in a session to lose about $ 600 billion, the bigger in history Drop of one day for an American company.

Huang also talked about something his company has revealed Until now at its conference, including the new AI infrastructure for robotics and for the company, highlighting associations with companies such as Give birth, HPE, Accentuate, Servicenow and Crowdstrike. He reflected widely in the AI ​​boom, pointing out the way Hype has changed purely generative to reasoning models. He also predicted that the world’s capital expenses seem to be on his way to reaching a billion dollars for the end of the decade, and that most of that money will be used for AI.

“Then, our opportunity as a percentage of billion dollars for the end of this decade is quite large,” said Huang. “We have a lot of infrastructure to build.”

Register now So that the CNBC Investing Club follows each movement of Jim Cramer in the market.

Discharge of responsibility The CNBC Investing Club Charitable Trust has actions of NVIDIA and Crowdstrike.