Nvidia's Jensen Huang Discusses DeepSeek's AI Model and its Computing Needs

Nvidia CEO Jensen Huang recently spoke with CNBC's Jim Cramer during the company’s annual GTC conference. Huang shared insights on the new artificial intelligence model developed by the Chinese startup DeepSeek. He emphasized that the model requires more computational power than most in the industry anticipated.

Huang described DeepSeek's R1 model as "fantastic," noting it is the first open-sourced reasoning model. This model is capable of addressing problems in a step-by-step manner, generating multiple possible answers, and verifying the accuracy of those answers.

In his words, "This reasoning AI consumes 100 times more compute than a non-reasoning AI. It was exactly the opposite conclusion that everybody had." This statement highlights a significant shift in understanding how powerful such models can be.

Earlier this year, in late January, the announcement of DeepSeek’s model led to a significant sell-off in AI stocks. Many investors were concerned that DeepSeek’s model could perform comparably to top competitors while using considerably less energy and financial resources. Following this news, Nvidia’s stock experienced a drastic decline of 17% in a single day, equating to a loss of nearly $600 billion—marking the largest one-day drop for a U.S. company in history.

During the interview, Huang also discussed Nvidia's recent announcements regarding new AI infrastructure tailored for robotics and business applications. He highlighted various partnerships that Nvidia has formed with major companies including Dell, HPE, Accenture, ServiceNow, and CrowdStrike.

Reflecting on the overall AI boom, Huang commented that the focus is shifting from solely generative AI toward increasingly complex reasoning models. He predicted that global expenditures on computing could reach a trillion dollars by the end of the decade, with a significant portion of that investment being funneled into AI technologies.

"Therefore, our opportunity as a percentage of a trillion dollars by the end of this decade is quite large," Huang noted. "We have substantial infrastructure to build out going forward."

Nvidia, AI, Computing