Why does Deepseek model need “100 times more calculations”

Executive Director Jensen Huang discussed impact of the Chinese Startup of the new Deepseek artificial intelligence model in an interview with CNBC on Wednesday Jim Cramer at NvidiaAnnual GTC Conference. Huang claims that the revolutionary model will actually require more calculations than many in the industry think.

According to Huang, the Deepseek R1 model is “fantastic” because it is “the first model of open source reasoning”. He explained that the model breaks the problems step by step, is able to come up with different answers and can check that his answer is correct.

“This reasoning AI consumes 100 times more explicit than AI that is not overwhelmed,” he said. “That was the opposite, it was the exact opposite conclusion they all had.”

At the end of January, the Deepseek model sparked a large-scale sale in AI’s shares, as investors feared that the model could be presented, as well as the best competitors using less energy and money. Nvidia fell 17% in one session to lose nearly $ 600 billion, the largest so far A day drop for an American company.

Huang also talks about some of what his company has discovered So far at his conference, including new II infrastructure for robotics and enterprise, emphasizing partnerships with companies such as DELL., HPE., Accent., Servicenow and CrowdstrikeS He thought widely on the AI ​​boom, noting the way Hype had shifted from purely generative AI to reasoning models. He also predicts that world calculation capital expenditures seem to be about to reach a trillion dollar by the end of the decade and that the bigger part of this money will be used for AI.

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

Sign up now For the CNBC investment club, follow Jim Cramer’s every move on the market.

Rejection The CNBC Club charity Trust Club holds NVIDIA and CROWDSTRIKE shares.