Huang of Nvidia says away from Deepseek, says artificial intelligence

Nvidia (NvdaCEO Jensen Huang does not highlight the introduction of artificial intelligence models such as China -trained Deepsek R1. Instead, Chip CEO said that the world needs a much more computing power to deal with the so -called AICEC AI applications than expected previously.

“Last year, Huang said,” Huang said in the keynote speech at Nvidia’s GTC 2025.

“The requirements of the account, and the law of scaling in artificial intelligence is more flexible, and in fact, excessive registration. The amount of the account we need at this stage is the result of the intelligence agent, as a result of thinking, is easily more than a hundred times than we thought we needed this time last year,” he explained.

AIGNIC AI is a type of artificial intelligence that can take action on behalf of the user. Thinking, or thinking about artificial intelligence, is a kind of artificial intelligence that mimics the way people think by breaking up problems step by step to find the best answer to the user’s inquiry.

Deepseek’s R1 hit the scene in late January and sent Wall Street to a collapse when the company said that the logic model matches the capabilities of the Openai model and that it trained its broader model Deepseek V3 for about 5 million dollars, compared to tens of millions of Silicon Valley, which spent comparing models training.

Read more: How do you earn Nvidia money?

Investors went to broadcasts for doors, and sent the market value of NVIDIA Nearly $ 600 billion decreased In fears that cloud companies will not need to spend billions on NVIDIA chips anymore.

Recently, NVIDIA has been concerned with concerns about President Trump’s introductory threats and the possibility of the United States imposing new export controls on China chips.

NVIDIA has decreased by 14 % so far, but it is still 30 % over the past 12 months.

While the customs tariffs and export controls outside the hands of NVIDIA, outside the pressure for exceptions, the company is fully able to deal with disturbing Deekseek concerns, Huang did this during his speech.

During almost two hours that lasted, CEO explained how thinking models will benefit from chips like the new Blackwele Ultra and Vera Rubin. He explained that this will only continue in the future, such as artificial intelligence, including human robots and self -driving cars in the market.

The CEO also expanded his plans for the NVIDIA CUDA program, which allows developers to take advantage of the company’s chips for public treatment, the omniverse simulation platform, and a set of other services.