While the launch of the Chinese AI startup DeepSeek dealt a heavy blow to AI companies worldwide, Nvidia took the biggest hit, experiencing its worst stock drop since 2020. However, despite this, Nvidia was among the few companies that considered DeepSeek an “excellent AI advancement,” believing it would ultimately benefit the green team by driving even greater demand for its AI-powered GPUs.

We recently learned that Nvidia’s US AI chips are around “60 times” faster than Chinese counterparts as a result of export controls and restrictions. Even still, that hasn’t stopped the company from continuing business in the region. Chinese-born AI models such as DeepSeek have had a massive impact on the industry, and one expert claims it could bring “the end of closed-source AI”.

Jensen Huang has a lot of respect for DeepSeek’s work

Nvidia’s CEO, Jensen Huang, once again praised DeepSeek, stating, “DeepSeek was fantastic” and adding, “It was fantastic because it open-sourced a reasoning model that's absolutely world-class,” says Huang in a recent interview with CNBC.

According to Huang, DeepSeek has popularized a reasoning model that will require more chips, and as we enter what he calls “the beginning of the reasoning AI era,” the demand for computing power is expected to grow substantially.

According to Huang, this new reasoning approach, which enables AI models to think “about how to best answer,” will require 100 times more computing power than older AI models. We’ve previously seen models like OpenAI’s GPT-4 use this reasoning process to think before responding, and DeepSeek has further fueled this advancement, improving it with its open-source approach.

“The amount of computation necessary to do that reasoning process is 100 times more than what we used to do”

Nvidia CEO, Jensen Huang. Source: CNBC

A recent example of just how demanding reasoning AI models can be in terms of computing power is Elon Musk's xAI, which recently unveiled its latest artificial intelligence model, Grok 3. This AI model is designed to “think like a human being,” and behind it is a colossal supercomputer training the model with over 100,000 Nvidia GPUs. So, for the green team, reasoning models like those from DeepSeek aren't a negative – in fact, they are driving even greater demand for Nvidia's AI chips.

About the Author

Hassam Nasir

Hassam boasts over seven years of professional experience as a dedicated PC hardware reviewer and writer.