NVIDIA’s stock might’ve first crashed when China’s DeepSeek released its R1 model, but its CEO believes that models such as DeepSeek’s will be a net positive for his company in the long run.
NVIDIA CEO Jensen Huang has said that the release of DeepSeek R1 was “fantastic”. Speaking to CNBC, Huang said that the increased computational demands of these models as a significant opportunity for NVIDIA, as opposed to the threat it was being initially perceived to be.
“R1 DeepSeek was a fantastic thing,” Huang began. “And the reason for that is because R1 is the first open-source reasoning model. What makes R1 incredible is that it ‘reasons’. That’s why the answer’s so good. And it breaks the problem down step by step. It asks itself while it’s thinking. It comes up with several different options for the answer. It might actually verify that the answer is correct; it solves a quadratic equation correctly, takes the number, plugs it back in, and makes sure that that actually solves the equation.”
He continued, “And so it’s a reasoning AI. This reasoning AI consumes 100 times more compute than a non-reasoning,” he said. “And that’s you (NVIDIA),” the anchor exclaimed. “It was the exact opposite conclusion that everybody has, and actually consumes 100 times more computing,” Huang smiled.
The release of DeepSeek’s R1 in January 2025 sent shockwaves through the tech world. Many feared that the open-source nature of the model would disrupt NVIDIA’s dominance in the AI hardware market, because DeepSeek had priced its models 90 percent cheaper than OpenAI. NVIDIA’s stock too had been hit, and had fallen more than 10 percent in a single day. Markets had assumed that DeepSeek had come up with a more efficient way to build AI models, which meant that NVIDIA’s chips — which power these models — wouldn’t be able to command the same premium they once did.
While DeepSeek did indeed make many technical innovations that led to its lower price, Huang seems to be focusing on how DeepSeek is a reasoning model, and uses much more compute than traditional AI models. If reasoning models become the new standard for AI computation, it would mean that the demand for computing power could shoot up, and with it a demand for NVIDIA’s chips. Only time will tell whether Huang’s optimistic outlook proves correct, but his statements offer a compelling counterpoint to the prevailing anxieties around open-source AI and its potential impact on established industry players.