Nvidia is treating DeepSeek not as a warning sign, but as another reason to expect more demand for AI compute. On the latest earnings call on Wednesday, CEO Jensen Huang pushed back on the idea that DeepSeek’s R1 model could weaken Nvidia’s sales outlook.
The discussion followed a sharp market reaction last month, when speculation that DeepSeek’s R1 model required far fewer chips to train helped fuel a record drop in Nvidia’s stock price. Huang’s answer was direct: reasoning models, including R1, increase the need for compute rather than reduce it.
Huang frames DeepSeek as demand creation
Huang described DeepSeek R1 as an important development for artificial intelligence, especially because it made a world-class reasoning AI model available as open source. His view is that wider use of reasoning models expands the amount of computation developers need.
“Reasoning models can consume 100 times more compute, and future reasoning models will consume much more compute,” Huang said. “DeepSeek R1 has ignited global enthusiasm. It’s an excellent innovation, but even more importantly, it has open sourced a world-class reasoning AI model. Nearly every AI developer is applying R1.”
That argument matters because it separates two different questions. One is whether a model can be trained with fewer chips. The other is how much compute is needed once reasoning models become widely used. Huang focused on the second question, saying the use of reasoning models points to greater compute demand.
In that framing, DeepSeek is not a substitute for Nvidia’s hardware market. It is a catalyst for more AI development, more experimentation, and more use of models that require significant computation.
Sales are still moving higher
Nvidia’s reported numbers support Huang’s confidence. The company posted another record-breaking quarter, with revenue reaching $39.3 billion. That result was above both its own projections and Wall Street estimates.
The company also said it expects revenue in the next quarter to rise again, to around $43 billion. That guidance suggests Nvidia does not see the DeepSeek reaction translating into weaker near-term demand.
Data center sales remain the center of the story. Nvidia’s data center sales nearly doubled in 2024 to $115 billion and rose 16% from the previous quarter, according to its earnings release.
Those figures are important because they show where AI demand is appearing most clearly in Nvidia’s business. The data center segment reflects the infrastructure side of AI, where companies need chips and compute capacity to train, run, and scale models.
Blackwell sits at the center of the reasoning AI pitch
Huang also used the call to highlight Nvidia’s latest Blackwell chip. He said Blackwell was custom-built for reasoning and described current demand for it as “extraordinary.”
That detail connects Nvidia’s product story with Huang’s broader DeepSeek argument. If reasoning models consume much more compute, then chips designed for reasoning become central to the company’s growth case.
The message from Nvidia is not that DeepSeek is irrelevant. It is that DeepSeek demonstrates the direction AI development is moving. More reasoning AI means more compute, and Nvidia is positioning Blackwell as a chip built for that workload.
Huang summed up the company’s confidence in a short statement: “We will grow strongly in 2025,” Huang said.
AI infrastructure spending remains intense
The wider market also appears to be moving in Nvidia’s favor. Even after the panic over DeepSeek last month, the market for AI chips shows no signs of cooling off.
Meta, Google, and Amazon have all unveiled massive AI infrastructure investments. Together, those companies are committing hundreds of billions for the coming years.
That continued spending is a key backdrop for Nvidia. If major technology companies keep building AI infrastructure, Nvidia remains tied to one of the largest areas of current technology investment.
The DeepSeek episode still raised a serious question for the AI chip market: what happens if powerful models can be built with fewer chips? Nvidia’s response is that model efficiency does not end the need for compute. Instead, it can make advanced AI more available, which may increase usage and create more demand for reasoning workloads.
For now, Nvidia’s numbers and guidance show no visible slowdown. Revenue has reached $39.3 billion, the company expects around $43 billion next quarter, and Blackwell demand is described as “extraordinary.” Huang’s view is clear: DeepSeek R1 may change the conversation around AI models, but it has not changed Nvidia’s confidence in AI chip demand.