Nvidia Acknowledges DeepSeek R1 as a Major AI Breakthrough Despite Market Impact
Nvidia has recognized the DeepSeek R1 model as a significant advancement in artificial intelligence, even as the Chinese startup’s emergence led to a 17% drop in Nvidia’s stock price on Monday.
In a statement to CNBC, an Nvidia spokesperson described DeepSeek R1 as “an excellent AI advancement and a perfect example of Test Time Scaling.” The spokesperson added that DeepSeek’s work demonstrates how new models can be developed using existing techniques while complying with export control regulations.
DeepSeek released R1 last week as an open-source reasoning model, which reportedly outperforms leading AI models from U.S. companies like OpenAI. Notably, the training cost of R1 was under $6 million—significantly lower than the billions invested by Silicon Valley firms in AI development.
Despite concerns over export compliance, Nvidia affirmed that DeepSeek used GPUs specifically designed for the Chinese market. This counters claims by Scale AI CEO Alexandr Wang, who suggested that DeepSeek might have used Nvidia models restricted from export to China.
DeepSeek R1’s success has sparked debates among analysts about the massive capital investments made by tech giants like Microsoft, Google, and Meta in Nvidia-based AI infrastructure. Microsoft recently announced an $80 billion AI infrastructure investment for 2025, while Meta CEO Mark Zuckerberg outlined plans to spend between $60 billion and $65 billion on AI development.
The emergence of Test Time Scaling—a concept championed by Nvidia CEO Jensen Huang and OpenAI CEO Sam Altman—suggests that AI performance can improve by increasing computational power during inference rather than just during training. This approach, used in OpenAI’s latest models and DeepSeek R1, could reshape AI development strategies in the future.
