Nvidia Stock Plunges on Fears of AI Model Cost Reduction
Nvidia's (NVDA) stock experienced a significant drop of over 11% in premarket trading on Monday. This decline was part of a broader market sell-off, particularly in chip stocks, following the release of a new AI model by China's DeepSeek.
DeepSeek's latest AI model, released on January 20th, is seen as a potential threat to American venture capitalist Marc Andreessen, who described it as "one of the most amazing and impressive breakthroughs." The model has raised concerns about the cost-effectiveness of AI investments, as DeepSeek reported training costs of just $5.6 million compared to OpenAI's GPT model, which cost over $100 million.
These announcements have fueled fears that AI models may require fewer chips and energy resources than previously anticipated, potentially impacting semiconductor companies like Nvidia, which has become the world's largest through its dominance in the AI chip market.
In a note to investors, Raymond James semiconductor analyst Srini Pajjuri suggested that if DeepSeek's innovations gain widespread adoption, "model training costs could come down significantly," potentially challenging the need for large-scale XPU/GPU clusters.
Other chip stocks also faced declines in premarket trading, with Broadcom (AVGO) dropping nearly 12%, Micron (MU) falling almost 8%, and Advanced Micro Devices (AMD) down over 4%.
However, Pajjuri also noted that the situation could drive greater urgency among U.S. hyperscalers to leverage their access to GPUs to differentiate themselves from cheaper alternatives.
Additionally, concerns arise over potential implications of tightened U.S. export restrictions on DeepSeek's ability to train models, as well as the growing U.S. investment in AI infrastructure.