Nvidia plunged 16% as U.S. stocks tumbled across the board, with the Nasdaq leading the selloff in a broad rout targeting artificial intelligence infrastructure makers. The decline accelerated after China released DeepSeek, a low-cost AI model that challenged the dominance of American tech giants in the generative AI space.
The market repriced expectations around AI chip demand and software superiority. Nvidia, the bellwether of the AI boom, bore the brunt of the selling as investors fled positions in semiconductor manufacturers that had benefited from the race to build out compute-intensive large language models. Other chipmakers and AI infrastructure vendors faced double-digit percentage drops.
DeepSeek's emergence raised questions about whether expensive, proprietary AI models could maintain pricing power if cheaper alternatives delivered comparable results. The model's efficiency suggested that companies might not need to purchase as many high-end GPUs from Nvidia to power their AI systems. This efficiency advantage threatened to compress margins across the AI supply chain.
The Nasdaq's underperformance reflected concentration risk in mega-cap tech stocks that had driven much of the broader market's gains through 2024. The index's heavy weighting toward semiconductor and software names meant the AI sell-off disproportionately hurt the index's performance.
The rout illustrated how quickly sentiment can shift in growth sectors dependent on sustained capital investment and technological moats. Nvidia's stock had surged on expectations that its chips would remain essential infrastructure for any company pursuing AI capabilities. DeepSeek challenged that assumption by delivering competitive performance through architectural innovations rather than brute-force computing power.
Investors reassessed whether the AI infrastructure buildout would generate the returns previously priced into semiconductor stocks. The selloff sent ripples through suppliers and service providers dependent on sustained demand for cutting-edge processors.