DeepSeek, a Chinese artificial intelligence startup, has upended the AI market, topping app download charts and triggering a sell-off in U.S. tech stocks.
The company launched its latest model, DeepSeek R1, in January, claiming it rivals OpenAI’s technology at a fraction of the cost. The announcement rattled investors, taking billions of dollars off the market value of chipmaker NVIDIA and raising questions about U.S. dominance in AI.
DeepSeek R1, a reasoning-based AI model, reportedly cost just $6 million to train – far less than OpenAI’s GPT-4, which CEO Sam Altman suggested exceeded $100 million.
The startup, founded in December 2023 by Liang Wenfeng, circumvented U.S. chip restrictions by stockpiling Nvidia A100 processors and integrating them with lower-cost alternatives. DeepSeek’s chatbot quickly became the most downloaded free app on Apple’s App Store. However, it has faced security concerns. Australia banned it from government devices, while Italy blocked the app over data privacy issues.
The model’s efficiency and lower costs have sparked uncertainty about AI development, challenging assumptions that cutting-edge models require vast budgets and high-performance chips.