On January 27th, Chinese startup DeepSeek's AI Assistant surpassed rival ChatGPT to claim the top spot as the highest-rated free application on Apple's App Store in the United States. The application, powered by the DeepSeek-V3 model, is acclaimed by its creators for its competitive edge among open-source models globally and its comparison to advanced closed-source models.
The surge in popularity among U.S. users since its January 10th launch, as reported by app data research firm Sensor Tower, underscores DeepSeek's impact on Silicon Valley. This success challenges prevailing beliefs about U.S. dominance in AI and the efficacy of Washington's export restrictions targeting China's advanced chip and AI capabilities.
Both ChatGPT and DeepSeek's AI models rely on advanced chips for their training. Despite the Biden administration's expanded bans on exporting such chips to China since 2021, DeepSeek researchers noted in a recent paper that the DeepSeek-V3 utilized Nvidia's H800 chips for training, at a cost of under $6 million.
While this assertion has been challenged, the suggestion that the chips were less powerful than those Washington aims to restrict, coupled with the relatively modest training expenses, has stirred skepticism among U.S. technology executives regarding the efficacy of technology export controls.
DeepSeek, a small Hangzhou-based startup established in 2023, has limited information publicly available. Originating from the year that saw the launch of Baidu's first Chinese AI large-language model, the company stands out among the many Chinese tech firms that have introduced their own AI models. DeepSeek earns distinction as the first to be lauded by the U.S. tech industry for matching or even surpassing the performance of cutting-edge U.S. models.