Chinese AI firm DeepSeek reports its R1 model was trained for only $294,000, significantly lower than US competitors. The company used 512 Nvidia H800 chips over 80 hours for training, as revealed in a Nature article co-authored by founder Liang Wenfeng. This announcement follows investor concerns about DeepSeek's rising influence in AI, echoing OpenAI CEO Sam Altman’s claim that training foundational models costs over $100 million. The cost efficiency marks a notable shift in the AI landscape.