In the 2025 AI arms race, China’s DeepSeek has become both a showcase of ambition and a cautionary tale on domestic chip limitations. Founded in Hangzhou in 2023 and backed by hedge fund High-Flyer, the startup’s R1 large language model — trained on Nvidia H100 GPUs — gained global recognition in January for advanced reasoning, coding, and multilingual performance, positioning it alongside OpenAI’s GPT-4.