Another factor is that DeepSeek is not just doing inference, but also training models, so they can use underutilized compute nodes for training during off-peak

Another factor is that DeepSeek is not just doing inference, but also training models, so they can use underutilized compute nodes for training during off-peak hours, as described in their DeepSeek v3 article: https://github.com/deepseek-ai/open-infra-index/blob/main/202502OpenSourceWeek/day_6_one_more_thing_deepseekV3R1_inference_system_overview.md#statistics-of-deepseeks-online-service https://github.com/deepseek-ai/open-infra-index/blob/main/20... But I agree that the main driver is that they are really good at optimizing. They will have chosen their architecture in such a way that it will be as efficient as possible on their own infrastructure, so they have a massive head start. Inference framework developers still have to catch up.