The core of artificial intelligence is to simulate human intelligence processes through algorithms. This technology has permeated various industries, ranging from natural language processing, computer vision to autonomous driving and medical diagnosis. The application areas of AI are expanding at an unprecedented speed. Each generation of AI computing power chips is increasing power consumption. For instance, NVIDIA has moved from the AI GPU chip A100 to the Hopper architecture's H100, H200, and the Blackwell architecture's B100, B200, etc. Each product represents a leap in computing power, while power consumption ranges from several hundred watts to several thousand watts, and the heat generated during workloads is also increasing.