The Tiangong 6000 Series AI training server is an ultra-high computing power density GPU server designed for intelligent computing centers. It is compatible with NVIDIA 8GPU modules and supports multi-GPU in a single server and multi-server cluster deployment, meeting the parallel training needs of trillion-parameter ultra-large models. Equipped with Intel® Xeon® processors, high-speed DDR5 memory and NVMe high-performance storage, it delivers robust computing performance for smart computing centers. It is widely applied in conversational AI, generative AI, industrial automation, business intelligence and analysis, drug discovery, climate and weather modeling, finance, economics and other fields.