The NVIDIA H200 NVL PCIe GPU is the latest product based on the NVIDIA Hopper architecture, designed for low-power, air-cooled enterprise racks. It offers flexible configurations to accelerate AI and high-performance computing (HPC) workloads of various scales. The H200 NVL has higher memory and bandwidth than the NVIDIA H100 NVL, enabling faster fine-tuning of large language models (LLMs) and providing up to 1.7 times the inference performance boost. Furthermore, the H200 NVL supports NVIDIA NVLink technology, allowing GPU communication speeds that are seven times faster than that of the fifth generation PCIe, addressing the demands of HPC, large language model inference, and fine-tuning. The H200 NVL also comes with robust software tools, including NVIDIA AI Enterprise, a cloud-native software platform for developing and deploying production AI.