NVIDIA H200 NVL Tensor Core GPU 141GB HBM3e

NVIDIA H200 NVL is the upgraded Hopper-generation data center GPU with 141GB ultra-large HBM3e memory, designed for generative AI and HPC workloads. It brings higher memory bandwidth and larger memory pool than H100 NVL, greatly accelerating inference for 70B, 400B and larger LLMs.

It supports FP8 native Tensor Core acceleration and MIG partitioning. Multiple H200 NVL can be linked via NVLink to form a unified high-speed memory pool. PCIe air-cooled dual-slot form factor, friendly for standard enterprise rack servers.It is widely deployed in private AI clusters, AI cloud operators and research institutes for batch LLM inference and large-scale scientific computing. No display outputs.