Nvidia designed the H100 GPU, a major accelerator used in artificial-intelligence data centers.
The H100 is based on Nvidia’s Hopper architecture and was announced in 2022. It uses specialized hardware, including fourth-generation Tensor Cores, to accelerate matrix calculations used in machine learning and other high-performance workloads.
Demand for H100 systems helped explain Nvidia’s dramatic rise in market capitalization during the generative-AI boom. Technology companies and cloud providers purchased large clusters of these accelerators to train and operate language models, image systems, recommendation tools, and scientific applications.
The H100 is not a consumer graphics card in the usual sense. It is a data-center product designed for servers, and complete systems also require networking, memory, software, cooling, and power infrastructure. Nvidia’s CUDA ecosystem is another important part of its competitive position because it gives developers tools for programming its GPUs.