NVIDIA H200 (cloud, 141 GB)

Rented NVIDIA H200 SXM GPU with 141 GB of HBM3e at 4.8 TB/s, 989 TFLOPS of FP16 dense compute, and NVLink at 900 GB/s when rented as an 8-GPU node.

Designed for bursty training and inference with no capital outlay, billed per GPU-hour from $3.72 per GPU-hour on-demand on specialist clouds, with spot capacity near $1.00, and scalable from one GPU to hundreds in minutes with CUDA, PyTorch, and the NVIDIA container stack preinstalled by every provider.

Lives in someone else's data center: nothing to power, cool, or secure, and the trade is per-hour cost and data leaving your premises. Runs a 120-billion-parameter model at 8-bit on one GPU; 405 billion at FP8 on eight with full context.

$3.72/hr on-demand floor, $3.99 Jarvislabs, ~$1.00 spot. Price verified September 2026.