GPU Per Hour

GPU Per Hour is a real-time cloud GPU pricing aggregator built for AI/ML practitioners who need to find the best compute value fast. Instead of checking dozens of vendor pages manually, the platform tracks live pricing and availability across 25+ cloud providers (including AWS, Lambda Labs, RunPod, and Vast.ai) and indexes 10,000+ GPU listings. You can compare hourly rates for popular training and inference hardware such as NVIDIA H100, A100, and RTX 4090, then quickly narrow results to the exact configuration you need.

The interface is designed around practical decision-making: filter by GPU model and VRAM, choose regions, and refine by infrastructure details such as NVLink support, security tiers, and deployment types. Results can be sorted by price per GPU to surface the lowest-cost options immediately. When you find a matching instance, GPU Per Hour sends you to the provider to deploy, so you can go from research to running jobs with minimal friction.

To help you estimate real spend, the site includes a cost calculator and head-to-head provider comparisons that make it easier to weigh tradeoffs between large enterprise clouds and specialized GPU providers. Whether you’re training a model, running experiments, or sourcing short-notice capacity for rendering or distributed workloads, GPU Per Hour aims to make GPU shopping transparent, fast, and data-driven.

Support is available at [email protected], with additional information on the Contact and About pages.