GpuPerHour

GpuPerHour is a GPU price-comparison tool. It tracks real-time pricing, hardware specs, and stock availability across cloud compute providers, so you can find the cheapest GPU for a given job.\n\n\n\n## What it tracks\n\n\n\n- Live pricing and availability for more than 50 GPU models, including the NVIDIA H100, A100, RTX 4090, RTX 5090, L40S, and V100.\n\n- More than 30 providers: hyperscalers (AWS, GCP, Azure) alongside specialized AI GPU clouds (RunPod, Vast.ai, Lambda Labs, CoreWeave, TensorDock, VERDA, Crusoe, and others).\n\n\n\n## Filtering\n\n\n\nSort and filter by GPU model, total VRAM, geographic region, and instance type (spot, on-demand, reserved). The tool aggregates pricing so you can compare apples-to-apples hourly rates.\n\n\n\n## Why this kind of tool matters\n\n\n\nThe pricing gap on the same GPU is enormous: an NVIDIA H100 SXM5 80GB can range from $0.80/hour on the low end to $11.10/hour on the high end, which is a multi-thousand-dollar monthly difference for the same hardware. A tool that surfaces the gap is genuinely valuable for any team buying compute.\n\n\n\n## Who it is for\n\n\n\nAI researchers, ML engineers, and startup founders buying cloud GPU. Useful for everything from a single RTX 4090 for a weekend project to a cluster of H100s for foundation-model training.