# Compute: what a GPU-hour costs

> 2319 per-GPU-hour prices for 20 accelerators (H100, H200, B200, GB200, MI300X, TPU, Trainium and more) across AWS, Azure, Google Cloud, Oracle, CoreWeave, Lambda, Nebius, Crusoe, Together, Runpod, DigitalOcean and Vast.ai, by region and pricing model (on-demand, spot, 1-year, 3-year, contract), with append-only history. By fru.dev.

Prices are read from each provider's own price list, API or pricing page (AWS price list JSON and spot feed, Azure Retail Prices API, Oracle price list API, Google Cloud and neocloud pricing pages, Vast.ai search API) and normalised to US dollars per GPU-hour: instance price divided by the GPUs in the instance; TPU and Trainium per chip-hour. List prices are read every Monday, AWS spot and Vast.ai daily. Last read: 2026-09-24 23:05:27 UTC.

## About this data

Prices from each provider's own price list, API or pricing page, per GPU-hour in US dollars before tax and discounts. Spot and marketplace prices are read every day and list prices every Monday, so check the provider before you buy. The list grows as providers add chips and offers. Not investment advice. Logos via logo.dev; trademarks belong to their owners.

## Chips

- [GB200](https://compute.fru.dev/chips/gb200): on-demand from $10.50 (CoreWeave) to $27.04 per GPU-hour, 3 providers
- [B300](https://compute.fru.dev/chips/b300): on-demand from $6.94 (Runpod) to $17.80 per GPU-hour, 6 providers
- [B200](https://compute.fru.dev/chips/b200): on-demand from $5.98 (Runpod) to $14.24 per GPU-hour, 9 providers
- [H200](https://compute.fru.dev/chips/h200): on-demand from $3.59 (Runpod) to $10.60 per GPU-hour, 11 providers
- [H100](https://compute.fru.dev/chips/h100): on-demand from $1.90 (Vast.ai) to $11.06 per GPU-hour, 12 providers
- [GH200](https://compute.fru.dev/chips/gh200): on-demand from $2.29 (Lambda) to $6.50 per GPU-hour, 2 providers
- [A100](https://compute.fru.dev/chips/a100): on-demand from $0.43 (Vast.ai) to $4.00 per GPU-hour, 9 providers
- [L40S](https://compute.fru.dev/chips/l40s): on-demand from $0.60 (Vast.ai) to $3.50 per GPU-hour, 8 providers
- [RTX PRO 6000](https://compute.fru.dev/chips/rtx-pro-6000): on-demand from $1.34 (Vast.ai) to $4.50 per GPU-hour, 5 providers
- [L4](https://compute.fru.dev/chips/l4): on-demand from $0.44 (Runpod) to $0.80 per GPU-hour, 3 providers
- [A10G / A10](https://compute.fru.dev/chips/a10g): on-demand from $1.01 (AWS) to $2.00 per GPU-hour, 3 providers
- [MI355X](https://compute.fru.dev/chips/mi355x): on-demand from $8.60 (Oracle Cloud) to $8.60 per GPU-hour, 2 providers
- [MI325X](https://compute.fru.dev/chips/mi325x): no on-demand price, 1 providers
- [MI300X](https://compute.fru.dev/chips/mi300x): on-demand from $2.59 (DigitalOcean) to $6.00 per GPU-hour, 4 providers
- [TPU Ironwood](https://compute.fru.dev/chips/tpu-ironwood): on-demand from $12.00 (Google Cloud) to $12.00 per GPU-hour, 1 providers
- [TPU v6e (Trillium)](https://compute.fru.dev/chips/tpu-v6e): on-demand from $2.70 (Google Cloud) to $2.70 per GPU-hour, 1 providers
- [TPU v5p](https://compute.fru.dev/chips/tpu-v5p): on-demand from $4.20 (Google Cloud) to $4.20 per GPU-hour, 1 providers
- [TPU v5e](https://compute.fru.dev/chips/tpu-v5e): on-demand from $1.20 (Google Cloud) to $1.20 per GPU-hour, 1 providers
- [Trainium](https://compute.fru.dev/chips/trainium): on-demand from $1.34 (AWS) to $1.34 per GPU-hour, 1 providers
- [Inferentia2](https://compute.fru.dev/chips/inferentia2): on-demand from $0.76 (AWS) to $0.76 per GPU-hour, 1 providers

## Pages

- [All prices](https://compute.fru.dev/)
- [Chips](https://compute.fru.dev/chips)
- [Providers](https://compute.fru.dev/providers)
- [AWS GPU prices](https://compute.fru.dev/providers/aws)
- [Microsoft Azure GPU prices](https://compute.fru.dev/providers/azure)
- [Google Cloud GPU prices](https://compute.fru.dev/providers/gcp)
- [Oracle Cloud GPU prices](https://compute.fru.dev/providers/oracle)
- [CoreWeave GPU prices](https://compute.fru.dev/providers/coreweave)
- [Lambda GPU prices](https://compute.fru.dev/providers/lambda)
- [Nebius GPU prices](https://compute.fru.dev/providers/nebius)
- [Crusoe GPU prices](https://compute.fru.dev/providers/crusoe)
- [Together AI GPU prices](https://compute.fru.dev/providers/together)
- [Runpod GPU prices](https://compute.fru.dev/providers/runpod)
- [DigitalOcean GPU prices](https://compute.fru.dev/providers/digitalocean)
- [Vast.ai GPU prices](https://compute.fru.dev/providers/vast)
- [Changes](https://compute.fru.dev/changes)
- [Sources and method](https://compute.fru.dev/about)

## API

- [For AI agents](https://compute.fru.dev/agents)
- [OpenAPI 3.1 spec](https://compute.fru.dev/openapi.json)
- GET https://compute.fru.dev/api/prices?chip=h100&model=on_demand: live prices, cheapest first
- GET https://compute.fru.dev/api/chips: cheapest price per chip and pricing model
- GET https://compute.fru.dev/api/providers
- GET https://compute.fru.dev/api/history?chip=h100: daily cheapest and median per provider
- GET https://compute.fru.dev/api/changes?since=YYYY-MM-DD: price moves, listed and delisted offers
- GET https://compute.fru.dev/api/companies?since=YYYY-MM-DD
- [Sitemap](https://compute.fru.dev/sitemap.xml)
- [Every live price as text](https://compute.fru.dev/llms-full.txt)