Cheapest cloud GPU right now: price per GPU-hour by chip and region

Lowest on-demand list price per GPU-hour on 2026-11-13 12:06 UTC: H100 USD 6.8800 per GPU-hour (AWS us-east-1); A100 80GB USD 3.4309 per GPU-hour (AWS us-east-1); T4 USD 0.5260 per GPU-hour (AWS us-east-1). Every chip and region is listed separately below.

Page generated 2026-11-13 12:06 UTC, from the providers' public price lists

H200

GPU memory per GPU: 141 GB. Cheapest on-demand first.

H100

GPU memory per GPU: 80 to 94 GB. Cheapest on-demand first.

A100 80GB

GPU memory per GPU: 80 GB. Cheapest on-demand first.

A100 40GB

GPU memory per GPU: 40 GB. Cheapest on-demand first.

L40S

GPU memory per GPU: 48 GB. Cheapest on-demand first.

L4

GPU memory per GPU: 24 GB. Cheapest on-demand first.

A10

GPU memory per GPU: 24 GB. Cheapest on-demand first.

A10G

GPU memory per GPU: 24 GB. Cheapest on-demand first.

T4

GPU memory per GPU: 16 GB. Cheapest on-demand first.

Rent or buy?

The API can do this sum for an agent: add a rent_vs_buy object to the same call with the purchase price, watts per GPU, electricity price and hours per month. It returns the cheapest matching cloud row, the monthly cost of renting and of owning, and the first month the purchase breaks even. Same price per call.

Worked example from today's snapshot: the cheapest on-demand H100 is AWS us-east-1 at USD 6.8800 per GPU-hour. Running one for 500 hours a month costs USD 3,440.00 in the cloud. A USD 30,000 card at 700 W and USD 0.15 per kWh costs USD 52.50 a month in electricity, so the purchase pays for itself in month 9 (no resale, PUE 1.0).

It excludes cooling beyond PUE, hosting, networking, maintenance, financing and tax, and uses cloud list prices, not discounts. Not advice.

How this is counted

• Official list and spot prices from Azure Retail Prices, AWS Price List, AWS EC2 Spot and Google Cloud Billing Catalog. Negotiated discounts, savings plans, committed use, reservations and credits are not included.
• Marketplaces such as Vast or RunPod are not included.
• Spot prices (AWS, Azure and Google Cloud; Google calls them Spot Preemptible) are for capacity that the provider can reclaim at short notice. They are not a price for a job that must finish.
• Prices exclude storage, network egress, licences and any support plan.
• Price basis (instance_share; AWS Spot rows: spot): AWS and Azure rows are the instance price divided by the number of GPUs in that instance, so they include the instance's vCPUs and memory. Google Cloud sells the GPU, vCPUs and memory as separate items, so a Google row is the GPU price plus the per-GPU share of the vCPU and memory prices of the named machine type (A2, A3, G2, or an N1 machine for T4). Local SSD bundled with some Google machine types, boot disks and licences are not included. Google regions are GCP region names. Google rows are limited to the regions Google lists for that GPU (its GPU regions and zones page); zones vary inside a region, so confirm the zone when you launch.
• For each provider, region and chip the page lists the lowest on-demand size; AWS Spot has a separate lowest-spot row. latest.json carries every size. AWS Spot dates are observation dates, not price-change times; refreshes are daily.
• No price is blended or averaged across chips, sizes, regions or providers.
• Dates: Azure price start, AWS list publication, AWS Spot observation, Google newest SKU effective date.
• Azure: Retail Prices API, Linux virtual machines, pay-as-you-go and Spot, regions eastus, westus2, westeurope, southeastasia, southcentralus. Status: ok, checked 2026-11-13 12:06 UTC.
• AWS: Price List bulk file for Amazon EC2, on-demand Linux with shared tenancy, regions us-east-1, us-west-2, eu-west-1, ap-southeast-1. Status: ok, price list published 2026-11-10. The AWS files are large, so they are read once a day.
• Google Cloud: Cloud Billing Catalog API, Compute Engine, USD, on-demand and Spot Preemptible, machine types a2-highgpu-1g, a2-ultragpu-1g, a3-highgpu-8g, a3-ultragpu-8g, g2-standard-4, g2-standard-48, n1-standard-4 + 1x T4, n1-standard-16 + 4x T4. Status: ok, checked 2026-11-13 12:06 UTC.
• AWS EC2 Spot: Linux USD hourly instance prices, divided by GPUs. Status: ok; last read 2026-11-13T10:42:46Z.
• Azure sizes left out because the GPU count is fractional or not in Tanod's size table: NC16ads_A10_v4, NC32ads_A10_v4, NC8ads_A10_v4, NCC40ads_H100_v5, ND96is_flex_H100_v5, ND96is_noIB_H100_v5, ND96isrf_H100_v5, NV12ads_A10_v5, NV18ads_A10_v5, NV6ads_A10_v5.
• The same rows are in latest.json. Browse the daily JSON history. Every daily snapshot since 13 October 2026 is a free dataset on Hugging Face: tanod/cloud-gpu-list-prices (CC BY 4.0).

Related: H100 price per hour, cheapest chain to send USDC, where an agent can get paid today, agent API prices by category. Back to tanod.dev or the guide index. Informational only, not advice. Results are automated.