A10G cloud GPU prices: instances and history

A10G 24 GB. On 2026-11-13, the lowest documented on-demand rate is USD 1.0060 per GPU-hour: AWS g5.xlarge in us-east-1 (USD 1.0060 per machine-hour for 1 GPU). Lowest spot: USD 0.3348 per GPU-hour (AWS g5.24xlarge, us-west-2), subject to interruption.

Snapshot 2026-11-13 10:42:15 UTC. Lowest among the documented regions, not a capacity guarantee.

Instances and regions

g5.12xlarge · g5.16xlarge · g5.24xlarge · g5.2xlarge · g5.48xlarge · g5.4xlarge · g5.8xlarge · g5.xlarge

A10G 24 GB daily history

32 recorded days, 2026-10-13 to 2026-11-13. Each day uses its own coverage; the cheapest instance or region can change. Missing dates are not interpolated.

What you pay for

On-demand and spot stay separate, as do GPU memory variants. A machine-hour is the recorded per-GPU rate multiplied by GPU count; per-GPU rates are rounded to four decimals, so reconstructed machine prices can differ slightly from the original quote. CPU and RAM specifications for AWS and Azure were read on 18 October 2026; Google shares are in the snapshot.

instance_share includes a share of the machine's vCPUs and system RAM. Google Cloud adds the GPU, vCPU and RAM SKU shares; AWS and Azure divide the instance rate by its GPU count. spot is an observed AWS Linux spot rate, also divided by GPU count; its date is an observation date. Other dates are the Azure price start, AWS list publication or newest Google component effective date.

Storage, local SSD, licences, networking, tax and discounts are excluded. Spot can be reclaimed. GPU memory and counts come from the dated price snapshot; verify region, zone and capacity with the provider before renting. Coverage is limited to the instances recorded here, not the whole rental market.

Related: cloud GPU overview, GPU price API, daily snapshots, rent vs buy.

Automated list prices, not a quote. All guides.