AWS · ai
AWS Inferentia2 cloud pricing
Compare every cloud instance with AWS Inferentia2 accelerators across AWS: 4 instance types with 1 to 12 Inferentia2s. Cheapest on demand: $0.7582 per Inferentia2-hour on AWS inf2.xlarge in US East (N. Virginia).
- on-demand / Inferentia2-hour
- $0.7582
- spot / Inferentia2-hour
- $0.0834
- memory per device
- 32 GB
- instance types
- 4
AWS inf2.xlarge
AWS inf2.xlarge
1 cloud
Every AWS Inferentia2 instance
cheapest region per instance · Linux · on-demand per GPU-hour = hourly price ÷ GPU count · government regions excluded
| Provider | Instance | GPUs | GPU memory | vCPU | RAM | On-demand /hr | Per GPU-hr | Spot per GPU-hr | Cheapest region |
|---|---|---|---|---|---|---|---|---|---|
| AWS | inf2.xlarge | 1 | 32 GiB | 4 | 16 GiB | $0.7582 | $0.7582 | $0.0834 | US East (N. Virginia) us-east-1 |
| AWS | inf2.24xlarge | 6 | 192 GiB | 96 | 384 GiB | $6.491 | $1.082 | $0.121 | US East (N. Virginia) us-east-1 |
| AWS | inf2.48xlarge | 12 | 384 GiB | 192 | 768 GiB | $12.98 | $1.082 | $0.1203 | US East (N. Virginia) us-east-1 |
| AWS | inf2.8xlarge | 1 | 32 GiB | 32 | 128 GiB | $1.968 | $1.968 | $0.2195 | US East (N. Virginia) us-east-1 |
Other GPUs
B300 $17.80B200 $14.24RTX PRO 6000 Blackwell $0.5596H200 $7.912Trainium2 $6.447Radeon Pro V710 $1.994H100 NVL $6.98L40S $1.861L4 $0.7045RTX 4000 Ada $0.52Qualcomm Cloud AI 100 $1.115H100 $6.88Trainium $1.344Habana Gaudi $1.639Radeon Pro V620 $2.54A10G $1.006A10 $2.20T4G $0.42A16 $0.471A100 80GB $2.397A40 $1.712A100 40GB $2.745Radeon Pro V520 $0.3785Inferentia $0.228Quadro RTX 6000 $1.50T4 $0.526V100 $3.06P40 $2.07P100 $2.07M60 $0.75K80 $0.90GRID K520 $0.65
AWS Inferentia2 FAQ
How much does an Inferentia2 cost per hour in the cloud?
On demand, from $0.7582 per Inferentia2-hour (AWS inf2.xlarge, $0.7582 per hour for 1 Inferentia2 in US East (N. Virginia)). Spot capacity starts at $0.0834 per Inferentia2-hour on AWS inf2.xlarge.
Which clouds offer Inferentia2 instances?
AWS (from $0.7582 per GPU-hour, e.g. inf2.xlarge).
How much memory does the Inferentia2 have?
32 GB per device. Cloud instances combine 1 to 12 of them.