Amazon Elastic Inference
Attachable GPU for inference
Discontinued (April 2024); replaced by AWS Inferentia. Elastic Inference used to let you attach a fraction of GPU capacity to EC2 or SageMaker to accelerate inference at a lower cost than a full GPU instance.
- Cost$ Low
- Difficulty●●●●●
- Costs money when idle?No
- Managed by AWS
⚠️ Retired or retiring service
Open Elastic Inference on the map →
Discontinued (April 2024); use AWS Inferentia
Think of it asLike renting an extra bit of engine to bolt onto your car: handy back then, but the manufacturer has stopped making it.
When to use it
✓ Use it when
Can no longer be used.
✗ Avoid it when
Low-cost inference → Inferentia instances with Neuron / SageMaker AI.
Common uses
- Migrate legacy inference workloads to Inferentia
What does it cost?
- Pricing model: The service is discontinued; no new charges.
