Amazon SageMaker AI Pricing Calculator
Estimate Studio, training, and real-time inference monthly spend before committing to a model training plan.
- No signup
- Instant results
- ~3 min
Rates reviewed
Read the full breakdown — AWS SageMaker AI Savings Plans covers billing dimensions, traps, and a 30-day cleanup plan.
Workload type and compute
us-east-1list rates, pricing as of 2026-06-25. Feature Store API, endpoint autoscaling overhead, and cross-region data transfer excluded. Source: AWS public pricing.
Who This Tool Is For
ML platform leads and CTOs planning SageMaker environments for training and production inference.
Why We Built This Tool
SageMaker bills across multiple dimensions, and teams often under-model training runtime and storage growth in early budget plans.
What Problem It Solves
- Compute mismatch. Oversized GPU instances stay attached to notebook and endpoint workflows.
- Underestimated storage. Model artifacts and feature data grow faster than expected.
- No Spot strategy. Training jobs miss obvious savings when interruption-tolerant workloads run on On-Demand only.
Frequently Asked Questions
Does this cover all SageMaker products?
This calculator focuses on common Studio, training, and inference cost drivers for planning decisions.
How are Spot savings modeled?
For training workloads with Spot enabled, this estimate applies a 30% compute discount as a planning baseline.
Can I compare SageMaker to Bedrock here?
Use this for SageMaker-side cost, then compare with the Bedrock calculator for token-driven workloads.
What storage is included?
EBS storage for instances and optional Feature Store offline storage assumptions are included.
Does endpoint autoscaling change cost?
Yes. Actual cost depends on scaling behavior and traffic shape; this model assumes fixed average runtime.
Is this suitable for annual budgeting?
Yes as a baseline model. Validate with Cost Explorer after launch and refine monthly.