Amazon SageMaker AI pricing calculator
On-Demand Studio, Training & Inference. Estimate your costs using AWS list prices.
Estimates from AWS list prices, not an AWS quote. Discounts, taxes and credit terms may affect your bill.
01 · How Amazon SageMaker AI pricing works
What you pay for
- Instance hours per ml.* instance type for each capability (Studio apps, notebook instances, processing, training, inference, batch transform), billed per second.
- Storage: ML storage volumes, Studio EFS/EBS, model artifacts and data in S3.
- Serverless inference (per ms of compute and data processed), async inference, HyperPod clusters, Feature Store, Canvas sessions, data transfer.
- Savings: SageMaker Savings Plans and Managed Spot Training.
Common cost traps
- Real-time endpoints bill 24/7 per instance even with no traffic.
- Studio apps and notebook instances keep billing until shut down; idle GPU notebooks are a classic cost leak.
- Training jobs on large GPU instances (p4d at about $25/hour) multiply quickly with hyperparameter tuning.
- Model artifacts, checkpoints and Studio storage accumulate in S3/EFS.
Ways to reduce cost
- Enable idle shutdown for Studio apps and use lifecycle configurations.
- Use Managed Spot Training with checkpoints for interruptible training (AWS cites up to 90% savings).
- Use serverless or async inference for low or bursty traffic; multi-model endpoints for many small models.
- Buy a No Upfront SageMaker Savings Plan for steady endpoints.
Free Tier & account eligibility
Free Tier depends on account age. Accounts created on or after 15 July 2025 get a $100 sign-up credit plus up to $100 more for completing onboarding activities; on the Free plan these last 6 months (or until used up), on the Paid plan unused Free Tier credits stay usable for 12 months from sign-up. Accounts created before 15 July 2025 keep the legacy Free Tier (12-month offers plus always-free offers). Always-free offers apply on both plans. Legacy accounts: a 2-month SageMaker free trial (for example 250 hours of ml.t3.medium Studio notebooks, 50 hours of ml.m5.xlarge training and 125 hours of ml.m5.xlarge real-time inference per month). New-plan accounts pay from Free Tier credits.
02 · AWS credits & runway
Do credits cover Amazon SageMaker AI?
SageMaker AI notebooks, Studio apps, processing, training, real-time/async/serverless inference, batch transform, HyperPod, feature store, JumpStart AWS-billed usage and the hourly part of No Upfront SageMaker Savings Plans are eligible usage and are offset by credits. General rules: credits only offset charges from the billing cycle in which they were applied onward (never past bills), they apply only to the services listed on the credit in Billing and Cost Management > Credits, they expire on the date shown there, and accounts on the AWS Free plan cannot receive Activate credits until they upgrade to the Paid plan.
Check my creditsExcluded charges & conditions
- Model packages and algorithms bought from AWS Marketplace (their software fees).
- Upfront payments on SageMaker Savings Plans.
- Upfront payments for Savings Plans and Reserved Instances (All Upfront, and the upfront part of Partial Upfront). The recurring hourly fee of a No Upfront commitment is billed as normal usage and can be offset.
- AWS Marketplace charges (Activate credits make an exception for third-party foundation models used through Amazon Bedrock).
- AWS Enterprise Support, Enterprise On-Ramp, Unified Operations, resold support and partner-led support.
- Amazon Route 53 domain name registration and transfer fees.
- Taxes (sales, use, VAT and similar), AWS Professional Services, Training and Certification, Mechanical Turk, AWS Managed Services, and any use for cryptocurrency mining.
Illustrative credit runway
Single-service examples from the supplied data, not your calculator result. Months = credit amount ÷ monthly cost, before credit expiry. Credits can expire before these theoretical months are used; coverage and other AWS services can shorten runway.
| Usage profile | Monthly estimate | $10,000 | $25,000 | $100,000 |
|---|---|---|---|---|
| Small | $185.10 | 54.0 mo. | 135.1 mo. | 540.2 mo. |
| Medium | $3,186.98 | 3.1 mo. | 7.8 mo. | 31.4 mo. |
| Large | $11,685.96 | 0.9 mo. | 2.1 mo. | 8.6 mo. |
Small profile assumptions
Data scientist: ml.t3.medium JupyterLab 160 h, 40 h ml.m5.xlarge training, 1 ml.m5.xlarge endpoint 24/7 | JupyterLab ml.t3.medium: 160 hours x $0.05 = $8.00; Training ml.m5.xlarge: 40 hours x $0.23 = $9.20; Real-time endpoint ml.m5.xlarge: 730 hours x $0.23 = $167.90 | Total ~$185.10/month, us-east-1 on-demand list prices, before Free Tier, taxes and support.
Medium profile assumptions
Team: 3 x ml.g5.xlarge notebooks 160 h each, 300 h ml.g5.2xlarge training, 2 ml.g5.xlarge endpoints 24/7 | JupyterLab ml.g5.xlarge: 480 hours x $1.41 = $676.80; Training ml.g5.2xlarge: 300 hours x $1.515 = $454.50; Real-time endpoints ml.g5.xlarge: 1,460 hours x $1.408 = $2,055.68 | Total ~$3,186.98/month, us-east-1 on-demand list prices, before Free Tier, taxes and support.
Large profile assumptions
Model training on ml.p4d.24xlarge 200 h, 6 ml.g5.2xlarge endpoints 24/7 | Training ml.p4d.24xlarge: 200 hours x $25.25 = $5,050.26; Real-time endpoints ml.g5.2xlarge: 4,380 hours x $1.515 = $6,635.70 | Total ~$11,685.96/month, us-east-1 on-demand list prices, before Free Tier, taxes and support.
04 · Questions
Amazon SageMaker AI pricing FAQ
Why is my SageMaker bill high with no traffic?
Real-time endpoints and open Studio apps bill per instance-hour whether or not they are used. Delete idle endpoints and stop apps.
Do credits cover SageMaker Savings Plans?
The hourly commitment of a No Upfront plan is billed as usage and offset by credits; upfront payments are excluded.
Is Spot training worth it?
For jobs that checkpoint, yes. You pay Spot prices and SageMaker handles interruptions and restarts.
Do Marketplace models cost extra?
Yes. Marketplace model packages add software fees billed through AWS Marketplace, which promotional credits exclude.
05 · Sources & estimate notes
Content pack dated 2026-10-19. Calculator rates remain their existing snapshots; this content update is not a price refresh.
Verification notes
- Legacy SageMaker free-trial quantities are quoted from the free-tier directory and may be dated.
- Runway prices come from the calculator repository snapshot rate card sagemaker-ai (no AWS Price List mapping in the build). Confirm the main unit prices on https://aws.amazon.com/sagemaker-ai/pricing/ before publishing.