Tool profile
AWS Step Functions: cost, fit and limits
AWS Step Functions is a serverless workflow service made by Amazon Web Services, Inc. that coordinates AWS services such as Lambda, Glue, ECS and SageMaker into state machines, billed per state transition for Standard workflows or per request and duration for Express workflows.
What to know first
AWS Step Functions bills Standard workflows at $0.000025 per state transition in the US East (N. Virginia) example, with 4,000 free transitions a month that do not expire. Express workflows are billed at $1.00 per million requests plus duration, with no free tier stated. The pages name no AI feature beyond a SageMaker integration, and they state no hosting region for your data.
AWS Step Functions at a glance
| Maker | Amazon Web Services, Inc., Not stated on the vendor pages |
|---|---|
| Free plan | Yes |
| Entry price | $0.000025 per state transition (Standard) |
| API access | Yes |
| Category | Best AI Data Orchestration Tools in 2026, rank 9 of 12 |
| Website | aws.amazon.com/step-functions/ |
Who AWS Step Functions is best for
Teams already on AWS
Lambda, ECS, EKS, Fargate, DynamoDB, SQS, SNS, Glue, EMR and SageMaker are listed integrations, so most steps need no glue servers.
High-event-rate workloads
Express Workflows handle event rates above 100,000 per second, which suits streaming or ingestion triggers that Standard workflows would price poorly or throttle.
Bursty or idle pipelines
Standard workflows charge nothing for idle time, so a weekly batch costs only its transitions, not a server that sits waiting.
AWS Step Functions plans and pricing
| Plan | Price | What it includes |
|---|---|---|
| Standard Workflows | $0.000025 per state transition | 4,000 state transitions a month free with no 12-month expiry. You do not pay for idle time, and retries count as transitions. |
| Express Workflows | $1.00 per million requests plus $0.00001667 per GB-second (example) | Duration is rounded up to 100 ms and billed in 64 MB chunks. No free tier is stated for Express. |
| Express tiered duration example | $0.06 per GB-hour for the first 1,000 GB-hours | The next 4,000 GB-hours cost $0.03 and anything beyond costs $0.01642, per the pricing page example. |
Checked November 2026.
What AWS Step Functions does
- Map state at scale
- Map iterates over S3 objects and can launch thousands of parallel workflows, which suits fan-out jobs such as processing every file in a bucket or partition.
- Built-in error handling
- Try/catch and retry logic is built in for failed or timed-out tasks, so retry rules are declared in the state machine rather than coded in each Lambda.
- Long-lived state
- State can persist for up to one year, which lets a workflow wait on a human approval or a slow upstream system without holding compute open.
- VPC endpoints
- AWS PrivateLink provides VPC endpoints, so calls from private subnets need not cross the public internet, which many security teams require for regulated data workloads.
- Wide service reach
- Integrations named on the features page include Lambda, ECS, EKS, Fargate, DynamoDB, SQS, SNS, Glue, EMR and SageMaker, covering most AWS data and compute services.
- Pay-per-use model
- Standard bills per state transition, metered daily and billed monthly, so costs track real workflow complexity and retries rather than provisioned capacity or idle servers.
How AWS Step Functions handles your data
The pricing and features pages for AWS Step Functions do not state whether customer data trains models, how long execution data is retained, which regions your data stays in or which certifications cover the service. They state that VPC endpoints are available through AWS PrivateLink and that state can persist for up to one year. Standard workflows meter state transitions daily and bill monthly. For compliance questions, use AWS's own compliance documentation, which was not part of this review. The vendor's policy was not checked for this.
Where AWS Step Functions falls short in practice
- Cloud lock-in. State machines call AWS services, so moving a workflow to another cloud means rewriting its steps and triggers, not exporting a definition file, and that cost grows with every workflow added.
- No Express free tier. The pricing page states no free tier for Express Workflows, and its memory estimate adds 50 MB plus definition and payload sizes, which makes cost modeling fiddly.
- Not a data-asset tool. Pages describe state machines and service calls, with no lineage, asset catalog or data-quality view, and no AI feature other than a SageMaker integration on the pages opened.
Works with
Alternatives to AWS Step Functions
AWS Step Functions: common questions
How much does AWS Step Functions cost?
Standard workflows cost $0.000025 per state transition in the US East example, so a million transitions is $25 after the free allowance. Express workflows cost $1.00 per million requests plus duration charges. Rates vary by region, so check the pricing page for yours.
Is there a free tier for Step Functions?
Standard workflows include 4,000 free state transitions a month that do not expire after 12 months. The pages state no free tier for Express workflows. Retries count as transitions, so a flaky step can use the free allowance faster than expected.
When should I choose Express over Standard?
Express suits high-volume, short workflows: the features page cites event rates above 100,000 per second. Standard suits longer workflows that persist state for up to a year. Express is billed on requests and duration, Standard on transitions, so model both.
Does Step Functions have AI features?
The pages opened name no AI feature for Step Functions. The nearest item is an integration with Amazon SageMaker for machine learning services. If you need a copilot or agent runtime inside the orchestrator, compare tools such as Kestra, Dagster or Astronomer.
Can Step Functions loop over files in S3?
Yes. The features page says the Map state can iterate over S3 objects and launch thousands of parallel workflows. Concurrency limits and per-execution quotas were not given on the pages opened, so check the service quotas documentation before processing very large buckets.