AWSAmazon AthenaAnalytics

`aws_athena_workgroup` cost estimation

Serverless query engine for S3 data. $5 per TB scanned (standard). Provisioned capacity option starts at $0.30/DPU-hour with 24-hour minimum. Workgroups themselves are free; the queries within bill.

Amazon Athena queries data in S3 using SQL. The aws_athena_workgroup is a container for queries with shared settings (data limits, result locations, encryption). The workgroup itself has no fee, queries run within the workgroup bill.

Two pricing models exist:

Standard (per-query): $5 per TB of data scanned. Data scanned is the amount Athena reads from S3 to satisfy the query, after partition pruning. A query over 100 GB scans costs $0.50.

Provisioned capacity: $0.30 per DPU-hour with a 24-hour minimum commitment. Useful for predictable, high-volume query workloads where the per-TB rate would be expensive. Break-even: roughly 14 TB scanned/day to make provisioned cheaper than standard.

Common cost surprises: - SELECT * scans every column. Use specific column lists to read less. - Unpartitioned data scans the entire table. Partition by date/region. - Text formats (CSV, JSON) scan more bytes than columnar (Parquet, ORC). Migrate hot tables to Parquet for 10-100x reduction. - Failed queries don't bill, but cancelled queries do (for data already scanned). - DML queries (INSERT, CTAS) bill both scan and write costs.

c3x estimates Athena based on declared workgroup configuration. For accurate cost, specify query volume (TB scanned/month) in c3x-usage.yml, Terraform doesn't know how much you'll query.

Terraform example

A minimal but realistic configuration that C3X can estimate.

resource "aws_athena_workgroup" "analytics" {
  name = "analytics-team"

  configuration {
    enforce_workgroup_configuration    = true
    publish_cloudwatch_metrics_enabled = true

    result_configuration {
      output_location = "s3://athena-results-prod/output/"

      encryption_configuration {
        encryption_option = "SSE_S3"
      }
    }

    bytes_scanned_cutoff_per_query = 10737418240  # 10 GB hard cap per query
  }
}

Pricing dimensions

What you actually pay for when you provision aws_athena_workgroup.

Dimension	Unit	What's being charged
Data scanned (standard)	per TB	Amount of data Athena reads from S3 to satisfy queries. Pay-per-query model. $5/TB scanned
Provisioned capacity	per DPU-hour	Reserved capacity for predictable workloads. 24-hour minimum commitment. Break-even at ~14 TB/day. $0.30/DPU-hour
Federated query (Lambda)	per invocation	When queries span non-S3 sources via federated connectors, the underlying Lambda invocations bill separately. Standard Lambda pricing
Result storage (S3)	per GB-month	Query results are stored in S3 (configured per workgroup). Standard S3 pricing applies. $0.023/GB-month

Optimization tips

Common ways to reduce aws_athena_workgroup cost without changing the workload.

Convert tables to Parquet

50-90% per query

Parquet is columnar, so SELECT col1, col2 only reads col1/col2 bytes from S3. Typical reduction: 50-90% of bytes scanned vs CSV/JSON. Use AWS Glue or a one-off CTAS to convert.

Partition aggressively

10-1000x reduction

Partition tables by frequently-filtered columns (date, region, customer_id). Queries with WHERE clauses on partitions only scan matching partitions. A daily partitioned 5-year table queried for last 7 days scans 0.4% of total data.

Set bytes_scanned_cutoff_per_query

Prevents runaway costs

Prevents accidental TB-scale queries. Set a 10 GB or 100 GB cap depending on workgroup purpose. Queries that would exceed the limit fail before billing.

Avoid SELECT *

30-90% per query

Athena scans every column referenced in the query. Listing specific columns reduces scan bytes proportionally. With Parquet, the reduction is substantial.

Consider provisioned capacity for steady high volume

20-40% on high-volume workloads

Above ~14 TB scanned/day, provisioned capacity at $0.30/DPU-hour can be cheaper. Calculate exactly: scan TB/day × $5 vs DPU count × $0.30 × 24.

FAQ

Does the Athena workgroup itself cost money?

No. Workgroups are organizational containers, they cost nothing on their own. Queries running within the workgroup bill at the standard or provisioned rate.

Do failed queries bill?

Failed queries (syntax errors, schema mismatches) don't bill, they fail before scanning data. Cancelled queries DO bill for the data already scanned before cancellation. Queries that hit the bytes_scanned_cutoff fail without billing.

How does Athena pricing compare to Redshift Spectrum?

Both query S3 data and charge similar per-TB rates ($5/TB). Spectrum requires a Redshift cluster (additional cost) but offers tighter integration if you already have one. Athena is fully serverless.

Can I set a budget cap on Athena queries?

Per-query: yes, via bytes_scanned_cutoff_per_query in the workgroup. Per workgroup, you can set data limits with alerts. AWS Budgets can also alert when monthly Athena spend exceeds a threshold.