Amazon Athena

Serverless SQL queries directly against data stored in Amazon S3

AI Data AnalysisOverseasβ˜…β˜…β˜…β˜†β˜† 3.0

What is Amazon Athena?

Amazon Athena is a serverless query service that runs SQL against data sitting in Amazon S3 or other supported sources. There is no cluster to size or maintain: you point a table definition at a prefix, and Athena queries the files directly, including CSV, JSON, Parquet, ORC and Avro. It is built on a Trino and Presto lineage, so familiar SQL works, and integration with the Glue Data Catalog means tables can be shared with other AWS analytics services. Billing is per query by data scanned, which makes Parquet partitioning and compression an economic decision rather than just a performance one.

Last updated: 2026-09-20. This site only provides an index; for exact features, pricing, and licensing, see the official website.

Key features

  • Serverless SQL over files in S3 with no cluster to manage
  • Reads CSV, JSON, Parquet, ORC and Avro directly from storage
  • Glue Data Catalog integration shares table definitions across AWS
  • Federated queries that reach relational and NoSQL sources
  • Athena for Apache Spark for notebook-style workloads
  • Workgroups with per-team query limits and cost controls

Pros & cons

Strengths

  • No infrastructure to provision and nothing to keep running
  • Query cost is transparent and tied to bytes scanned
  • Fits naturally into an existing S3-based data lake

Watch out for

  • Poorly organised files can make each query surprisingly expensive
  • Cold-start latency on the first query of a session
  • Not designed for interactive dashboards with heavy concurrency

Best for & use cases

log analysis, ad hoc s3 querying, data lake exploration and cost-aware reporting

If you're comparing similar products, check the alternatives below, or browse all tools in the AI Data Analysis category.

FAQ

Why does the same query cost different amounts on different days?

Athena bills by bytes scanned, so a query against a daily partition costs far less than one scanning a year of files. Converting to Parquet with partitions and selecting only needed columns can cut a bill by an order of magnitude.

Is Athena an AI tool?

No. It is a conventional SQL engine over object storage. Machine learning work would use services such as SageMaker or Athena's Spark integrations for feature engineering before training elsewhere.

When is a warehouse a better choice than Athena?

When many users run concurrent dashboards, when workloads need consistently fast interactive response, or when complex joins repeat constantly. Athena suits ad hoc exploration and log analysis where infrastructure-free operation matters more.