The news
Amazon Web Services, the cloud unit of Amazon (AMZN), said on September 30, 2026, that its Aurora PostgreSQL database can now query Apache Iceberg and Parquet files in a company's data lake directly. Iceberg and Parquet are open formats widely used to store large analytical datasets in Amazon S3.
The feature works by embedding DuckDB, an open-source analytical query engine, inside Aurora. According to AWS, analytical scans run within the database, so a single query can combine live operational rows with data lake files without extra network hops or ETL (extract, transform and load) pipelines that duplicate data.
SiliconANGLE reported that the queries can include uncommitted writes in the live database. AWS's example combined seven days of customer transactions held in Aurora with five years of historical transactions stored in S3.
The capability is generally available in all commercial AWS Regions and in AWS GovCloud (US), on Aurora PostgreSQL 17.11 and later and 18.6 and later, AWS said. It supports Iceberg tables through the AWS Glue Data Catalog and catalogs compatible with the Iceberg REST Catalog specification, plus S3 and S3 Tables. Customers turn it on with an extension called aurora_analytics and an IAM role that grants access to S3 and Glue.
AWS charges nothing extra for the feature itself. Customers pay for the incremental Aurora compute their queries consume and for S3 requests to read the files. AWS said workloads that need single-digit-millisecond latency should still copy data into native Aurora tables. SiliconANGLE also reported that AWS acquired DuckDB's developer in August 2026.
The numbers
- Supported Aurora PostgreSQL versions
- 17.11+ and 18.6+
- Additional feature charge
- $0 (pay for compute and S3 requests)
- AWS example: live data window joined with history
- 7 days in Aurora + 5 years in S3
Why CEOs should care
For CIOs and data leaders, the main gain is fewer moving parts. Many companies copy transaction data out of production databases into a warehouse each night, which adds cost, lag and another system to secure. If teams can answer some questions inside Aurora, ask which ETL jobs exist only to support reports that could now run in place, and whether retiring them lowers spending.
CFOs should look carefully at the pricing model. The feature is free, but heavy analytical queries consume Aurora compute and S3 requests, and those charges grow with use. Before moving dashboards over, ask the data team to estimate query volume and test it against current warehouse bills. Also ask whether large scans could slow the transaction workloads that share the same database.
CISOs and architects get a simpler data flow, though not a smaller responsibility. Queries reach into S3 and Glue through an IAM role, so access rules need review. AWS's own framing points at AI agents: its solutions architect Esra Kayabali said, as quoted by SiliconANGLE, that agents make it impractical to predict and pre-copy every dataset they might need. Leaders should decide which lake tables agents connected to operational databases are allowed to read.
The bigger picture
The launch fits a broad push to blur the line between transaction databases and analytics. Iceberg has become a common table format across cloud data platforms, and putting an analytical engine next to the operational database targets the work that warehouses and lakehouse products have long handled. AWS did not compare the feature with Snowflake or Databricks, and specialized warehouses still offer scale and features a single database instance does not. For many mid-sized workloads, though, buyers now have another option to price.
Using DuckDB also signals how much weight AWS puts on open-source engines. DuckDB has gained a following for fast analytics on files, and bundling it into a managed database lets AWS offer it without asking customers to run it themselves.
What’s next
Watch for AWS to publish performance guidance and customer results, and for whether it extends the approach to other databases. Buyers renewing warehouse contracts should test their heaviest Aurora-linked reports on the new feature first. Teams should also track query metrics the feature exposes, such as rows scanned, bytes read from S3 and cache hits, which SiliconANGLE said are available. Those numbers will show whether moving reports into Aurora actually saves money or simply shifts the bill from the warehouse to the database.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
