Key Takeaways
- AWS Glue 6.0 is now generally available at 30% lower cost with full Apache Iceberg v3 support.
- New runtime features include Spark 4.1, Python 3.12, the VARIANT data type, and nanosecond-precision timestamps.
- The upgrade strengthens open table formats with faster real-time streaming and deeper Iceberg integration.
Table of Contents
AWS Glue 6.0 Reprices Serverless Spark and Completes Iceberg v3
AWS Glue 6.0 is now generally available in every region where AWS Glue operates, carrying a 30% lower price than previous versions and full Apache Iceberg v3 support.
The release moves serverless Spark onto Apache Spark 4.1, Python 3.12, and Scala 2.13, a modernization that AWS describes as faster across ETL, PySpark, and streaming workloads.
AWS News Blog positions the rollout as a milestone for open table performance on managed infrastructure.
AWS Glue provides the most complete Iceberg v3 implementation on any fully serverless managed Spark service.
As reported in the AWS News Blog, the price cut lands while data teams are reassessing open storage formats, making the delivery of Iceberg v3 features alongside lower unit economics especially timely.
Inside the Iceberg v3 and Spark 4.1 Runtime Upgrade
The most consequential change is the complete Apache Iceberg v3 specification built on Iceberg 1.11.0.
At its center is the VARIANT data type with shredding support, which accelerates query reads against semi-structured data compared with traditional string columns.
Teams can store and query JSON, logs, and event data without flattening schemas, removing duplicate data copies, custom parsing code, and pipeline breakage when schemas evolve.
Iceberg v3 also brings several additional capabilities into the Glue runtime.
- Geometry and geography data types: native spatial processing for GIS analytics, location intelligence, and geospatial pipelines.
- Nanosecond-precision timestamps: support for IoT sensor data, scientific computing, and high-frequency financial workloads that need precision beyond milliseconds.
- Unknown type handling: resilience against upstream schema changes so unexpected or evolving data does not stop pipelines.
On the Spark 4.1 side, declarative pipelines simplify ETL authoring by letting engineers declare transformations while the engine determines execution order and optimization.
Arrow-native Python UDFs and UDTFs remove serialization overhead between Python and the JVM, improving PySpark performance for complex transformations.
For stateless streaming use cases, a new real-time streaming mode achieves single-digit millisecond latency on Glue-optimized execution.
Existing APIs remain unchanged; users can select version 6.0 through the existing glue-version parameter in create-job or update-job calls, Glue Studio, SageMaker Unified Studio, or a preferred IDE.
The Open Table Race Moves Deeper Into Real-Time Territory
This release sharpens the competitive pressure on proprietary warehouse formats in three ways.
First, Iceberg v3 has materially narrowed the operational gap with warehouse-grade performance.
The specification mandates binary deletion vectors stored in Puffin files, permits at most one deletion vector per data file in a snapshot, and deprecates position delete files.
Those changes pair with row-level deletes and advanced compaction to make streaming ingestion through Flink or Spark Structured Streaming far more practical for enterprise workloads.
AWS had already shipped v3 deletion vectors and row lineage across Athena, Glue, EMR, S3 Tables, and SageMaker in November 2025.
Snowflake reports a twofold query improvement for Iceberg Tables compared with its older external tables, a vendor-reported figure that underscores how much open table performance has advanced.
The broader field agrees directionally: AWS, Snowflake, Databricks, and Google now support Iceberg natively, turning proprietary warehouses into compute engines over open storage.
That shift makes enterprise format decisions hinge less on table format alone and more on query engines, catalogs, and governance controls.
The competitive tension with Delta Lake is no longer a simple format war.
It now revolves around which engine can execute most efficiently against the same open metadata layer, and AWS Glue 6.0 is a direct push toward winning that execution layer on serverless Spark.
What the Price Cut and Variant Support Change for Data Teams
For enterprise data teams, AWS Glue 6.0 collapses much of the distance between serverless Spark economics and warehouse-grade Iceberg operations.
The remaining decision is execution speed and governance fit, not format lock-in.
For data platform teams translating this shift into search-visible authority content, programmatic SEO AI automation is how Andres SEO Expert approaches it — contact the team.
Frequently Asked Questions
How much cheaper is AWS Glue 6.0 compared to previous versions?
AWS Glue 6.0 is priced 30% lower than previous versions while bringing full Apache Iceberg v3 support and faster performance for ETL, PySpark, and streaming workloads.
What is the new VARIANT data type in Iceberg v3 and how does it help?
VARIANT is a data type that supports semi-structured data like JSON, logs, and event data without flattening schemas. It accelerates query reads, removes duplicate copies, custom parsing code, and prevents pipeline breakage during schema evolution.
What are the key runtime upgrades in AWS Glue 6.0?
AWS Glue 6.0 runs on Apache Spark 4.1, Python 3.12, and Scala 2.13. Spark 4.1 introduces declarative pipelines, Arrow-native Python UDFs and UDTFs, and a new real-time streaming mode with single-digit millisecond latency.
Which Iceberg v3 features are included besides VARIANT?
Iceberg v3 in AWS Glue 6.0 includes geometry and geography data types for spatial processing, nanosecond-precision timestamps for high-frequency workloads, and unknown type handling for resilience against upstream schema changes.
How does AWS Glue 6.0 improve real-time streaming?
With Spark 4.1, a new real-time streaming mode for stateless workloads achieves single-digit millisecond latency on Glue-optimized execution, making streaming ingestion more practical for enterprise workloads.
Can I still use existing AWS Glue APIs and tools with version 6.0?
Yes, existing APIs remain unchanged. You can select Glue 6.0 through the glue-version parameter in create-job or update-job calls, Glue Studio, SageMaker Unified Studio, or a preferred IDE.
Why is the open table race shifting toward execution engines rather than formats?
AWS, Snowflake, Databricks, and Google now support Iceberg natively. The competitive tension with Delta Lake is no longer a simple format war; it now revolves around which engine can execute most efficiently against the same open metadata layer, and AWS Glue 6.0 is a direct push toward winning that execution layer on serverless Spark.
