Amazon Redshift、Gravitonベース RGインスタンスとデータレイク統合クエリエンジンを発表 Amazon Redshift introduces AWS Graviton-based RG instances with an integrated data lake query engine

AWS has announced a new generation of Amazon Redshift compute, the RG instance family, powered by AWS Graviton processors, along with an integrated query engine for directly accessing data lake content. The combination targets better price-performance over the existing RA3 family while blurring the line between warehouse and lake.

RG instances are built on Graviton, AWS's Arm-based silicon that the company has been steadily rolling out across its data services portfolio, including Amazon RDS, Aurora, OpenSearch, and EMR. Graviton is generally positioned for improved performance-per-dollar and power efficiency, and Redshift customers running large analytic workloads may see meaningful cost optimization, though actual gains will depend on workload characteristics.

The more strategically significant piece is the integrated data lake query engine. It allows Redshift to query open table formats such as Apache Iceberg stored in Amazon S3, working with the AWS Glue Data Catalog and Amazon S3 Tables. Rather than ingesting or copying data into the warehouse, users can run high-performance SQL directly over lake data. This evolves the longstanding Redshift Spectrum capability into something closer to a first-class lakehouse experience, simplifying ETL and reducing data duplication.

The announcement lands in a competitive context. Snowflake has been investing heavily in Iceberg support and externally managed tables, while Databricks continues to push its lakehouse architecture with Unity Catalog and Delta Lake. AWS appears to be positioning Redshift not just as a traditional MPP warehouse but as a broader analytics platform that can natively span warehouse-managed and lake-managed data. Combined with prior moves like Redshift Serverless and the growing set of zero-ETL integrations with Aurora, RDS, and DynamoDB, the direction is clearly toward reducing operational friction between transactional sources, the warehouse, and open lake storage.

Customers currently on RA3 will likely want to evaluate RG as a migration target once availability and pricing details are confirmed in their regions. Organizations already standardizing on Iceberg for their lake should find the new query engine particularly compelling, as it offers an additional engine option alongside Athena, EMR, and third-party tools without requiring data movement. As always with newly launched instance families, real-world benchmarking against representative workloads will be the best way to validate the price-performance claims.