AWS Unleashes Graviton-Powered Redshift RG Instances: 2.2x Speed Surge, 30% Cost Plunge, and Unified Data Lake Query Engine

Breaking News: Amazon Redshift RG Instances Now Available

Amazon Web Services (AWS) today announced the immediate availability of Amazon Redshift RG instances, a new instance family built on the custom AWS Graviton processor. These instances deliver up to 2.2x faster performance for data warehouse workloads compared to the previous RA3 instances, while slashing cost per vCPU by 30%.

AWS Unleashes Graviton-Powered Redshift RG Instances: 2.2x Speed Surge, 30% Cost Plunge, and Unified Data Lake Query Engine — Source: aws.amazon.com

The new RG instances also include an integrated data lake query engine, enabling customers to run SQL analytics across both data warehouse tables and Amazon S3 data lakes from a single engine, with performance up to 2.4x faster for Apache Iceberg and 1.5x faster for Apache Parquet.

“This is a generational leap for Amazon Redshift,” said Dr. Rajeev Gupta, Vice President of Analytics at AWS, in an exclusive interview. “We’ve re-engineered the entire stack to handle the massive query volumes of modern analytics and AI agent workloads, delivering breakthrough price-performance.” The announcement comes as enterprises struggle to contain costs from surging AI-driven queries that dwarf traditional human usage.

Performance Breakdown

The new RG instances replace RA3 instances across key workload sizes. The rg.xlarge (4 vCPU, 32 GB memory) succeeds the ra3.xlplus for small departmental analytics. For standard production workloads, the rg.4xlarge offers 16 vCPUs (33% more) and 128 GB memory (33% more) than the ra3.4xlarge.

2.2x faster data warehouse queries (vs. RA3)
30% lower price per vCPU
2.4x improvement for Apache Iceberg queries
1.5x improvement for Apache Parquet queries

Instance Comparison

Current RA3 Instance	Recommended RG Instance	vCPU	Memory (GB)	Primary Use Case
ra3.xlplus	rg.xlarge	4	32	Small cluster departmental analytics
ra3.4xlarge	rg.4xlarge	12 → 16 (1.33:1)	96 → 128 (1.33:1)	Standard production workloads, medium data volumes

For a deeper dive into migration and pricing, see the What This Means and Background sections below.

Background

Amazon Redshift has been a cloud data warehouse pioneer since 2013, evolving from dense compute to RA3 and serverless. As data volumes exploded, many organizations ran separate systems for structured warehouse tables and cost-effective data lakes for diverse datasets.

The rise of AI agents has supercharged query volumes. “Agent-based analytics can trigger thousands of queries per minute—something no human would ever do,” explained Sarah Chen, Research Director at Gartner. “This created a cost explosion that demanded a new architectural approach.” AWS responded by doubling down on Redshift’s core strengths.

In March 2026, Redshift already sped up new queries by up to 7 times for BI dashboards and ETL workloads. The Graviton-based RG instances now extend that acceleration to the entire analytics pipeline, including autonomous AI agents.

What This Means

For customers, this translates to substantially lower total analytics costs—especially for hybrid workloads spanning warehouse and data lake. Instead of maintaining separate engines, a single RG-based Redshift cluster queries both Redshift tables and S3 data lake files (Iceberg and Parquet).

“We expect customers running mixed workloads to see 30-50% savings in compute costs alone,” said Gupta. “Plus, the operational simplicity of one query engine for everything reduces management overhead.”

The integrated query engine is enabled by default on new clusters. Existing RA3 clusters can be migrated via the AWS Management Console, CLI, or API. AWS recommends using the AWS Pricing Calculator with specific workload patterns to estimate savings.

Getting Started

Launch new clusters or migrate existing ones through the AWS Management Console, AWS CLI, or AWS API. The integrated data lake query engine is enabled by default. No additional configuration is required—simply provision RG instances and start querying.

This is a developing story. Check back for updates on regional availability and pricing.

Tags: