AWS has released new Graviton-powered RG instances for its Amazon Redshift data warehouse service aimed at helping enterprises reduce both rising analytics costs and the operational complexity of modern lakehouse architectures.
At the core of the new instances is an integrated data lake query engine that AWS says can run SQL analytics across both Redshift warehouse data and Amazon S3 data lakes, delivering faster query performance and lowering analytics costs.
“Earlier, Amazon Redshift RA3 systems operated as two separate engines, with Redshift handling warehouse data and Spectrum handling S3 data lake queries. When a query required both, AWS had to coordinate between the two systems, which added complexity, slowed performance, and made Spectrum scan costs unpredictable,” said Pareekh Jain, principal analyst at Pareekh Consulting.
“The new RG instances combine those worlds into one integrated engine running directly inside Redshift itself. That means Iceberg, Parquet, and S3 lake data can now be queried natively alongside warehouse data with less movement, lower overhead, and better performance optimization while also eliminating separate Spectrum per-scan charges,” Jain added.
The separate Spectrum charges, the analyst further added, were increasingly becoming a pain point for enterprises as AI workloads drove higher query volumes, more machine-generated analytics, and greater data-processing demands, with many customers disliking Spectrum’s separate scan-based pricing because of the possibility of sudden bill spikes.
The new instances could be AWS’ response to growing enterprise demand for AI-scale analytics platforms that avoid added architectural complexity, as rivals including Databricks, Snowflake, Google Cloud with BigQuery, and Microsoft through Microsoft Fabric push unified lakehouse platforms to reduce operational sprawl, Jain said.
“RG instances do strengthen Amazon Redshift competitively, but mostly as a defensive move rather than a breakthrough disruption,” Jain said.
While Databricks leans on AI and data science capabilities, Snowflake on multi-cloud simplicity, Google Cloud on AI-native analytics through BigLake, and Microsoft on tight integration between Fabric, Power BI, and Microsoft Copilot, AWS is betting on the scale of Amazon S3 and tighter Redshift optimization to keep enterprise analytics workloads within its cloud stack, Jain added.
What enterprises should take note of
That differentiation in strategy, according to Greyhound Research Chief Analyst Sanchit Vir Gogia, is what CIOs and enterprise teams should be cognizant of when evaluating the new instances.
“The best fit is not every workload. The best fit is the painful overlap. That overlap is where Redshift, S3, open formats, BI, recurring analytics, cost pressure, and AI-assisted querying meet. That is where RG can materially reduce friction,” Gogia said.
“CIOs should inventory external schemas, identify high-scan and high-frequency lake queries, benchmark Iceberg and Parquet workloads under real concurrency, test month-end reporting pressure, model AI-agent query patterns, and measure whether savings are real after considering compute, S3, Glue, KMS, monitoring, and operational overhead,” Gogia added.
The biggest beneficiaries of the new RG instances will be enterprises already storing large amounts of data in S3 using formats like Iceberg and Parquet, especially in industries such as banking, telecom, retail, manufacturing, media, advertising, and IoT, Jain said. “These companies often deal with huge datasets, expensive data duplication between lakes and warehouses, unpredictable bills, and multiple systems that are hard to manage,” Jain said.
AWS, too, cautioned enterprises against assuming uniform savings across workloads, recommending that customers use the AWS Pricing Calculator with their own workload patterns to estimate potential cost reductions. The new RG instances, currently, have been made available across US East, US West, Canada, São Paulo, Frankfurt, Ireland, Milan, London, Paris, Spain, Stockholm, Mumbai, Hyderabad, Singapore, Sydney, Seoul, Tokyo, and Hong Kong regions.
Go to Source
Author:
