Estimate Amazon Redshift costs for Serverless RPU and provisioned nodes with Managed Storage, Spectrum, and Concurrency Scaling.
Last verified: May 2026
$0.375 per RPU-hour. Billed per second (min 60s per query).
Average hours per day that RPUs are actively processing queries.
$0.024 per GB-month for Redshift Managed Storage (RMS).
$5.00 per TB scanned from S3 via Redshift Spectrum.
Serverless automatically scales compute in RPU (Redshift Processing Units) and bills per second of active usage. Ideal for variable or unpredictable workloads. Provisioned clusters run 24/7 with fixed node capacity, offering better cost efficiency for steady-state workloads with Reserved pricing discounts up to 50%.
dc2 nodes use local SSD storage and are best for datasets under ~2 TB per node. ra3 nodes decouple compute from storage using Redshift Managed Storage (RMS) backed by S3, allowing you to scale storage independently. ra3 is recommended for most new clusters.
Redshift Spectrum lets you query data directly in S3 without loading it into Redshift. Charged at $5 per TB of data scanned. Use columnar formats (Parquet, ORC) and partitioning to minimize scan costs. Great for extending your warehouse to a data lake architecture.
Concurrency Scaling adds transient clusters to handle burst read queries. Each cluster gets 1 free hour per day of concurrency scaling credit. Beyond that, you pay the same per-second rate as your main cluster. Credits accumulate over 30 days, providing up to 30 free hours of scaling per month.
Output will appear here...The Redshift Cost Estimator projects your monthly Amazon Redshift bill for both Serverless and provisioned deployment models. For Serverless, configure your base RPU capacity and expected query hours to see RPU-hour charges alongside Managed Storage costs. For provisioned clusters, select node types (RA3, DC2, or DS2), node counts, and Reserved Instance terms. The estimator also covers Redshift Spectrum charges for querying external S3 data and Concurrency Scaling costs for burst capacity. Use it to model different configurations and find the price-performance sweet spot for your data warehouse workload.
Your data team is migrating from a 4-node DC2.8xlarge cluster ($21,000/month at On-Demand) to Redshift Serverless. The estimator with their workload (16 hour-per-day active query window, 32 RPU peak) shows Serverless cost at ~$10,500/month. They commit to 1-year RIs on the existing cluster instead at 40% discount = $12,600/month. Serverless wins by $2,100/month with no upfront commitment. They migrate, and after 6 months Serverless usage averages 24 RPU — bill drops further to $8,000/month as workload patterns settle.
Redshift Serverless's base RPU setting is the floor, not the average. You'll pay for the base capacity 24/7 even when idle. For workloads that have long idle periods (overnight, weekends), the minimum 8 RPU base ($2.88/hr × 730 hr = ~$2,100/month) often costs more than a small RA3 cluster running fewer hours.
Concurrency Scaling is a hidden cost driver: it adds clusters during peak load at On-Demand rates. The first hour per day is free per main cluster, but heavy use can add hundreds per day. Always monitor the SVL_QLOG and STL_USAGE_CONTROL views to see how often you're scaling.
Redshift Spectrum charges $5 per TB scanned in S3 — same as Athena. If you're already using Athena for the same data, layering Spectrum on top doesn't save query costs. Use Spectrum primarily to JOIN external S3 data with local Redshift tables, not as a replacement for Athena.
The estimator handles two billing models: Serverless (base RPU × hours × per-RPU rate, plus actual RPU-hour usage above base, plus Managed Storage GB × $0.024) and Provisioned (node count × per-node-hour rate, with RI discounts applied if 1-year or 3-year terms are selected). Spectrum and Concurrency Scaling are added as separate line items based on your inputs.
Was this tool helpful?
Disclaimer: This tool runs entirely in your browser. No data is sent to our servers. Always verify outputs before using them in production. AWS, Azure, and GCP are trademarks of their respective owners.