At a glance
| SQL support | PostgreSQL-derived SQL with QUALIFY, PartiQL for SUPER, stored procedures; Lambda UDFs |
| Storage and compute architecture | Leader and compute nodes with managed storage offloaded to S3, or Serverless workgroups in RPUs |
| Workload types | SQL analytics and BI, ELT, lake queries, data sharing, ML through SageMaker AI |
| Integrations | Amazon S3, Glue Data Catalog, IAM, Lambda, SageMaker AI, AWS Data Exchange |
| Deployment | AWS only: provisioned clusters or Serverless workgroups in AWS Regions |
| Pricing model | Node-hours (reserved available) or RPU-hours; managed storage per GB-month; Spectrum per TB scanned |
Redshift suits teams committed to AWS who want the warehouse inside their AWS account model, IAM and billing, with direct queries over Iceberg tables in the Glue Data Catalog. Serverless removes cluster sizing for variable workloads, while reserved provisioned clusters suit steady ones; either way there are more capacity and table design decisions than in a fully serverless warehouse.
Overview
Amazon Redshift is a massively parallel data warehouse with SQL descended from PostgreSQL. A provisioned cluster has a leader node that plans queries and compute nodes that run them. AWS recommends RG (Graviton based) or RA3 nodes, both using Redshift managed storage so that compute and storage are paid for separately; DC2 nodes are listed as previous generation.
Redshift Serverless removes clusters: capacity is measured in Redshift Processing Units (RPUs, 16 GB of memory each), with a default base of 128 RPUs that can be set from 4 to 512, or up to 1,024 in five large Regions. AI-driven scaling adjusts capacity towards a price-performance target, and maximum capacity and maximum RPU-hours limits cap spend.
Key features
- Provisioned RG and RA3 clusters with managed storage, pause and resume, and reserved nodes
- Redshift Serverless with base and maximum RPUs, AI-driven scaling and RPU-hour limits
- Concurrency Scaling clusters for WLM queues, with one free hour of credit per 24 hours per cluster
- SUPER type for semi-structured data, queried with PartiQL
- Data sharing across clusters, workgroups, AWS accounts and Regions, including writes
- Queries over Iceberg and other lake tables in the AWS Glue Data Catalog
- Redshift ML, which trains models through Amazon SageMaker AI
Pros and cons
Pros
- Choice of Serverless or provisioned clusters with reserved pricing
- Native IAM, VPC, S3 and Glue Data Catalog integration on AWS
- Serverless spend limits through maximum capacity and maximum RPU-hours
- Data sharing with write access across accounts and Regions
Cons
- AWS only, with no option to run the warehouse on other clouds
- More capacity and design decisions: provisioned or Serverless, node type, base RPUs, distribution and sort keys
- AWS has ended support for Python UDFs after 30 June 2026; existing ones must move to Lambda UDFs
- SUPER values are limited to 16 MB, and Concurrency Scaling has documented exclusions
Pricing
Redshift has two compute models. Provisioned clusters are billed per node-hour by node type (RG and RA3, with DC2 as previous generation), on demand or with one- and three-year reserved nodes; the AWS pricing page lists, for example, rg.4xlarge at USD 3.04267 per hour in US East (N. Virginia). Redshift Serverless is billed in RPU-hours per second with a 60-second minimum and no charge for start-up time, at USD 0.375 per RPU-hour in US East (N. Virginia), with base capacity adjustable from 4 RPUs. Redshift managed storage costs USD 0.024 per GB-month in US East (N. Virginia), and Redshift Spectrum is USD 5.00 per TB scanned (10 MB minimum per query). Prices vary by Region. First-time Serverless users get a USD 300 credit that expires after 90 days.
Pricing checked on the vendor's official page on 7 November 2026. Prices change; confirm before buying.
Who it is best for
AWS-based teams that want a warehouse integrated with S3, Glue and IAM, with a choice of serverless or reserved provisioned capacity.
Who should look elsewhere
- Organisations that need one warehouse across several clouds: see Snowflake vs Redshift.
- Small operational datasets that fit in PostgreSQL: see Redshift vs PostgreSQL.
Verdict
Redshift is the natural warehouse for teams whose data and applications are on AWS, and Serverless has made it far easier to start. Expect to make more capacity and table design choices than with Snowflake or BigQuery, and check any Python UDFs, which AWS no longer supports.
Frequently asked questions
How much does Redshift Serverless cost?
AWS lists USD 0.375 per RPU-hour in US East (N. Virginia), billed per second with a 60-second minimum while the warehouse is active. Managed storage is billed separately at USD 0.024 per GB-month there. New Serverless users get a USD 300 credit valid for 90 days.
Should I choose Redshift Serverless or a provisioned cluster?
Serverless suits variable or unpredictable workloads because you pay for RPU-hours while queries run and can cap spend. Provisioned RG or RA3 clusters suit steady workloads, where reserved nodes lower the hourly rate.
Does Redshift still support Python UDFs?
No. AWS states that Redshift no longer supports Python UDFs after 30 June 2026, enforced in phases, and recommends migrating them to Lambda UDFs.
Sources
- Amazon Redshift pricing
- Amazon Redshift Management Guide: Provisioned clusters and node types
- Amazon Redshift Management Guide: Serverless compute capacity
- Amazon Redshift Developer Guide: Concurrency scaling
- Amazon Redshift Developer Guide: Python UDF language support
Checked November 2026.
How we research reviews and what star ratings mean: our editorial method.