Amazon Redshift Review
AWS's fully managed, petabyte-scale cloud data warehouse
Is this your product? Claim this page · Request a change
Looking for a Amazon Redshift alternative? See our ranked comparison.→What is Amazon Redshift?
Amazon Redshift is AWS's managed cloud data warehouse. Teams use it to run SQL analytics and BI/ETL workloads over large volumes of structured and semi-structured data. It launched in 2013 as one of the first cloud-native columnar data warehouses and is still one of AWS's most used data services.
You can run Redshift as provisioned clusters, where you choose the node type and count and pay per node-hour, or as Redshift Serverless, where you pay per RPU-hour and AWS handles scaling and cluster management. Storage for RA3/RG node types is billed separately as managed storage, so compute and storage costs are split, similar to Snowflake and BigQuery.
Through 2025 and 2026, AWS has added more automation and tighter integration with the rest of its data stack: AI-driven serverless scaling by default, zero-ETL integrations from operational databases and S3, multi-warehouse data sharing, and query performance work that cuts cold-start query execution time by up to 7x.
Who it's for
- ✓ Teams already on AWS who want a data warehouse tightly integrated with S3, Glue, QuickSight, and other AWS services
- ✓ Workloads with spiky or unpredictable usage, since Redshift Serverless bills per RPU-hour and drops to no charge when idle
- ✓ Teams that need to combine data warehouse queries with data lake queries over S3, using Redshift Spectrum, without moving the data
Who should look elsewhere
- ✗ Teams outside the AWS ecosystem who don't need deep AWS integration and want simpler, more transparent per-second or per-credit pricing, like Snowflake or BigQuery
- ✗ Small teams or startups that want a real free tier to experiment before committing spend. Redshift's only no-cost option is a time-limited trial credit
- ✗ Teams that need true multi-cloud portability, since Redshift only runs on AWS
Pros
- + Deep native integration with the AWS ecosystem (S3, Glue, Lambda, QuickSight, SageMaker, IAM) means less glue code for teams already on AWS
- + Redshift Serverless removes cluster sizing and management. You pay per RPU-hour and pay nothing while idle
- + Redshift Spectrum lets you query data sitting in S3 directly, without loading it into the warehouse first, at $5/TB scanned
- + Reserved Instance pricing (1- or 3-year, with upfront payment options) gives real discounts for steady, predictable workloads
- + Recent query-execution changes, moving compilation off the critical path, cut cold-start query latency for BI dashboards and ETL by up to 7x as of 2026
Cons
- – No standing free tier. The only no-cost options are a 90-day/$300 Serverless credit or a two-month provisioned trial. After that, standard on-demand billing kicks in automatically
- – Pricing is split across compute (provisioned node-hour or serverless RPU-hour), managed storage (per GB-month), Spectrum scans (per TB), and Concurrency Scaling, so total cost is harder to estimate up front than with single-metric competitors
- – AWS-only. It doesn't run on GCP or Azure, unlike Snowflake, Databricks, or Starburst, which limits it for multi-cloud strategies
- – Serverless has a 4-RPU minimum, so even light, intermittent workloads carry an effective floor of roughly $1.50/hour of active compute
- – DS2 node types are already fully retired and can no longer be created. AWS is also pushing remaining DC2 users toward RA3/RG, which bills managed storage separately from compute instead of bundling local SSD storage into the node price
Amazon Redshift pricing
What you pay for
Redshift bills by usage, not by seat. You pay for compute (per RPU-hour on Serverless or per node-hour on provisioned clusters) plus managed storage per GB-month, and Spectrum and Concurrency Scaling are billed separately on top. The cheapest way in is Redshift Serverless at $0.375 per RPU-hour, though the 4-RPU minimum means an active workgroup costs about $1.50 an hour while running, with no charge when idle. There's no standing free tier. New users get a $300 credit that expires after 90 days, or a two-month trial on provisioned clusters where Serverless isn't available. Pricing is publicly disclosed on AWS's pricing page, though Reserved Instance rates need a custom quote.
You pay for what you consume rather than a per-seat fee, so cost scales with usage.
| Plan | Price | Highlights |
|---|---|---|
| Serverless | $0.375/mo | Billed per RPU-hour (US East N. Virginia rate) · Minimum 4 RPUs, so effective floor is about $1.50/hour while active · Scales up to 1,024 RPUs · No charge during idle periods · AI-driven scaling is now the default for new workgroups |
| Provisioned (RA3, on-demand) | $3.26/mo | ra3.4xlarge example rate, US East N. Virginia · Compute and managed storage billed separately · Managed storage (RMS) billed at roughly $0.024/GB-month |
| Provisioned (RG, on-demand, newest generation) | $3.04267/mo | rg.4xlarge example rate, US East N. Virginia · AWS-recommended node family for new provisioned clusters |
| Reserved Instances | Custom | 1-year or 3-year commitments · No Upfront, Partial Upfront, or All Upfront payment options · Significant discount vs. on-demand rates |
Redshift has no flat per-seat price. You pay for compute-hours and storage. Provisioned on-demand starts at $0.543/hour for the smallest node (AWS doesn't itemize the exact node type beyond its examples). The RA3 and RG rates above are for the 4xlarge size in US East (N. Virginia) and vary by region. Redshift Serverless bills per RPU-hour with a 4-RPU minimum and 60-second billing granularity, so an active workgroup costs at least about $1.50/hour before any usage. Redshift Spectrum, for querying data directly in S3, is billed separately at $5.00 per TB scanned. All figures are USD, US East (N. Virginia), from the live AWS pricing page.
Pricing verified July 6, 2026 · source
How Amazon Redshift's pricing compares
Amazon Redshift next to its closest alternatives on entry price, billing, and whether pricing is public.
| Tool | Starting price | Billing | Free option | Pricing disclosed |
|---|---|---|---|---|
| Amazon Redshift | $0.375/RPU-hour | usage-based | Trial ($300 credit (90-day expiration) for first-time Redshift Serverless users; in regions without Serverless, a two-month free trial for provisioned clusters (up to 750 hours/month)) | Partly public |
| Databricks | Free tier + custom | usage-based | Yes | Not disclosed |
| Snowflake | Custom / quote | usage-based | Trial (30 days with $400 in free credits) | Not disclosed |
| Google BigQuery | $6.25/TiB scanned | usage-based | Yes | Public |
| Dremio | $0.2/DCU | usage-based | Yes | Partly public |
| Starburst | $0.5/credit | usage-based | Yes | Public |
Is Amazon Redshift still actively developed?
Last significant update: April 2026. AI-driven scaling and optimization became the default for all new Redshift Serverless workgroups. It uses machine learning to predict compute needs and adjust resources before queries queue up. The supported Base RPU range also expanded to 8-512 RPU, up from the prior 32-512 RPU floor.
Top Amazon Redshift alternatives
Amazon Redshift FAQ
Does Amazon Redshift have a free tier?+
No standing free tier. First-time Redshift Serverless users get a $300 credit that expires after 90 days. In regions without Serverless, provisioned clusters get a two-month trial capped at 750 hours a month. After either expires, standard on-demand billing applies automatically.
How is Amazon Redshift priced?+
Provisioned clusters are billed per node-hour by node type, plus separate managed storage charges. On-demand provisioned pricing starts at $0.543/hour, with example rates of $3.26/hour for ra3.4xlarge and $3.04267/hour for rg.4xlarge. Redshift Serverless is billed per RPU-hour ($0.375/RPU-hour in US East N. Virginia) with a 4-RPU minimum, so an active workgroup costs at least about $1.50/hour.
What is the difference between Redshift Serverless and provisioned Redshift?+
Provisioned Redshift requires you to choose and manage node types and cluster size, billed per node-hour with optional Reserved Instance discounts. Redshift Serverless removes cluster management entirely. It auto-scales compute, measured in RPUs, up and down based on load and bills only for what you use, with no charge while idle.
Can Amazon Redshift query data stored in S3 without loading it?+
Yes, through Redshift Spectrum. It runs SQL queries directly against data in S3 and is billed separately at $5.00 per terabyte scanned, on top of regular Redshift compute costs.