Snowflake vs Redshift Pricing and Cost With Which Warehouse to Choose
Per-second credits against RPU-hours and node-hours, priced from both vendors' own rate cards, with the number of hours a day at which each one becomes the cheaper choice. Try the console while you read: it is answering questions against a demo database.
PostgreSQL · MySQL · Snowflake · BigQuery · read-only
Ask the demo shop a question, or type your own:
Reading the schema, writing SQL… Writing SQL… Running (read‑only)… SQL AgentSQL wrote
▋
A real MySQL shop: customers, products, orders and signups over the last year. AgentSQL writes the SQL, runs it read-only, and answers.
Direct answer
Snowflake bills compute per second at $2.00 a credit on Standard and $3.00 on Enterprise, and an X-Small warehouse uses 1 credit an hour. Redshift Serverless bills $0.375 per RPU-hour from a 4 RPU minimum, $1.50 an hour, and a provisioned rg.large node costs $0.3801 an hour whether it is busy or not, $277.47 a month. So Snowflake is cheaper for light, bursty use, and an always-on Redshift node is cheaper once the warehouse is busy more than about 4.6 hours a day. Storage is nearly identical at $23 against about $24.58 per TB a month. Choose Snowflake for low operations and multi-cloud, Redshift for steady AWS workloads.
›_ list prices, US East, on demand
What Snowflake and Redshift actually charge, and what each unit covers.
Read from the vendors on 29 September 2026, before any negotiated or committed discount.
| Meter | List price | What it covers | Source |
|---|---|---|---|
| Snowflake credit | $2.00 Standard, $3.00 Enterprise, $4.00 Business Critical | Warehouse compute with the machines included. An X-Small warehouse burns 1 credit an hour, billed per second after a 60-second minimum, and suspends itself when idle. | Snowflake Credit Consumption Table, effective 25 Sep 2026, AWS US East |
| Snowflake storage | $23.00 per TB per month | Compressed data, averaged over the month, billed separately from compute. | Same PDF, Table 3(a) |
| Redshift Serverless | $0.375 per RPU-hour | All-in compute, including concurrency scaling and data lake queries. Base capacity starts at 4 RPUs (16 GB of memory each), so the smallest running workgroup costs $1.50 an hour. | AWS Price List API, offer file of 11 Sep 2026 |
| Redshift rg.large node | $0.3801 per node-hour | The newest provisioned generation, 2 vCPU and 16 GiB. Bills every hour it runs, busy or idle, unless you pause the cluster. | AWS Price List API, offer file of 11 Sep 2026 |
| Redshift ra3.large node | $0.543 per node-hour | The previous generation at the same 2 vCPU and 16 GiB. rg.large does the same job for exactly 70 percent of the price. | AWS Price List API, offer file of 11 Sep 2026 |
| Redshift managed storage | $0.024 per GB per month | About $24.58 per TB (1,024 GB), the same rate on Serverless, RG and RA3. | AWS Price List API, offer file of 11 Sep 2026 |
Both sets of numbers can be checked without talking to sales. AWS publishes every Redshift SKU in its Price List API, a machine-readable file that the pricing web page is built from; the one we read was published on 11 September 2026. Snowflake's pricing page draws its table in JavaScript, so the authoritative list is its Credit Consumption Table, a legal PDF on snowflake.com that carries the line "Effective: September 25, 2026". The full rate cards, including every node type and reservation term, are on our Redshift pricing and Snowflake pricing pages.
One caution before comparing rows. These are the smallest billable units, not equal amounts of horsepower. A 4 RPU Serverless workgroup has 64 GB of memory, four times an rg.large node, and Snowflake does not publish the hardware behind an X-Small. What the table does let you compare honestly is the shape of each bill: what you pay while nothing is running, and what you pay per hour while something is.
›_ worked out from the rate cards
Four numbers that decide Snowflake vs Redshift.
None of them appears in the comparison articles currently ranking for this search.
finding one
An always-on Redshift node wins after about 4.6 hours a day
One rg.large node running all month is $277.47. A Snowflake X-Small on Standard costs $2.00 an hour, so it passes that figure at 138.7 active hours, roughly 4.6 hours a day. On Enterprise at $3.00 the break-even arrives at 3.1 hours a day.
finding two
Redshift Serverless is the cheaper pay-per-use meter, not the cheaper warehouse
The smallest Serverless workgroup is $1.50 an hour against $2.00 for a Snowflake X-Small. But Serverless scales above its base when a query needs it, and bills the extra RPUs, so a heavy query can cost more than the base rate suggests.
finding three
Storage is a wash
Snowflake charges $23.00 per TB month and Redshift managed storage $0.024 per GB month, about $24.58 per TB. On 5 TB the gap is under $8 a month. Pick the warehouse on compute and on who runs it, not on storage.
finding four
Redshift gives 1,000 AI questions a month away
Amazon Q generative SQL in Query Editor v2 allows 1,000 prompts a month across the whole AWS account on the Free tier. Snowflake Cortex Analyst bills 67 credits per 1,000 messages, $134 on Standard. Past the free pool, Q Pro is $19 per user a month.
›_ monthly compute against busy hours
Where pay-per-second stops being the cheap option.
Snowflake and Redshift Serverless rise with every busy hour. A provisioned node is a flat line.
A Snowflake X-Small on Standard crosses one always-on rg.large node at 138.7 busy hours, about 4.6 hours a day over a 30-day month. Redshift Serverless at its 4 RPU floor crosses it at 185 hours, about 6.2 hours a day.
Commit to the node for a year with no upfront payment and the flat line drops to $194.23, which pulls Snowflake's break-even down to 97 hours, just over 3 hours a day. On Snowflake Enterprise at $3.00 a credit the break-even with the on-demand node is 92.5 hours.
The one thing the chart cannot show is idle time you forget about. A Snowflake warehouse with auto-suspend set to 60 seconds stops billing almost at once; a provisioned cluster nobody pauses bills all night.
›_ three workloads, five ways to pay
What a month of compute costs on each option.
List prices, US East, smallest unit of each, compute only. Add storage at roughly $23 to $25 per TB on either side.
| Workload | Snowflake X-Small, Standard | Snowflake X-Small, Enterprise | Redshift Serverless, 4 RPU | rg.large, on demand | rg.large, 1 yr reserved |
|---|---|---|---|---|---|
| Nightly load, 1 hour a day (30 hours) | $60 | $90 | $45 | $277.47 | $194.23 |
| Business-hours BI, 8 hours on 22 workdays (176 hours) | $352 | $528 | $264 | $277.47 | $194.23 |
| Always on, 730 hours | $1,460 | $2,190 | $1,095 | $277.47 | $194.23 |
Read the table row by row, because the winner changes. A nightly load that runs an hour is cheapest on Redshift Serverless at $45 and dearest on an idle node at $277.47, unless someone pauses the cluster between runs. A business-hours BI workload is close to a tie everywhere, which is exactly where operations, not price, should decide. Anything that runs around the clock, such as a customer-facing dashboard or a streaming ingest, is where provisioned Redshift is hard to beat: $194.23 reserved against $1,460 for a Snowflake X-Small left running.
Two footnotes keep this honest. Redshift Serverless grows above its base when a query needs more capacity and bills the extra RPUs, so a heavy query on a 4 RPU workgroup can cost more than $1.50 an hour; set MaxRPU if you need a ceiling. And a reserved node bills even while the cluster is paused, which AWS states on its pricing page, so reserve only what you will run. The same always-on against pay-per-use logic, with Google's per-query model added, runs through BigQuery vs Redshift.
›_ ten dimensions, both directions
Snowflake vs Redshift on what actually differs.
Including where each one is the weaker choice.
| Dimension | Snowflake | Redshift |
|---|---|---|
| Cloud | AWS, Azure or Google Cloud, with the same product on each. | AWS only, wired into S3, Glue, IAM and Lake Formation. |
| How compute is sold | Virtual warehouses in T-shirt sizes, billed per second in credits, auto-suspend when idle. | Two models. Serverless billed per RPU-second, or provisioned node clusters billed per node-hour. |
| Smallest running unit | X-Small warehouse, 1 credit an hour ($2.00 on Standard). | Serverless at 4 RPUs ($1.50 an hour), or one rg.large node ($0.3801 an hour, always on). |
| Idle cost | Zero once the warehouse suspends. You set the auto-suspend timer. | Zero on Serverless. A provisioned cluster bills until you pause it, and reserved nodes bill even when paused. |
| Concurrency | Multi-cluster warehouses add copies of the warehouse when users queue (Enterprise and above). | Concurrency scaling adds clusters; included on Serverless, one free hour a day per provisioned cluster. |
| Semi-structured data | VARIANT columns take JSON natively and query with dot notation. | SUPER type and PartiQL handle JSON; workable, less forgiving. |
| Operations | Almost nothing to tune. No vacuum, no distribution keys. | Serverless is low-touch. Provisioned rewards someone who manages sort keys, workload queues and node sizing. |
| Data sharing | Live sharing across accounts and clouds, plus a data marketplace. | Data sharing across Redshift warehouses and AWS accounts, inside AWS. |
| Plain-English questions | Cortex Analyst API at 67 credits per 1,000 messages, Snowflake Intelligence on AI Credits. | Amazon Q generative SQL in Query Editor v2, 1,000 free prompts a month per account, Pro $19 a user. |
| Price transparency | Full list prices in a public legal PDF. Contract discounts are negotiated. | Every SKU in AWS's public Price List API, including every reservation term. |
The row most buyers underweight is operations. A provisioned Redshift cluster is cheapest when someone owns it: picks the node type, sets sort and distribution keys, tunes workload queues and pauses it when nobody is working. Without that person, Redshift Serverless is the fair comparison with Snowflake, and at the smallest size it bills $1.50 an hour against Snowflake's $2.00. The row most buyers overweight is performance, where both vendors publish benchmarks showing themselves ahead and the honest answer is to test your own slowest queries on both.
If the choice is really about where your data engineers work rather than where analysts query, Snowflake vs Databricks covers the lakehouse side, and the Amazon Redshift alternatives guide prices the other warehouses AWS teams move to.
›_ the part of the bill nobody else prices
What it costs to let 10 business users ask 1,000 questions a month.
List prices, US dollars, per month, on top of warehouse compute.
Most warehouse decisions now come with a second question: once the data is in, can sales, finance and operations get answers without filing a ticket? AWS and Snowflake both sell a plain-English layer for it, priced in very different ways. Here is the same modest team, ten managers asking about a hundred questions each a month, on each option.
| Option | Per month | How that number is built |
|---|---|---|
| Amazon Q generative SQL, Free tier | $0 | 1,000 prompts a month for all users in the AWS account together, so 1,000 questions fits exactly and the 1,001st waits for next month. Redshift compute to run each generated query is billed as usual. |
| Amazon Q Developer Pro, 10 users | $190 | $19 per user a month, each user then gets 1,000 prompts a month. Needs IAM Identity Center set up for every user. Plus Redshift compute. |
| Snowflake Cortex Analyst (API) | $134 to $201 | 1,000 messages x 67 credits per 1,000 = 67 credits, at $2.00 Standard or $3.00 Enterprise. Plus warehouse credits to run each query, plus the app and the YAML semantic model you build around the API. |
| AgentSQL Team on Snowflake | $59 | Flat, billed yearly, up to 10 seats, no per-question charge. Queries run on your Snowflake warehouse, so its credits apply as they would for any BI tool. Does not connect to Redshift. |
Redshift wins this table on list price, and it is worth being clear about why. AWS's documentation for Query Editor v2 says that on the Amazon Q Developer Free tier "the total number of prompts of all users of an AWS account is limited to 1,000 in a month". That is a shared pool: one enthusiastic analyst can use it up for everyone by the middle of the month. The Pro tier lifts it to 1,000 prompts per user for $19 each. Cortex Analyst on Snowflake is the only option you can price exactly per question, $0.134 on Standard, but it is an API, so someone has to build the app and write the semantic model first.
If you are on Snowflake and want a finished tool rather than an API project, a read-only product that connects to the warehouse is the third route. How the Snowflake-native options compare in practice is on Snowflake natural language query.
›_ pick by how the warehouse is used
Three situations, three different right answers.
Choose Snowflake if nobody wants to run a warehouse
Your analysts write SQL, the load is spiky, and there is no platform engineer. Snowflake suspends itself when idle, adds warehouse clusters when users queue, and needs no vacuuming or key design. You pay a premium per busy hour for that, which is a good trade below about 4 or 5 busy hours a day and a poor one for a warehouse that never sleeps.
Choose Redshift if the warehouse runs all day on AWS
Your data already lives in S3, your team knows IAM and Glue, and the warehouse is busy most of the day. A reserved rg.large node at $194.23 a month is a fraction of what the same hours cost on any per-second meter. Budget for someone to own sort keys, queues and node sizing, or use Serverless and accept a higher hourly rate.
Re-price before you migrate
Many Redshift to Snowflake moves were justified against RA3 nodes and on-demand prices. The rg generation now does the same job for 70 percent of RA3, and Serverless removes idle cost. Before a migration project, price your real workload on both with the numbers above; our Redshift to Snowflake migration cost breakdown shows the one-off costs to add.
›_ for teams choosing snowflake
AgentSQL puts plain-English questions on Snowflake without a build project.
We build AgentSQL, so read this as interested rather than neutral. It connects to your Snowflake account with a read-only role, reads the live schema, turns a typed question into Snowflake SQL, runs it on the warehouse you choose and returns the number, a chart and the exact query so an analyst can check it. There is no semantic model to write before the first answer and no per-question charge: Team is $59 a month billed yearly for up to 10 seats, and Scale is $179 with role-based access and an audit log of every question. It never trains on your data, and it also connects to BigQuery, PostgreSQL and MySQL.
What it does not do, stated plainly: AgentSQL does not connect to Redshift, so if you stay on Redshift, Amazon Q generative SQL is the tool to evaluate first. It is not a warehouse and replaces neither. What it removes is the queue of one-off questions that lands on the data team every week. How the SQL generation works is on text to SQL, and Snowflake setup is on the Snowflake connection page.
›_ frequently asked
Snowflake vs Redshift questions, answered.
Is Snowflake better than Redshift?
Snowflake is better for teams that want a warehouse with almost no administration, spiky or unpredictable query load, or the option to run on Azure or Google Cloud. Redshift is better for AWS-committed teams with steady, all-day workloads, where a reserved node is far cheaper per hour than any pay-per-second meter. Neither wins everywhere.
Is Snowflake cheaper than Redshift?
For light, bursty use, usually yes against provisioned Redshift, because Snowflake bills nothing while suspended. For workloads that run most of the day, no. One rg.large node costs $277.47 a month always on, which a Snowflake X-Small on Standard passes after about 4.6 active hours a day. Redshift Serverless is cheaper per hour than Snowflake at the smallest size.
What is the difference between Snowflake and Redshift?
Snowflake is a multi-cloud warehouse that bills compute per second in credits and suspends when idle. Redshift is AWS's warehouse, sold either as Serverless, billed per RPU-second, or as node clusters billed per hour. Snowflake needs almost no tuning; provisioned Redshift rewards someone who manages it. Storage costs are nearly identical.
Is Redshift faster than Snowflake?
Neither is reliably faster. Both vendors publish benchmarks showing themselves ahead, and results turn on warehouse size, data layout and caching more than on the engine. For typical BI queries on a few terabytes, a sensibly sized warehouse on either returns in seconds. Test your own ten slowest queries on both before you believe any benchmark.
Can Snowflake replace Redshift?
Yes. Snowflake runs on AWS, reads from S3, and most Redshift SQL ports with modest changes, so companies move from one to the other regularly. The work is in rewriting Redshift-specific features such as distribution and sort keys, stored procedures and workload management rules, and in re-pointing every pipeline and dashboard. Plan it as a project, not a switch.
Why do companies move from Redshift to Snowflake?
The reasons given most often are concurrency, operations and idle cost. A provisioned Redshift cluster queues when many users hit it at once and needs ongoing tuning, while Snowflake adds warehouse clusters and suspends when idle. Redshift Serverless and the new RG nodes answer part of that, so re-price Redshift before assuming the move saves money.
Does Snowflake run on AWS?
Yes. Snowflake runs on AWS, Microsoft Azure and Google Cloud, and most US customers run it in an AWS region. That is why the choice is not really AWS against another cloud. It is AWS's own warehouse against a third-party warehouse hosted on AWS, with Snowflake billing you for the infrastructure inside its credit price.
Can business users query Snowflake or Redshift without SQL?
Yes. On Redshift, Amazon Q generative SQL in Query Editor v2 turns a typed question into SQL, with 1,000 free prompts a month per AWS account. On Snowflake there is the Cortex Analyst API or a separate tool that connects read-only. AgentSQL is one: it connects to Snowflake, writes and runs the SQL, and shows it. It does not connect to Redshift.
›_ pricing, read from the vendor
›_ the other warehouse side-by-sides
›_ connect your database
Ask Snowflake a question in plain English.
Read-only on Snowflake, BigQuery, Postgres or MySQL. Every answer comes back with the SQL beside it.