BigQuery vs Redshift Pricing with Amazon Redshift vs BigQuery Cost and Which Warehouse to Choose
Bytes scanned against RPU-hours and node-hours, priced from Google's and AWS's own rate cards, with the monthly scan volume at which each one becomes the cheaper warehouse. The console beside this is live against a demo database, so ask it something while you read.
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
BigQuery on demand charges $6.25 per TiB a query scans, with the first TiB each month free, and nothing between queries. Redshift Serverless charges $0.375 per RPU-hour from a 4 RPU base, $1.50 an hour while busy, and a provisioned rg.large node costs $277.47 a month whether it is busy or not. So BigQuery is cheaper for most teams scanning under about 45 TiB a month, and an always-on Redshift node wins above that, or above 32 TiB if you reserve it for a year. Storage list prices are close, $0.023 per GiB against $0.024 per GB, but BigQuery counts uncompressed bytes by default. Choose BigQuery for zero capacity planning, Redshift for steady AWS workloads.
What $100 of compute buys
Different units on purpose: BigQuery sells bytes, Redshift sells time. The Redshift bars show the share of a 730-hour month that $100 keeps running.
›_ list prices, US regions, pay as you go
What BigQuery and Redshift actually charge, and what each unit covers.
Read from Google on 2 October 2026 and from AWS's price list of 11 September 2026, before any negotiated or committed discount.
| Meter | List price | What it covers | Source |
|---|---|---|---|
| BigQuery on demand | $6.25 per TiB scanned | Every byte a query reads from the columns it references. The first 1 TiB a month is free per billing account. Nothing bills while no query runs. | cloud.google.com/bigquery/pricing |
| BigQuery Editions | $0.04, $0.06, $0.10 per slot-hour | Standard, Enterprise and Enterprise Plus capacity instead of bytes, per second with a 1-minute minimum. One and three year commitments cut the rate by 10 and 20 percent. | Same page, Iowa (us-central1) |
| BigQuery storage | $0.023 per GiB a month | Active logical (uncompressed) bytes, metered per GiB-hour at $0.000031507. Long-term logical $0.016. Physical (compressed) billing is $0.040 active and $0.020 long-term. First 10 GiB free. | Same page |
| Redshift Serverless | $0.375 per RPU-hour | Compute billed per second while queries run, 60-second minimum. Base capacity starts at 4 RPUs, so the smallest workgroup costs $1.50 an hour while busy. | AWS Price List API, offer file of 11 Sep 2026 |
| Redshift rg.large node | $0.3801 per node-hour | Provisioned compute, 2 vCPU and 16 GiB. Bills every hour the cluster runs, busy or idle, $277.47 for a 730-hour month. One year reserved, no upfront, is $194.23. | AWS Price List API, offer file of 11 Sep 2026 |
| Redshift managed storage | $0.024 per GB a month | The compressed columnar data the warehouse actually stores, about $24.58 per TB. Same rate on Serverless and on provisioned nodes. | AWS Price List API, offer file of 11 Sep 2026 |
A warning about the articles you will find on this comparison: several still say BigQuery costs $5 per TB. That was the on-demand price until Google raised it to $6.25 per TiB in July 2023, and a TiB is also about 10 percent bigger than the TB those pages imply. Google meters storage per GiB-hour rather than per month, which is why its page shows $0.000031507 and not a round number; multiplied by 730 hours it is $0.023 per GiB a month, not the $0.02 most guides repeat. The full BigQuery rate card, including the per-slot-hour break-even between on demand and Editions, is on our BigQuery pricing page, and every Redshift node type and reservation term is on Redshift pricing.
One caution before you compare 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 BigQuery on demand can borrow up to 2,000 slots for a single project. What the table does let you compare honestly is the shape of each bill: what you pay while nothing is running, and what drives the bill when something is.
›_ worked out from the rate cards
Four numbers that decide BigQuery vs Redshift.
None of them appears in the comparison articles currently ranking for this search.
finding one
BigQuery stays cheaper than a Redshift node until about 45 TiB scanned a month
One always-on rg.large node costs $277.47 a month. At $6.25 per TiB after the free first TiB, BigQuery on demand reaches that figure at 45.4 TiB scanned. Against a one-year reserved node at $194.23 the break-even is 32.1 TiB.
finding two
BigQuery bills bytes, Redshift bills time
A query that reads 50 GiB costs $0.31 on BigQuery however fast it runs. The same query on a 4 RPU Serverless workgroup that finishes in 30 seconds bills the 60-second minimum, about $0.025. Repeated dashboard queries over big tables favor Redshift; occasional questions favor BigQuery.
finding three
Equal storage prices hide a compression gap
BigQuery's default logical billing counts uncompressed bytes at $0.023 per GiB. Redshift bills the compressed columns it stores at $0.024 per GB. With 4 to 1 compression, 4 TiB of raw data is $94 a month on BigQuery logical and about $25 on Redshift. Switching BigQuery to physical billing closes most of that gap.
finding four
Plain-English questions are free on both, for now
Amazon Q generative SQL gives every AWS account 1,000 prompts a month. Google's Conversational Analytics agent for BigQuery is in a free trial through 31 December 2026, then bills $3 per million input and $20 per million output data tokens.
›_ monthly compute against TiB scanned
Where paying per byte stops being the cheap option.
BigQuery on demand rises with every TiB your queries read. A provisioned Redshift node is a flat line.
BigQuery on demand crosses one always-on rg.large node at 45.4 TiB scanned a month: $277.47 buys 44.4 TiB at $6.25, plus the free first TiB. Against the same node reserved for a year with no upfront payment, the crossover drops to 32.1 TiB.
The number that matters is your own monthly scan, and BigQuery shows it to you. In the console, the job history lists bytes billed for every query, and the INFORMATION_SCHEMA.JOBS view sums a month of it in one query. Most small and mid-sized teams find they scan far less than they assumed.
What the chart cannot show is whether one 2 vCPU node can carry your workload. Past roughly 30 to 45 TiB a month you are probably also past what a single small node handles comfortably, so price two or four nodes before declaring Redshift the winner.
›_ three workloads, four ways to pay
What a month of compute costs on each option.
List prices, US regions, compute only. Add storage at roughly $23 to $25 per TB on either side.
| Workload | BigQuery on demand | Redshift Serverless, 4 RPU | rg.large, on demand | rg.large, 1 yr reserved |
|---|---|---|---|---|
| Startup dashboards, 3 TiB scanned, busy 60 hours | $12.50 | $90 | $277.47 | $194.23 |
| Business-hours BI, 20 TiB scanned, busy 176 hours | $118.75 | $264 | $277.47 | $194.23 |
| Always-on analytics, 80 TiB scanned, busy 730 hours | $493.75 | $1,095 | $277.47 | $194.23 |
The first two rows are where most companies asking this question actually sit, and BigQuery wins both by a wide margin: a startup scanning 3 TiB a month pays $12.50 against $90 for the smallest Serverless workgroup used 60 hours. Business-hours BI on 20 TiB is $118.75 against $194.23 for the cheapest Redshift option. Only the always-on row flips, and it flips hard: $493.75 on BigQuery against $194.23 for a reserved node, as long as that node keeps up.
Two footnotes keep the comparison fair. BigQuery's cost depends on how much each query reads, so an unpartitioned table and a habit of SELECT * can multiply the first column several times; partitioning by date and clustering on common filters usually cuts it more than any pricing plan change. And Redshift Serverless grows above its 4 RPU base when a query needs it and bills the extra, so set a maximum RPU if you need a ceiling. Spend controls for the BigQuery side are covered in BigQuery cost control for ad hoc queries.
›_ ten dimensions, both directions
BigQuery vs Redshift on what actually differs.
Including where each one is the weaker choice.
| Dimension | BigQuery | Redshift |
|---|---|---|
| Cloud | Google Cloud. BigQuery Omni can query data sitting in AWS S3 or Azure, billed separately. | AWS only, wired into S3, Glue, IAM and Lake Formation. |
| How compute is sold | On demand per TiB scanned, or Editions capacity per slot-hour. | Serverless per RPU-second, or provisioned node clusters per node-hour. |
| What a single query costs | The bytes it reads, regardless of how long it runs. Cached repeats of the same query within 24 hours are not billed. | The capacity time it uses. On Serverless a short query bills the 60-second minimum. |
| Idle cost | Zero on demand. Editions autoscaling bills only while slots are in use. | Zero on Serverless. A provisioned cluster bills until paused, and reserved nodes bill even when paused. |
| Cost guardrails | Maximum bytes billed per query, custom daily quotas per project or per user, dry runs that show bytes before you run. | Base and maximum RPU settings and usage limits on Serverless; fixed node count on provisioned. |
| Tuning | No indexes or keys. Partitioning and clustering cut bytes scanned, which cuts the bill directly. | Automatic on Serverless. Provisioned rewards someone who manages sort keys, distribution and workload queues. |
| Semi-structured data | Native JSON type plus nested and repeated STRUCT and ARRAY columns. | SUPER type queried with PartiQL; workable, less forgiving. |
| Plain-English questions | Conversational Analytics agent and Gemini in BigQuery, free trial through 31 Dec 2026, then per data token. | Amazon Q generative SQL in Query Editor v2, 1,000 free prompts a month per account, Pro $19 a user. |
| Migration tooling | BigQuery Data Transfer Service has a Redshift connector, and the SQL translation service converts Redshift SQL. | AWS Schema Conversion Tool and Database Migration Service for the reverse direction. |
| Price transparency | Full public pricing page, storage metered per GiB-hour. | Every SKU in AWS's public Price List API, including every reservation term. |
The row buyers underweight is guardrails. BigQuery's per-byte model means one careless query against a 50 TiB table can cost $312 on its own, so set a maximum bytes billed on every account that writes ad hoc SQL and a daily per-user quota on the project. Redshift's failure mode is the opposite: a provisioned cluster nobody pauses bills all night, every night. Both are fixable in an afternoon, and both are the most common reason a team's first month costs more than the estimate.
If the shortlist also includes Snowflake, the same time-against-bytes logic runs through BigQuery vs Snowflake and Snowflake vs Redshift, and the Amazon Redshift alternatives guide prices every warehouse AWS teams usually 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 loaded, can sales, finance and operations get answers without filing a ticket? Google and AWS both sell a plain-English layer for that, and both are free at the moment for different reasons. 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 shared by every user in the AWS account, so 1,000 questions fits exactly and the 1,001st waits for next month. Redshift compute to run each 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. Plus Redshift compute. |
| BigQuery Conversational Analytics agent | $0 until 31 Dec 2026 | Free trial of all Data Cloud agents through 31 December 2026. After that $3 per million input and $20 per million output data tokens. Google publishes no per-question figure, so forecast it from a pilot. Plus BigQuery bytes scanned. |
| AgentSQL Team on BigQuery | $59 | Flat, billed yearly, up to 10 seats and 1,500 questions a month, no per-token charge. Queries run on your BigQuery project, so normal bytes-scanned billing applies. Does not connect to Redshift. |
The dates matter more than the prices here. Google priced its Data Cloud agents, which include the Conversational Analytics agent for BigQuery, at $3 per million input and $20 per million output data tokens, and originally ended the free trial on 30 September 2026. Its pricing page now runs the trial through 31 December 2026, so a pilot started this quarter is free and the first bill lands in January. Output tokens cost 6.7 times input, which means long answers drive the bill more than the number of questions, and Google publishes no per-question estimate. AWS's limit is a shared pool instead: Amazon Q's free tier allows 1,000 prompts a month across the entire account, so one enthusiastic analyst can use it up for everyone.
How BigQuery's own natural-language options compare with a separate read-only tool is covered in querying BigQuery without writing SQL, and the per-vendor meters for every platform are side by side on conversational analytics pricing.
›_ pick by how the warehouse is used
Three situations, three different right answers.
Choose BigQuery if nobody wants to size a warehouse
Your query load is uneven, the team is small, and nobody owns capacity. BigQuery bills nothing between queries, the first TiB is free, and partitioning a few large tables is the only tuning that moves the bill. Below a few tens of TiB scanned a month, it is very hard to beat on price or on effort.
Choose Redshift if the warehouse runs all day on AWS
Your data already lives in S3, your team knows IAM and Glue, and dashboards hit the same big tables all day. A reserved rg.large node at $194.23 a month turns heavy repeated scanning into a flat cost. Budget for someone to own sort keys and queues, or use Serverless and accept a higher hourly rate.
Measure your scan before you migrate
A move from Redshift to BigQuery only saves money if your monthly scan volume is below the break-even. Redshift's system tables show how much data your queries read; convert that to TiB, multiply by $6.25 and compare it with your current node bill before starting a migration project. If you are weighing Snowflake as well, our Redshift to Snowflake migration cost breakdown lists the one-off costs to add.
›_ for teams choosing bigquery
AgentSQL puts plain-English questions on BigQuery without a build project.
We build AgentSQL, so read this as interested rather than neutral. It connects to your BigQuery project with a read-only service account, reads the live schema, turns a typed question into GoogleSQL, runs it and returns the number, a chart and the exact query so an analyst can check it. There is no per-question or per-token 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. Because every answer shows its SQL, you can see which tables it reads and keep bytes scanned in check. It never trains on your data, and it also connects to Snowflake, PostgreSQL and MySQL.
What it does not do, stated plainly: AgentSQL does not connect to Redshift, so if you stay on AWS, Amazon Q generative SQL is the tool to evaluate first. It is not a warehouse and replaces neither. What it removes is the weekly queue of one-off questions that lands on the data team. BigQuery setup takes a few minutes and is described on the BigQuery connection page, and how the SQL generation works is on text to SQL.
›_ frequently asked
BigQuery vs Redshift questions, answered.
Is BigQuery better than Redshift?
BigQuery is better for teams that want no capacity to manage and whose query volume is uneven, because it bills per byte scanned and costs nothing between queries. Redshift is better for AWS teams with steady, all-day workloads, where a reserved node is cheaper than any per-query meter. Neither wins everywhere, so price your own monthly scan volume first.
Is BigQuery cheaper than Redshift?
Usually, below about 45 TiB scanned a month. BigQuery on demand costs $6.25 per TiB with the first TiB free, while one always-on Redshift rg.large node costs $277.47 a month. Past roughly 45 TiB, or 32 TiB against a reserved node, the flat Redshift node is cheaper, provided one node can carry the workload.
What is the difference between BigQuery and Redshift?
BigQuery is Google Cloud's serverless warehouse that bills by the bytes each query reads, or by slot-hours on Editions. Redshift is AWS's warehouse, sold as Serverless billed per RPU-second or as node clusters billed per hour. BigQuery needs no capacity planning; provisioned Redshift rewards someone who tunes it. Storage list prices are close.
Is Redshift faster than BigQuery?
Neither is reliably faster. Speed depends on how the data is partitioned or sorted, the capacity you give each one and caching more than on the engine. BigQuery is quick on large ad hoc scans; a well-tuned Redshift cluster is quick on repeated dashboard queries. Run your ten slowest real queries on both before believing any benchmark.
What is the AWS equivalent of BigQuery?
Amazon Redshift is the AWS data warehouse that competes with BigQuery, and Redshift Serverless is the closest match in how it is run. If you want BigQuery's pay-per-byte billing specifically, Amazon Athena is the nearer equivalent: it charges $5.00 per TB scanned against files in S3, with no warehouse to load.
Can BigQuery query data in AWS?
Yes. BigQuery Omni runs BigQuery queries against data stored in AWS S3 and Azure Blob Storage without copying it to Google Cloud, billed under its own on-demand and capacity prices. For moving data the other way, BigQuery Data Transfer Service includes an Amazon Redshift connector used for migrations.
How hard is it to migrate from Redshift to BigQuery?
It is a project of weeks to months, not a switch. Google's Data Transfer Service copies Redshift tables and its SQL translation service converts most Redshift SQL, but sort and distribution keys, stored procedures, workload queues and every pipeline and dashboard connection still need rework. Price the target bill on your real scan volume before you start.
Can business users query BigQuery or Redshift without SQL?
Yes. On Redshift, Amazon Q generative SQL turns typed questions into SQL with 1,000 free prompts a month per account. On BigQuery, Google's Conversational Analytics agent is free through 31 December 2026. AgentSQL is a third route on BigQuery: it connects read-only, writes and runs the SQL and shows it for $59 a month. It does not connect to Redshift.
Pricing, read from the vendor
›_ the other warehouse side-by-sides
›_ connect your database
Ask BigQuery a question in plain English.
Read-only on BigQuery, Snowflake, Postgres or MySQL. Every answer comes back with the SQL beside it.