Data warehouse cost compared: Redshift vs BigQuery vs Synapse

Quick answer

The three warehouses price compute differently: Redshift offers provisioned nodes or serverless RPUs, BigQuery offers on-demand per-TB-scanned or capacity slots, and Synapse offers dedicated SQL pools (per DWU) or serverless (per TB scanned). Storage is cheap and similar on all three; compute dominates. The cheapest depends on your query pattern, steady heavy load favors provisioned, spiky or ad-hoc favors serverless/per-scan, not the platform itself.

Redshift, BigQuery, and Synapse all separate cheap storage from expensive compute, but they meter compute on different axes. That makes a headline comparison misleading; the cheapest for you depends on how you query, not which cloud's warehouse you pick. The pattern of your workload maps to a pricing mode on each.

Three compute models

Warehouse	Compute options
Redshift	Provisioned nodes (always on) or Serverless (per RPU-hour)
BigQuery	On-demand (per TB scanned) or capacity slots (reserved/autoscale)
Synapse	Dedicated SQL pool (per DWU, pausable) or serverless (per TB scanned)

Each offers a provisioned/reserved mode (cheaper for steady, heavy query load) and a consumption mode (per-scan or per-RPU, cheaper for spiky or ad-hoc querying). BigQuery's on-demand per-TB-scanned model and Synapse's serverless per-TB model reward scanning less; Redshift and Synapse dedicated reward keeping a right-sized cluster busy. The details are in Redshift versus Athena, BigQuery on-demand versus flat-rate, and Synapse cost.

Storage is cheap and similar

On all three, storage is a minor, similar cost (a few cents per GB-month, with long-term or compressed discounts), so storage rarely decides the platform. Compute is the bill, which is why the query pattern, not the storage rate, drives the comparison.

Match the mode to the query pattern

Steady, heavy, predictable query load favors a provisioned or reserved mode on any of the three, keeping a right-sized cluster or slot pool busy. Spiky, ad-hoc, or intermittent querying favors a consumption mode, per-TB-scanned or serverless, that costs nothing when idle. The same workload lands on the same kind of mode across all three, so the cheapest is usually whichever cloud you already use, run in the right mode.

The shared optimization: scan less

Whatever the platform, scanning less data is the universal lever, through partitioning, columnar storage, compression, and selective queries. On per-TB-scanned modes it cuts cost directly; on provisioned modes it lets a smaller cluster serve the load. Choose the warehouse on your cloud and team, pick the mode that matches your query pattern, and optimize to scan less, which pays on all three.

FAQ

Is Redshift, BigQuery, or Synapse cheapest?

It depends on your query pattern, not the platform. All three separate cheap, similar storage from expensive compute, and each offers a provisioned mode (cheaper for steady heavy load) and a consumption mode (cheaper for spiky or ad-hoc querying). The cheapest is usually whichever cloud you already use, run in the mode that matches your workload.

How do the data warehouse pricing models differ?

Redshift offers provisioned nodes or serverless RPUs, BigQuery offers on-demand per-TB-scanned or capacity slots, and Synapse offers dedicated SQL pools (per DWU, pausable) or serverless per-TB. Each has a reserved mode for steady load and a consumption mode for spiky querying; the axes differ but the pattern maps across all three.

Does storage decide the data warehouse choice?

Rarely. Storage is a minor, similar cost on all three (a few cents per GB-month with long-term or compressed discounts), so it seldom decides the platform. Compute dominates the bill, which is why the query pattern, not the storage rate, drives the comparison.

When should I use provisioned versus consumption warehouse pricing?

Steady, heavy, predictable query load favors a provisioned or reserved mode (Redshift nodes, BigQuery slots, Synapse dedicated pool), keeping a right-sized cluster busy. Spiky, ad-hoc, or intermittent querying favors a consumption mode (per-TB-scanned or serverless) that costs nothing when idle.

What is the universal data warehouse cost lever?

Scanning less data, through partitioning, columnar storage, compression, and selective queries. On per-TB-scanned modes it cuts cost directly; on provisioned modes it lets a smaller cluster serve the load. This lever pays on Redshift, BigQuery, and Synapse alike, whatever the pricing mode.

Does C3X estimate data warehouse cost?

C3X prices provisioned warehouse resources (Redshift clusters, Synapse dedicated pools) from your Terraform. Consumption-mode cost (per-TB-scanned, serverless, slots) depends on query volume and bytes scanned, which you model as usage assumptions to compare across platforms.

What to do next

Compare warehouse options for your workload. C3X reads your Terraform and prices your resources against a live catalog. Start with the quickstart.