What Does a Data Warehouse Cost? Understanding Cloud Analytics Pricing

Data warehouse costs in cloud environments come from three categories: storage, compute, and data transfer. Understanding the pricing model of the platform you use — and the behaviors that drive costs in each category — is the foundation of cost governance. Organizations that do not actively manage their cloud data warehouse costs routinely discover invoices 2-5x their estimates when usage patterns differ from assumptions.

Storage Costs

Cloud data warehouse storage costs are relatively low compared to compute and are the easiest cost category to forecast.

**Snowflake storage** is priced per terabyte per month (typically $23-40/TB/month depending on region and contract). Snowflake's columnar compression reduces physical storage significantly — a table that would be 10TB uncompressed might be 2TB compressed, reducing actual charges. Time Travel (retaining historical versions of data) and Fail Safe (disaster recovery data copies) add to storage costs. Organizations with heavy Time Travel usage (the default 90-day Fail Safe period) may find storage growing faster than raw table size growth.

**BigQuery storage** separates active storage (recently modified data) and long-term storage (data not modified for 90+ days, priced at roughly half the active rate). For large historical datasets that are rarely modified, BigQuery's long-term storage tier provides automatic cost reduction.

**Redshift storage** varies by instance type and managed storage (RA3 instances with separate managed storage) versus dense storage (DS2 instances). RA3's managed storage scales independently of compute; DS2's compute-to-storage ratio is fixed.

**Storage cost management:** Identify and purge unused tables (information_schema last_modified dates), configure Snowflake Time Travel periods appropriate to recovery requirements (30 days rather than 90 for non-critical tables), and compress data appropriately before loading.

Compute Costs

Compute is typically the dominant data warehouse cost category, and the most variable.

**Snowflake compute** is charged in credits per second while virtual warehouses are running. The credit rate varies by warehouse size (X-Small: 1 credit/hour, Small: 2 credits/hour, and doubling upward). The primary levers for compute cost control:

- Auto-suspend aggressively (2-5 minutes of inactivity rather than the default 10+ minutes)

- Right-size warehouses for the workload — do not run X-Large for queries a Medium would handle

- Separate compute by workload type so development activity does not inflate production warehouse costs

- Use Snowflake result cache — identical queries return cached results without consuming compute

- Resource monitors to prevent runaway jobs from consuming unbounded credits

**BigQuery compute** on on-demand pricing charges per byte scanned by each query ($5-6 per TB scanned depending on region). The primary levers:

- Partition and cluster tables to reduce bytes scanned per query

- Avoid SELECT star queries — select only the columns needed

- Materialize frequently queried aggregations as scheduled query outputs or materialized views

- Use flat-rate slot reservations for predictable high-volume workloads (more economical than on-demand above a threshold)

- Set maximum bytes billed per query to prevent accidental full-table scans

**Redshift compute** is charged by provisioned instance-hours for traditional provisioned clusters, or query-based with Redshift Serverless. Provisioned cluster costs are fixed regardless of usage — the lever is right-sizing the cluster and using workload management (WLM) to prioritize queries appropriately.

Data Transfer Costs

Data transfer costs are frequently overlooked in cloud data warehouse budgets. Cloud providers charge for data transferred out of their network (egress) and sometimes for data moved between regions or availability zones.

**Snowflake data transfer** is charged when data is transferred out of cloud provider regions — exporting large result sets to external systems, replication to other regions, or data sharing with accounts in different cloud regions.

**BigQuery data transfer** is charged for data exported to non-Google services and for cross-region queries against externally partitioned data.

Minimize transfer costs by: keeping analytics work (queries, transformations) within the same cloud region as the data, avoiding repeated large exports to external systems (use the warehouse as the serving layer rather than exporting to intermediate stores), and configuring replication only for data that genuinely requires cross-region presence.

Total Cost of Ownership

Cloud data warehouse costs are not only platform charges. The total cost of ownership includes:

**Data team labor:** Engineers designing and maintaining the warehouse and its pipelines. A warehouse that requires constant tuning and optimization consumes engineering hours that have real cost.

**EL tool costs:** Fivetran, Airbyte, and similar tools charge per Monthly Active Row or connector, adding to the total analytical infrastructure cost.

**BI tool costs:** Tableau Creator/Explorer/Viewer licenses, or equivalent for Power BI or Looker, are often larger than the warehouse itself for large user bases.

**Monitoring and observability tools:** Data catalog, data quality monitoring, and cost alerting tools.

For mid-market organizations, a complete modern data stack (cloud warehouse + EL tool + dbt Cloud + Tableau + monitoring) typically runs $5,000-$25,000 per month in tooling costs depending on scale, before considering engineering labor.

Our data architecture services helps organizations understand, forecast, and actively govern their data warehouse costs — from Snowflake warehouse configuration through BigQuery query optimization and total stack cost analysis. Contact us to discuss your data infrastructure requirements.

A former Microsoft data architect audits your data foundation, identifies your top priorities, and sends you a written plan. Free. No pitch.