Snowflake: The Cloud Data Warehouse, Its Credits, and Its Rivals
What it is: Snowflake is a cloud data warehouse with auto-scaling, pay-per-use pricing, and complete separation of storage and compute. It runs identically on AWS, Azure, and GCP with zero infrastructure to manage.
Quick answer: Snowflake is a fully managed cloud data warehouse that bills compute in "credits" — one credit ≈ $2-$4 depending on edition, and an X-Small virtual warehouse burns 1 credit/hour only while running. Storage is separate at ~$23-$40/TB/month. It's easier to operate than Redshift and more multi-cloud than BigQuery, at a price premium for always-on workloads.
How does Snowflake pricing actually work?
Snowflake charges for three things: compute (credits), storage (per TB), and cloud services (usually free below 10% of compute). A credit is the unit of compute time: an X-Small warehouse consumes 1 credit per hour, and each size up (S, M, L, XL...) doubles the burn — an XL uses 16 credits/hour. Credit price depends on edition: roughly $2 (Standard), $3 (Enterprise), $4+ (Business Critical). Two details dominate real bills: warehouses bill per second after the first 60, and auto-suspend means an idle warehouse costs nothing — so a query cluster that runs 2 hours/day on a Medium costs about 2 × 4 credits × $3 ≈ $24/day, not $288.
-- The single most important cost control in Snowflake
ALTER WAREHOUSE analytics_wh SET
AUTO_SUSPEND = 60 -- suspend after 60s idle
AUTO_RESUME = TRUE;
-- And right-size: try smaller first, sizes double cost
ALTER WAREHOUSE analytics_wh SET WAREHOUSE_SIZE = 'SMALL';Snowflake vs Redshift vs BigQuery: which should you pick?
Pick Snowflake for the least operational work and true multi-cloud; pick Redshift when you're deep in AWS and want reserved-instance economics for steady workloads; pick BigQuery for serverless simplicity and tight Google integration. Rough differences:
Pricing model: Snowflake credits per second of warehouse uptime; Redshift per node-hour (or RPU-hour for Serverless); BigQuery $6.25/TB scanned or slot reservations
Operations: Snowflake near-zero tuning; Redshift needs distribution/sort key thinking; BigQuery zero-ops serverless
Cloud: Snowflake on AWS/Azure/GCP; Redshift AWS-only; BigQuery GCP-only
Best-case cost: Redshift cheapest for 24/7 predictable load; Snowflake and BigQuery cheaper for spiky, intermittent analytics
What It Does Best
Scale without thinking. Query performance stays consistent. Add compute instantly, remove when done.
Multi-cloud. Runs on AWS, Azure, GCP. Same SQL, same features everywhere.
Zero maintenance. No servers to manage. No tuning. No indexes. Just write SQL.
Key Features
Separation of storage and compute: Scale independently, pause compute when not in use
Virtual warehouses: Multiple isolated compute clusters that never contend
Zero-copy cloning: Instant database copies for dev/test
Time Travel: Query historical data up to 90 days back
Data sharing: Share live data across accounts securely
Pricing
Compute: Credits at ~$2-$4 each by edition; X-Small = 1 credit/hour, each size doubles it; per-second billing with auto-suspend
Storage: ~$23/TB/month (capacity, pre-purchased) to ~$40/TB/month (on-demand)
Typical cost: $200-2,000/month for small teams, $10k+/month for heavy users
Free trial: 30 days with $400 of credits
When to Use It
✅ Analytics on large datasets (TB+)
✅ Variable workload (scale up/down)
✅ Multiple teams needing isolated compute
✅ Don't want to manage infrastructure
✅ Need data sharing across organizations
When NOT to Use It
❌ Transactional database needs (use Postgres)
❌ Small datasets (< 100GB, Postgres cheaper)
❌ Need sub-second query response (not optimized for OLTP)
❌ Budget very tight (can get expensive without warehouse discipline)
❌ Heavy real-time streaming (use specialized tools)
Common Use Cases
Data warehouse: Central repository for business analytics
BI dashboards: Power Tableau, Looker, Power BI
Data science: Large-scale data preparation for ML
Data sharing: Share data with partners/customers
ELT pipelines: Load and transform data at scale
Snowflake vs Alternatives
vs BigQuery: Snowflake multi-cloud with predictable warehouse sizing; BigQuery serverless with per-TB-scanned billing and tighter Google integration
vs Redshift: Snowflake easier to operate and scales elastically; Redshift cheaper on AWS for steady 24/7 workloads with reserved pricing
vs Databricks: Databricks better for ML/Spark and data engineering; Snowflake better for pure SQL analytics
Unique Strengths
True multi-cloud: Same experience on AWS, Azure, GCP
Compute isolation: Workloads never interfere with each other
Zero-copy clone: Instant environments without data duplication
Data marketplace: Access to third-party data sets
Bottom line: Data warehouse made easy. The credit model rewards spiky workloads and punishes forgotten always-on warehouses — set auto-suspend and right-size, and Snowflake is often cheaper than it looks. If you're drowning in data and query performance, it solves the problem with the least engineering effort of the big three.