Overview
Snowflake is a cloud-native data platform with separation of storage and compute, multi-cluster scaling, and native support for structured, semi-structured, and unstructured data. The platform runs as a managed service on AWS, Azure, and GCP, with the same architecture and customer experience across clouds. Snowflake's data sharing and clean room capabilities are differentiators that have driven adoption in financial services, retail, and media verticals where cross-organisation data collaboration is increasingly required.
Snowflake has expanded materially beyond data warehousing into Snowpark (Python/Java/Scala compute), Cortex AI (managed LLM access), Streamlit-based applications, and the Native App Framework. Pricing is consumption-based, which is both a strength (no capacity planning) and a risk (runaway queries). FinOps discipline and warehouse-sizing governance are essential for any production deployment.
Key Features
- Multi-cluster shared-data architecture with separate compute and storage
- Auto-suspend and auto-resume virtual warehouses
- Time Travel for historical data queries (up to 90 days)
- Zero-copy cloning of databases and tables
- Native semi-structured data types (VARIANT, OBJECT, ARRAY)
- Snowpark for Python, Java, Scala compute
- Cortex AI for managed LLM functions and embeddings
- Snowflake Data Sharing and Snowflake Marketplace
- Native App Framework for distributable applications
- Streamlit in Snowflake for embedded analytics apps
- Dynamic Tables for declarative data pipelines
- Streams and Tasks for change data capture workflows
Pricing
| Edition | Model | Typical Cost |
|---|---|---|
| Standard Edition (compute) | Per credit-hour | $2.00/credit-hour (AWS us-east-1) |
| Enterprise Edition | Per credit-hour | $3.00/credit-hour |
| Business Critical Edition | Per credit-hour | $4.00/credit-hour |
| VPS / Government | Per credit-hour | Quote required |
Pricing verified May 2026. Storage priced separately (~$23/TB/month on-demand). Annual commitments yield significant discounts vs on-demand. Cortex AI and Snowpark have separate consumption metering.
Strengths
- Cleanest separation of compute and storage among major data warehouses
- Multi-cloud parity — same product across AWS, Azure, GCP
- Data sharing and clean rooms enable cross-organisation use cases
- Cortex AI delivers production-ready LLM functions without leaving the platform
- Strong ecosystem of partners, ISVs, and data providers
Limitations
- Consumption pricing produces unpredictable bills without governance
- Snowpark is improving but trails Databricks for ML and large-scale data engineering
- Per-second billing introduced 2023, but minimum 60-second charge persists
- Some advanced features (Search Optimization, Materialized Views) drive separate costs
- Egress costs apply for cross-region or cross-cloud queries
Buyer Considerations
Snowflake cost governance is the single biggest determinant of total ownership cost. The platform's elasticity is a strength when paired with disciplined warehouse sizing, query auditing, and resource monitoring; without those practices, costs escalate rapidly. Mature deployments invest in dedicated FinOps capability within the data platform team and treat query optimisation as a continuous operational discipline rather than a one-time tuning exercise.