Snowflake
Also known as: Snowflake DW, Snowflake data cloud, SFDC (data)
Snowflake is a cloud-based data warehouse platform.
Companies store all their business data in Snowflake and query it with SQL. Its unique architecture separates storage from compute -- you can run 10 query workloads simultaneously without them competing, and you pay for compute only while queries are actually running.
Snowflake, explained at three levels
At home
Snowflake is a place where companies store and analyze all their data. You write SQL queries and get answers from billions of rows of data in seconds -- without managing any database servers. The costs scale with how much computing power your queries use.
At work
Data teams use Snowflake as their central data warehouse. All the company's data -- from sales systems, marketing tools, product databases -- is loaded into Snowflake, where analysts can query it with SQL to answer business questions.
At home
Snowflake's key innovation is separating storage (data sits in S3/Azure Blob/GCS) from compute ('virtual warehouses' that query it). Different teams get their own virtual warehouses that don't compete for resources. You pause a warehouse when not in use and pay nothing. Auto-suspend and auto-resume happen automatically.
At work
Snowflake features: Time Travel (query data as it existed up to 90 days ago), Data Sharing (share live data with partners without copying), Streams and Tasks (change data capture and scheduled jobs), and Snowpark (run Python, Java, or Scala inside Snowflake). dbt (data build tool) is the most popular tool for transforming data in Snowflake.
At home
Snowflake's storage layer uses micro-partitions -- small (50-500 MB) columnar files that are automatically clustered and pruned at query time. The metadata layer tracks min/max values per micro-partition, letting the query planner skip vast amounts of data. This eliminates the need for manual indexing.
At work
Snowflake's multi-cluster warehouse feature automatically adds compute clusters when concurrency demand spikes, then removes them. Data Sharing uses a unique architecture: no data is copied -- recipients query the provider's storage directly through a cross-account metadata pointer. Snowflake Horizon provides unified governance: column masking, row access policies, object tagging, and cross-cloud lineage.
Real-world examples
At home
Carlos's team loaded 5 years of customer transaction data (3 billion rows) into Snowflake. Analysts can now query the entire history in under 10 seconds and build dashboards in Tableau, which connects to Snowflake directly.
At work
Fatima's company uses Snowflake's Data Sharing feature to share live inventory data with three retail partners. Partners query Snowflake directly without receiving data exports -- the data is always current and Fatima's team controls access.
Common confusions
- Snowflake is a cloud provider — Snowflake is a data warehouse service that runs on top of AWS, Azure, or GCP. You choose which cloud provider's region to deploy in, but Snowflake manages all the infrastructure.
- Snowflake replaces your production database — Snowflake is optimized for analytics -- large reads and complex queries. It's not designed for transactional workloads (lots of individual row inserts and updates). Your application database (PostgreSQL, MySQL) stays separate; data is replicated to Snowflake for analysis.
For analysts
Snowflake is the central tool for most modern data teams. Business analysts write SQL queries, build views, and connect BI tools (Tableau, Power BI, Looker) directly. Snowflake's Time Travel feature lets analysts undo accidental data changes by querying previous versions.
History & origin
Snowflake was founded in 2012 by former Oracle engineers who set out to build a data warehouse designed for the cloud from scratch. It launched publicly in 2014. In September 2020, Snowflake held the largest software IPO in history at the time, valuing the company at over $70 billion. Its success came from solving real pain points: traditional data warehouses required expensive, rigid hardware, while Snowflake offered elastic, pay-per-query compute on any cloud (AWS, Azure, GCP).