A cloud data warehouse is a managed, analytics-first database that stores large volumes of structured and semi-structured data and answers complex queries over it quickly. What separates the cloud generation from the on-prem warehouses that came before is an architectural shift: the separation of storage and compute. Data sits in cheap, durable cloud storage, and query engines spin up against it on demand - so you scale the two independently, pay for compute only while it runs, and stop sizing a single box for your worst-case workload.
What makes them fast
Cloud warehouses are built for analytics, not transactions. They store data in columnar format, so a query that touches three columns of a hundred-column table reads only those three. They use massively parallel processing (MPP), splitting a query across many nodes that each handle a slice of the data. And they lean on pruning - skipping blocks of data that can't match a query's filters - to avoid scanning what they don't need. Together these let a warehouse aggregate billions of rows in seconds.
The major platforms
Snowflake popularized the storage-compute split with its virtual-warehouse model and features like zero-copy cloning and data sharing. Google BigQuery takes a serverless approach with no clusters to manage. Amazon Redshift and Azure Synapse integrate tightly with their respective clouds. Databricks blurs the line with the lakehouse, putting warehouse-style SQL over open data-lake storage. And ClickHouse serves the real-time analytics end, where sub-second queries on fresh data matter most.
Warehouse, lake, and lakehouse
The category is converging. A data lake stores raw files cheaply but lacks the structure and performance of a warehouse; the lakehouse pattern adds table formats like Delta Lake and Apache Iceberg over lake storage to get warehouse-like reliability and speed on open files. The practical result is that "cloud data warehouse" now spans classic warehouses, lakehouse platforms, and fast analytical databases - and many organizations run more than one.
Choosing one
- General-purpose analytics with rich SQL and ecosystem? Snowflake or BigQuery.
- Already deep in AWS or Azure? Redshift or Synapse for tight integration.
- Lakehouse with Spark and ML alongside SQL? Databricks.
- Real-time, high-ingest analytics with sub-second queries? ClickHouse.
The differences that matter in practice are pricing model, concurrency behavior, ecosystem fit, and how each handles semi-structured data. Match the platform to your cloud, your workload shape, and how you want to pay for compute.