Google BigQuery is a fully managed, serverless data warehouse that separates compute from storage completely — there is no cluster to provision, no nodes to size, and no infrastructure to maintain. SQL queries execute against petabyte-scale datasets in seconds by dynamically allocating compute slots from Google's shared infrastructure pool. BigQuery's architecture represents the most mature implementation of the serverless analytics warehouse model: the same SQL interface works whether the dataset has 1,000 rows or 1 trillion rows, and cost scales with data scanned rather than with cluster uptime.
BigQuery stores data in Capacitor — Google's proprietary columnar format — distributed across thousands of nodes in Google's Colossus distributed file system. Unlike Redshift, there are no explicit compute nodes to configure: BigQuery uses a shared pool of query workers called slots. One slot is approximately one virtual CPU for query execution. On-demand pricing allocates slots automatically for each query and charges per TB of data scanned. Flat-rate pricing purchases a dedicated slot commitment (100-slot minimum) and all queries run against that commitment, making costs predictable for high-volume analytical workloads.