Google BigQuery

Overview

BigQuery is Google Cloud's data warehouse, and the word that defines it is serverless. There are no clusters to size or suspend: you submit SQL and Google allocates the compute, with storage and compute scaling separately so that neither affects the other's performance or availability.

Because of that, the pricing question differs from its rivals. You either pay per terabyte scanned on demand, or buy capacity as slots through the editions model, where Standard, Enterprise and Enterprise Plus differ in features and rate. Slots come with a 50-slot minimum and increments of 50.

The query language is ANSI-standard GoogleSQL, alongside a pandas-compatible Python DataFrames API, so analysts and Python teams use the same warehouse without a bridge between them.

BigQuery ML is the part that surprises people. Models are trained and served in SQL, from regression and clustering to ARIMA forecasting and the pre-trained TimesFM model, with generative and embedding functions and vector indexes for semantic search on top.

Openness is delivered through BigQuery tables for Apache Iceberg, which are read-write and keep data in your own Cloud Storage bucket, exporting Iceberg metadata so Spark and other engines can read the same files.

Multi-cloud is handled by BigQuery Omni, which runs Enterprise-edition queries inside your own AWS or Azure account so the data never leaves it.

Governance covers IAM down to table level, column and row access controls, dynamic masking, VPC Service Controls, audit logging, data profiling, quality scans and lineage, plus zero-copy sharing and clean rooms.

Key features and capabilities

The same headings are used for every cloud data warehouses entry, so two tools can be read side by side.

Architecture and storage

• Serverless, with no infrastructure to provision or suspend
• Storage and compute are separate layers, scaled independently
• Columnar managed storage, billed on either a logical or a physical model
• Capacity bought as slot reservations, minimum 50 slots and increments of 50

Performance and scaling

• Slots are the unit of compute, with an autoscaler or a fixed baseline
• Standard bills per second with a one-minute minimum; opt-in fluid scaling removes the minimum
• BI Engine adds in-memory acceleration for dashboards
• One and three-year commitments cut the slot rate

SQL and workloads

• ANSI-standard GoogleSQL, including DML against Iceberg tables
• A pandas-compatible Python DataFrames API alongside SQL
• Streaming ingestion through the Storage Write API
• Notebooks and version control in BigQuery Studio

Open formats and sharing

• BigQuery tables for Apache Iceberg are read-write, stored in your own Cloud Storage bucket
• Iceberg V2 snapshots are exported and refreshed so Spark and other engines can read them
• Iceberg REST catalogue and Delta Lake support are not published on the Iceberg tables page
• Zero-copy live sharing and data clean rooms, with free inbound data transfer

Security and governance

• IAM permissions on projects, datasets, tables and views
• Column-level and row-level access control, plus dynamic data masking
• Encryption in transit and at rest, with VPC Service Controls perimeters
• Audit logging, data profiling, quality scans, and table and column lineage
• Named compliance certifications are not listed on the governance page

AI and machine learning

• BigQuery ML trains and serves models in SQL, including ARIMA forecasting and the pre-trained TimesFM model
• Externally trained models, AutoML, and remote models against deployed endpoints
• Generative functions, embeddings and vector indexes for semantic search and retrieval
• Gemini assistance for SQL and Python, plus conversational analytics

Where it runs and what it costs

• More than 60 Google Cloud regions, plus US and EU multi-regions
• BigQuery Omni runs queries in your own AWS or Azure account, across seven listed regions
• No self-hosted option
• Charged per terabyte scanned or per slot-hour, plus storage

Pricing

Usage-based1 TiB of queries free each month, then $6.25

Two models. On demand, the first terabyte each month is free and further queries cost $6.25 a TiB. Editions bill per slot-hour in Iowa at $0.04 Standard, $0.06 Enterprise and $0.10 Enterprise Plus, falling by up to a fifth on a three-year commitment, with a 50-slot minimum. Storage is billed per GiB with the first 10 GiB free, active logical storage working out near $0.023 a GiB a month and physical storage cheaper to hold. Inbound transfer is free. A sandbox gives 10 GiB and 1 TiB of queries a month with no card.

Vendor pricing page →

Demos and videos

• What is BigQuery? Product demo · YouTube
• How to get started with BigQuery Tutorial · YouTube

About Google Cloud

BigQuery is a first-party Google Cloud service, so it has no funding history or ownership of its own. Google announced it on its Cloud Platform blog in November 2011, which makes it one of the oldest cloud warehouses still in mainstream use. It is proprietary, with no open-source or self-hosted edition, billed through a Google Cloud billing account and quoted in US dollars, with other currencies handled through Cloud Platform SKUs. Openness comes from adopting Iceberg and keeping table data in buckets the customer owns.

cloud.google.com