Dremio

Overview

Dremio is a lakehouse query platform rather than a full stack: it owns no storage and runs no Spark. What it brings is a fast SQL engine built on Apache Arrow, federation across dozens of sources, and an Iceberg-native catalogue, so data can be queried where it already lives.

Its open-source pedigree is unusual. Dremio co-created Apache Arrow, is a lead contributor to Apache Iceberg and co-created Apache Polaris, which its Open Catalog is built on.

The performance trick is reflections: materialised structures the engine maintains and transparently substitutes into queries. Autonomous reflections go further, analysing workloads and creating them without anyone tuning by hand, alongside automatic Iceberg compaction and metadata cleanup.

Federation is broad: Redshift, BigQuery, Snowflake, SQL Server, Postgres, Oracle, MongoDB, SAP HANA, Teradata and object storage on all three clouds, with catalogues including Glue, Hive, Nessie, Unity Catalog and OneLake.

The recent direction is agents: an AI semantic layer that generates documentation and labels, a natural-language agent that writes and validates SQL, and an MCP server so external agents can query through the same governed layer. It does not train or serve models itself.

The ownership change matters for anyone choosing today: SAP completed its acquisition of Dremio in July 2026, and SAP says it will make its Business Data Cloud an Iceberg-native lakehouse. Terms were not disclosed.

Key features and capabilities

The same headings are used for every lakehouse platforms entry, so two tools can be read side by side.

Architecture and storage

• No storage of its own: it queries your object storage and source systems in place
• Distributed engine that scales to a thousand nodes, or a single node for development
• Dremio Cloud is fully managed with automatic scaling; engines bill only while running
• Compute and storage fully separated

Engines and workloads

• One engine: SQL on Apache Arrow. No Spark
• Reflections accelerate queries, with automatic compaction and table maintenance
• BI served through Arrow Flight SQL drivers rather than a bundled BI tool
• Notebooks and streaming are not first-party features

Table formats and catalogues

• Table formats: Apache Iceberg and Delta Lake
• File formats: Parquet, CSV, JSON, Excel and ORC, the last only from a metastore source
• Catalogues: its own Open Catalog on Polaris, plus Glue, Hive, Nessie, Unity Catalog and OneLake
• Federation to Redshift, BigQuery, Snowflake, SQL Server, Postgres, Oracle, MongoDB, SAP HANA and more

Security and governance

• Role-based access control with fine-grained permissions
• Open Catalog provides unified governance across data assets
• Row and column policies come through integrations such as Lake Formation, Privacera or Ranger
• Identity through OIDC or LDAP, with audit logging; named certifications are not published

AI and machine learning

• An AI agent that turns questions into validated SQL with visualisations
• An MCP server so external agents can query through the governed layer
• Semantic search and auto-generated documentation over the semantic layer
• No model training, serving or vector search; it is the data layer, not the ML platform

Ecosystem and integrations

• BI clients including Power BI, Tableau, Looker, Excel, Superset and ThoughtSpot
• Arrow Flight SQL JDBC and ODBC drivers, and PyArrow for custom applications
• TLS enabled by default for Arrow Flight clients
• No marketplace; dbt support is not published

Where it runs and what it costs

• Dremio Cloud, fully managed on AWS, with Azure stated as coming
• Dremio Enterprise self-managed on Kubernetes, cloud, or on-premises
• Quickstarts for AWS, Azure, Google Cloud, Kubernetes, Docker and YARN
• Region lists are not published

Pricing

Usage-basedCloud at $0.20 per compute unit

Dremio Cloud is $0.20 per Dremio Compute Unit, measured as engine size multiplied by minutes running, so a 20-unit engine for ten hours consumes 200 units. A 30-day trial comes with $400 of credit. Dremio Enterprise, the self-managed edition, is quote-only. A free community edition runs on a local machine or server.

Vendor pricing page →

Demos and videos

• Dremio channel Channel or playlist · YouTube
• Dremio resource library Product demo · dremio.com

About Dremio (part of SAP)

Dremio was founded in 2015 by people from the open-source big data world, including co-creators of Apache Arrow, and is based in Santa Clara. It raised through to a $160m Series E in 2022 at a $2bn valuation, with investors including Insight Partners and Norwest. SAP completed its acquisition of the company in July 2026 on undisclosed terms, and Dremio's own pages now describe it as part of SAP. The product is commercial, with a free community edition; Arrow, Iceberg and Polaris remain Apache-licensed upstream projects.

Founded 2015 · Santa Clara, California · dremio.com