Firebolt

Firebolt is an open-source analytical database and data warehouse for engineers, run yourself or as a usage-billed managed service on AWS.

Firebolt answers analytical SQL queries over large and fast-changing data, from a single laptop to a cluster in the cloud. It runs as one binary for local development and scales out to hundreds of nodes and petabytes on object storage. The engine speaks a PostgreSQL dialect, and clients connect over the Postgres wire protocol, ADBC or a JSON REST interface. Firebolt, the independent company of the same name, builds the engine, and its open-source release is currently in preview. Compute runs on engines sized by node type and node count, while storage sits apart on object storage. Engines start in seconds and stop when idle, and multi-cluster autoscaling adds and removes clusters as concurrent demand changes. Data arrives through READ_STREAM from Kafka topics, through change data capture from Postgres and MongoDB, and through incremental loading from Amazon S3. External tables and the READ_ICEBERG and READ_DUCKLAKE functions query lake files where they sit. Aggregating indexes, a result cache and subresult reuse cut the work behind repeated queries. CREATE TABLE CLONE copies a table without duplicating its data, and Python UDFs, in private preview, run custom logic beside the data. Firebolt is written for engineers who put analytics inside products: dashboards their customers open, real-time applications and backends that AI agents query. Data teams that already keep Apache Iceberg or DuckLake tables on Amazon S3 can query them without a separate load step. Teams running Postgres or MongoDB can keep analytical tables current through change data capture. A team can start on Firebolt Core, the self-managed edition, and move to the managed service on AWS when it wants Firebolt to operate the engines. The Enterprise and Dedicated editions add AWS PrivateLink, HIPAA support and single-tenant isolation for regulated workloads. An engineer installs the binary with one curl command, builds against it locally and deploys the same engine to the cloud. A typical pipeline lands Kafka events or database changes in tables, then reshapes them with SQL inside Firebolt as ELT. Analysts read the results in Tableau, while application code reaches the same tables through the Python SDK or SQLAlchemy. That puts one engine behind both the internal dashboard and the analytics feature a product ships to its own users.