Firebolt

Overview

Check the licence before comparing Firebolt against what you remember of it, because it changed direction. It now describes itself as a fully open-source analytical database that can be self-hosted without limits, with a managed service and a bring-your-own-cloud option sold alongside.

Read the status carefully though: the documentation states that Firebolt OSS is in Preview, that general availability launches in September, and that code access currently comes by emailing support. So the open-source claim on the pricing page is ahead of what is actually downloadable, and anyone planning to self-host should confirm where that has landed.

The technical shape is a single binary that runs on a laptop and scales out to hundreds of nodes over object storage in your own cloud, with the same behaviour at both ends. It is PostgreSQL-compatible, reads open formats including Iceberg, and targets low-latency, high-concurrency queries and efficient ELT rather than long batch analytics.

Pricing is the most concrete in this category. There is a published engine calculator: choose a node size from S to XL, choose compute-optimised or storage-optimised hardware, set the node count from 1 to 128 and a data volume, and it returns an all-in monthly cost. Compute runs on Arm and bills per second, storage is passed through at object-storage cost, and engines scale to zero.

Per-second billing with scale-to-zero is the part worth holding against the rest of the category: an idle engine costs nothing, which changes the economics of bursty or intermittent workloads compared with anything billed by reserved capacity.

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

• A single binary that runs on a laptop and scales out to hundreds of nodes
• Data sits on object storage in your own cloud
• Engines scale to zero when idle
• Scale up node size before scaling out node count, which the pricing tool makes explicit

Performance and scaling

• Built for low-latency queries and high concurrency rather than long batch jobs
• All nodes in an engine work together on a single SQL query
• Compute runs on Arm hardware, with compute-optimised and storage-optimised node types
• Efficient ELT is named as a target workload alongside querying

SQL and workloads

• PostgreSQL-compatible, so existing tooling and dialect knowledge carry over
• The same engine behaves identically from laptop to production cluster

Open formats and sharing

• Positioned as fully open source, self-hostable without limits
• Reads open formats, including Apache Iceberg
• The open-source edition is marked Preview in the documentation, with code access by emailing support

Security and governance

• Data remains in your own cloud account under the bring-your-own-cloud and self-hosted options
• Governance detail is thinner in public documentation than performance and cost detail

AI and machine learning

• Explicitly built for AI agents as well as engineers, per its own documentation
• Low-latency querying is the stated fit for agent-driven access patterns

Where it runs and what it costs

• Self-hosted, managed service, or bring-your-own-cloud
• Six published regions across the US, Europe and Asia Pacific
• Per-second billing on compute, with storage passed through at object-storage cost

Pricing

Usage-basedFree to self-host; managed billed per second

Self-hosting is free and unlimited. The managed service bills compute per second with scale to zero, and passes storage through at object-storage cost. A public calculator gives an all-in monthly figure from node size, hardware type, node count and data volume — rare transparency for an analytical database.

Vendor pricing page →

Demos and videos

• Firebolt documentation Tutorial · docs.firebolt.io (the vendor)

About Firebolt

Firebolt develops an analytical database it now positions as open source, sold as a managed service and as a bring-your-own-cloud deployment, with self-hosting offered without limits. Its documentation is candid about maturity: the open-source edition is marked Preview, with general availability stated for September and code access by request through support. Unusually for the category it publishes a working cost calculator rather than a contact form, covering six regions across the US, Europe and Asia Pacific.

firebolt.io