Databricks
It bills you in a unit you've never heard of, sends two separate invoices for one workload on AWS or GCP, and changes the price depending on whether a human is watching the cluster. A quoted "$1,000/month" honestly lands closer to $2,000-3,000.
What is Databricks?
Databricks, founded by the original creators of Apache Spark, is a unified analytics platform built around the "lakehouse" architecture — combining the structure and performance of a data warehouse with the flexibility and scale of a data lake, all in one system spanning data engineering, data science, machine learning, and SQL analytics. Unity Catalog provides centralized governance across the whole platform, the Photon query engine accelerates SQL workloads, and Mosaic AI extends the platform into full generative and agentic AI territory. A genuinely current, notable addition worth knowing about is Agent Bricks, which lets you assemble specialized AI agents — a Knowledge Assistant that answers questions about your documents with citations, Genie Spaces for natural-language SQL querying — under a coordinating Supervisor Agent. In February 2026, Knowledge Assistant became available by default for Enhanced Security and Compliance customers with HIPAA controls selected, a real, current expansion into regulated healthcare use cases specifically.
The genuinely honest, blunt thing worth knowing before budgeting for Databricks seriously: one detailed 2026 review puts it plainly — "Databricks is a genuinely great platform. It's also priced like a riddle. It bills you in a unit you've never heard of, sends two invoices for one workload, and changes the price depending on whether a human is watching the cluster." That unit is the Databricks Unit (DBU), a normalized measure of compute capacity, with real per-DBU rates varying significantly by workload type: Jobs Compute for automated workloads runs roughly $0.10-0.15 per DBU, All-Purpose Compute for interactive development runs $0.40-0.55, SQL Warehouses for analytics run $0.22-0.40, and Serverless SQL and Compute carry premium rates typically 20-40% higher than standard. The genuinely important, easy-to-miss catch: on AWS and GCP, your cloud provider bills infrastructure costs entirely separately from your Databricks DBU bill, and that separate cloud infrastructure routinely runs 50% to 100% or more of your DBU charges — meaning a quoted "$1,000 a month for Databricks" honestly lands closer to $2,000-3,000 once the cloud bill arrives. Azure is the one genuine exception, since Databricks operates there as a first-party Microsoft service with everything consolidated onto a single Azure bill.
The "priced like a riddle" characterization and the concrete "$1,000 becomes $2,000-3,000" example are drawn directly from a detailed independent 2026 pricing analysis (Mammoth.io). The current Standard tier retirement timeline is drawn directly from Flexera's independent FinOps pricing guide, corroborated by Revefi's separate breakdown. The Agent Bricks Supervisor billing mechanic is drawn directly from Databricks' own official Agent Bricks pricing documentation.
Key features
Lakehouse architecture
Combines data warehouse structure with data lake flexibility and scale.
Unity Catalog
Centralized governance, access control, and audit logging across the platform.
Agent Bricks
Assemble Knowledge Assistant, Genie Spaces, and other agents under a Supervisor.
Photon query engine
Accelerated SQL performance for analytics workloads at scale.
Mosaic AI
Full generative and agentic AI development tooling built into the platform.
Apache Spark & Delta Lake core
Built on genuinely open-source foundations, usable independently outside Databricks.
Pricing
Free Trial
- No permanent free tier for the managed platform
- Apache Spark and Delta Lake remain genuinely free to use independently
- Good for testing real workloads before committing to a paid tier
Premium
- Includes Unity Catalog, RBAC, audit logging, serverless compute, Mosaic AI
- Standard tier retired (AWS/GCP: Oct 2025; Azure: blocked Apr 2026, full migration Oct 2026)
- Real total cost typically 2-3x the DBU charges alone once cloud infra is added
Enterprise
- Negotiated directly, prepaid DBU bundles available above $100K/year spend
- Buyers commonly secure below-list pricing via volume and multi-year commitments
- Best for organizations needing the deepest compliance and governance controls
Agent Bricks' Supervisor Agent bills for itself plus every sub-agent used — a Supervisor running Knowledge Assistant and Genie Spaces together generates three separate charged line items, not one. Prices reflect Databricks' published pricing as of July 2026.
Available models
Integrations & platforms
Pros, cons & best for
Pros
- Genuinely powerful, unified platform spanning data engineering through agentic AI
- Real, ongoing investment in agent orchestration through Agent Bricks and Mosaic AI
- Built on genuinely open-source foundations, avoiding total platform lock-in
Cons
- Genuinely complex, two-invoice billing on AWS and GCP inflates real total cost
- Standard tier retirement forces real migration work for existing Azure customers
- Agent Bricks' stacked sub-agent billing can surprise teams expecting one flat charge
Best for
- Enterprises needing a genuinely unified data engineering and AI platform at scale
- Organizations with dedicated FinOps capacity to manage DBU and cloud cost together
- Not the pick for smaller teams wanting simple, single-invoice, predictable billing
Take a look inside
Alternatives
For a simpler cloud data warehouse or a more accessible predictive analytics tool instead:
Our verdict
Databricks' genuine strength is real, unified platform depth — spanning data engineering, data science, machine learning, and SQL analytics on a lakehouse architecture, with genuinely current investment in agentic AI through Agent Bricks and Mosaic AI, all built on open-source Apache Spark and Delta Lake foundations that avoid total vendor lock-in. The honest, important thing worth budgeting for carefully: pricing genuinely earns its "riddle" reputation, since Databricks Units bill separately from real cloud infrastructure costs on AWS and GCP, routinely adding 50-100% or more on top of the quoted DBU rate — meaning a real total cost commonly runs 2-3x what a bare DBU estimate alone would suggest. Combined with the currently unfolding Standard tier retirement (already complete on AWS and GCP, actively migrating on Azure through October 2026) and Agent Bricks' stacked sub-agent billing, this is a platform that rewards teams with real FinOps discipline far more than a quick, naive cost estimate. For enterprises needing genuine unified data and AI infrastructure at scale, Databricks remains a strong, powerful choice; for smaller teams wanting simple, predictable, single-invoice billing, the complexity here deserves careful planning first.
FAQ
Why does Databricks send two bills for one workload?
On AWS and GCP, your cloud provider bills the actual server infrastructure entirely separately from Databricks' own Databricks Unit (DBU) charges, and that separate infrastructure bill routinely runs 50-100% or more on top of the DBU charges — Azure is the exception, since it's consolidated onto one bill there.
If Databricks quotes me $1,000/month, what will I actually pay?
Realistically closer to $2,000-3,000 a month once the separate cloud infrastructure bill lands, on AWS or GCP specifically — factor this in from day one rather than being surprised by it later.
Is the Standard tier still available on Databricks?
Not on AWS or GCP, where it reached end-of-life in October 2025, and on Azure, new Standard workspace creation was blocked as of April 1, 2026, with all remaining Azure Standard workspaces set to automatically upgrade to Premium by October 1, 2026.
How does Agent Bricks billing work if I use multiple agents?
The Supervisor Agent bills for itself plus every sub-agent it coordinates — if your Supervisor includes both Knowledge Assistant and Genie Spaces, you're charged for three separate items: the Supervisor, the Knowledge Assistant agent, and Genie Spaces' SQL queries.
Does Databricks offer a free plan?
Not a permanent one for the managed platform — a 14-day free trial with $400 in credits is available, though the underlying open-source components (Apache Spark, Delta Lake) remain genuinely free to use independently outside Databricks itself.
What is a Databricks Unit (DBU)?
A normalized measure of compute capacity that Databricks uses as its own billing currency, with real per-DBU rates varying by workload type — roughly $0.10-0.15 for automated jobs, $0.40-0.55 for interactive development, and $0.22-0.40 for SQL analytics.