SKU mix, DBU commit shape, Mosaic AI and Model Serving add-ons, and the marketplace stack interaction across AWS, Azure and GCP. Written by the practice that has negotiated more than forty Databricks commits since the platform began offering enterprise contracting.
The Databricks Unit, the DBU, is the unit of consumption that sits underneath every Databricks contract. The DBU rate varies by SKU, by cloud, by region and by purchase channel, and the commit shape determines whether the buyer pays the lowest DBU rate available or a materially higher one for three full years. The decisions that set the commit shape are taken in a six-to-eight week window, and the buyer who walks into that window without an internally validated consumption model will commit to a shape that is wrong.
This guide walks through the decisions, in order, that determine whether a Databricks commit becomes a controlled three-year cost or a lock-in to consumption that the business has already begun to outgrow. It is written for IT, data, procurement and finance leaders who are now sitting across from a Databricks account team for the first or second time.
The guide is divided into eight chapters. Each chapter has a checklist, a decision tree and a worked example drawn from a real engagement. The chapters move from SKU selection through commit shape modelling, marketplace channel choice, Mosaic AI and Model Serving add-on economics, the multi-cloud interaction, the true-up posture and finally the negotiation choices that determine the renewal commercial three years downstream.
This is not a primer on Databricks architecture. We assume readers already understand the difference between Jobs, All-Purpose, SQL Warehouse and Serverless SKUs, the basic Photon mechanics, and the difference between the Workspace and Account-level commits. We have a separate reference paper for that audience — ask us for it directly.
The lead author runs the Databricks and modern-data-stack practice at SoftwareContractNegotiation. The practice has supported more than forty Databricks commits across financial services, healthcare, retail, manufacturing and public sector. The paper draws on outcomes from those engagements, anonymised for confidentiality. Independent firms such as Redress Compliance are referenced where their published analysis informs a specific decision.
SKU-level DBU rates, the difference between Jobs, All-Purpose, SQL Warehouse and Serverless, and why the SKU mix is the first commercial decision.
Flat vs. ramped commit, the three-year horizon, and the over-commit penalty that quietly erodes year-three economics.
AWS Marketplace, Azure Marketplace, GCP Marketplace and direct contracting — the discount stack interaction with each cloud’s commitment programme.
The fast-growing add-on economics, the GPU-backed SKUs, and the line items that look small in year one and dominate by year three.
Workspace-level commits across AWS, Azure and GCP, the account-level pooled commit, and the cross-cloud price-floor mechanics.
The mid-year true-up, the overage rate exposure, and the contractual language that prevents a benign overshoot becoming a price reset.
What Databricks will ask for at renewal, the leverage that survives year three, and the carry-over credit posture.
A complete redacted engagement — SKU mix, commit shape, marketplace channel and the three-year net outcome against the as-quoted proposal.
If your Databricks commit closes within the next 90 days, the work starts now. The first conversation is free of charge and free of obligation.