Home / Case Studies / Databricks DBU Commit Restructure
DatabricksLife SciencesDBU CommitMosaic AI

Databricks DBU commit reshaped, Mosaic AI at parity.

A life sciences group was negotiating a three-year Databricks commit with an opening proposal that bundled All-Purpose and Jobs DBUs at a blended rate and priced Mosaic AI outside the master discount. Fourteen weeks later, the commit was restructured for $2.7M of measurable saving, with All-Purpose and Jobs SKUs separated to actual workload mix and Mosaic AI brought inside master-discount parity.

Cloud data platform engineering workspace
$2.7M
Three-year saving
21%
Reduction vs. opening commit
14 wk
Kick-off to signature
Parity
Mosaic AI master discount
The contract going in

Blended DBU rate, Mosaic on the side.

The group's existing Databricks commit had been negotiated against an early-stage adoption plan. Two years in, the workload mix was settled: roughly two-thirds Jobs and one-third All-Purpose, with a strong push from research teams to standardise on Mosaic AI for model training. The opening renewal continued to price All-Purpose and Jobs at a blended rate that overpaid for the Jobs portion.

  • Annual DBU spend running at $5.4M, with consumption tracking 8% above commit.
  • Workload mix settled at 65% Jobs, 35% All-Purpose, with the prior commit priced as a blend.
  • Mosaic AI adoption decision already made by the research function, pricing held outside master discount.
  • Renewal cycle overlapping with a clinical data platform consolidation programme.
Databricks's opening position

Larger blended commit, Mosaic at list, growth uplift.

The Databricks account team proposed a commit increase of approximately 16%, held the blended DBU rate, and quoted Mosaic AI at list with a separate growth uplift. The proposal also framed Mosaic adoption as a future commitment that would unlock incremental discount only at year two.

What we flagged

A blended DBU rate masks the difference between Jobs and All-Purpose. When the actual workload skews toward Jobs, the blend systematically overpays. SKU separation is the lever, not headline commit size.

The work

Fourteen weeks. Four workstreams.

1. Workload mix evidence

We pulled twelve months of DBU consumption by SKU and built a defensible workload split. The output: actual Jobs/All-Purpose ratio, with a forecast that named the consolidation programme as a likely shift toward Jobs.

2. SKU separation

Negotiated separate Jobs and All-Purpose unit prices reflecting actual workload mix, removing the blended-rate overpay on the Jobs portion.

3. Mosaic AI parity

Brought Mosaic AI inside the master discount at SKU parity from year one, with no separate growth uplift schedule.

4. Commercial position

We drafted the position paper. The group's data platform director presented it. Our team ran the three follow-up sessions across commercial, finance and technical reviews.

Lesson

Databricks commit conversations default to blended rates because they're easier to model. The leverage comes from refusing the blend and pricing SKUs against actual workload mix. The Mosaic question negotiates better when it's part of the master, not a separate add-on.

The contract going out

Workload-priced commit, Mosaic at master discount.

The signed commit separated Jobs and All-Purpose SKUs at unit prices reflecting actual workload mix, brought Mosaic AI inside the master discount at parity, and removed the separate growth uplift schedule.

$2.7M
Saved
Versus the opening Databricks proposal, measured over the three-year term.
Separated
Jobs vs All-Purpose
SKUs priced to actual workload mix rather than blended rate.
Parity
Mosaic AI
Inside master discount from year one, no separate uplift.
“Once we separated the SKUs, the discussion became measurable. The Mosaic parity meant the research function did not have to defend their adoption choice in commercial terms. The negotiation worked because both points were anchored in evidence.”
Director of Data Platform · Life Sciences · Anonymised by client request
Related resources

More on Databricks strategy.

← Back to all case studies

Renewing a Databricks DBU commit?

Tell us the commit shape, the workload mix and the Mosaic position. We will respond within one business day with the lead and the most relevant precedent.