Fabric vs Azure Databricks: monthly TCO inputs
Planning tool, not a vendor quote. No live pricing is fetched. Rates start blank. Enter dated official or contracted rates for one currency and region. Hypothetical scenarios describe workloads; they do not predict runtime or platform sizing.
Fabric
Enter one whole-capacity hourly rate, not a per-CU rate. For reservations, use amortised monthly commitment plus uncovered PAYG separately. Storage/operations and eligible separate Spark charges are additional only when applicable.
Azure Databricks
Enter total billed DBUs for each workload, not DBU/hour. For included serverless infrastructure, use zero external VM node-hours. For classic, enter actual billable driver and worker node-hours and a blended rate, or split externally for unequal node prices.
| Workload | Total DBUs | Rate/DBU |
|---|---|---|
| Jobs / batch | ||
| SQL | ||
| Pipelines / streaming | ||
| ML / other |
Hybrid and project cost
Model controls and assumptions
- GB/TB are decimal planning units here: 1 TB = 1,000 GB. Convert if the vendor uses a different billing definition.
- Billable capacity hours are service availability hours, not job runtime. Two simultaneous jobs do not double active capacity hours.
- Utilisation scenarios are workload sensitivity inputs; multiplying a purchased-capacity bill by utilisation would incorrectly discount idle capacity.
- Reservations are commitments. Pausing capacity does not remove an amortised commitment.
- CU consumption already covered by base capacity is not a second charge. Add separate Spark or overage only as billed.
- DBU quantities and runtime must be measured per platform; this tool does not translate CUs into DBUs.
- Zero optional costs must be justified. Do not interpret defaults as proof that networking, BI or support is free.
- Shared-cost subtraction needs a line-item reconciliation. The hybrid total is an accounting exercise, not a recommendation.
- No tax, FX, inflation or discount assumptions are inserted. No claim of comparative savings is made.