Microsoft Fabric has restructured enterprise analytics licensing more fundamentally than any Microsoft platform change in the last decade. This guide provides the complete framework: how Fabric is licensed, how to size capacity correctly, how to migrate from Power BI Premium, how to control OneLake storage costs, and how to negotiate Fabric pricing in your next EA. After 500+ Microsoft licensing engagements managing $2.1B in spend, these are the insights that produce 25–40% cost reductions vs Microsoft's standard Fabric pricing.
Guide Contents
Chapter 1
What Microsoft Fabric Is — And Why It Changes Everything
Fabric unifies Power BI, Azure Data Factory, Azure Synapse, and Azure Data Lake into a single capacity model. Understanding the architectural consolidation is the foundation for all licensing and cost decisions.
Chapter 2
The F SKU Licensing Model: Capacity Units, SKU Sizes, and Pricing
F2 through F2048 — the complete SKU table, pay-as-you-go vs reserved pricing, pause/resume mechanics, and MACC eligibility. The rates your account team won't volunteer.
Chapter 3
F SKU vs P SKU: The Migration Decision Framework
Feature parity analysis (every Power BI Premium feature works on F SKUs), cost comparison (F64 vs P1, F128 vs P2), and the commercial migration strategy for organisations mid-EA-term on P SKUs.
Chapter 4
Per-User Licensing Requirements: Who Needs What
Fabric consumers vs creators, Power BI Pro vs PPU requirements, M365 E5 reconciliation, and the break-even model that shows when capacity licensing beats per-user licensing for your organisation.
Chapter 5
Capacity Sizing: The Right-Sizing Methodology
The Capacity Metrics App, CU consumption by workload type, the smoothing model, burst vs sustained demand, and the 30-day measurement protocol for accurate F SKU sizing before committing to reserved pricing.
Chapter 6
OneLake Storage: Pricing, Architecture, and Cost Optimisation
OneLake vs ADLS Gen2, shortcut strategy (reference without copying), Delta table storage growth and VACUUM maintenance, and the 3-year storage cost model for enterprise data volumes.
Chapter 7
EA Negotiation Strategy: Getting the Best Fabric Pricing
Four negotiation levers: reserved capacity vs EA P SKU, MACC linkage, competitive positioning (Databricks, Snowflake, BigQuery), and Power BI Pro bundle discounts. Benchmark pricing data from 500+ engagements.
Chapter 1: What Microsoft Fabric Is
Microsoft Fabric launched in May 2023 and reached general availability in November 2023. It is Microsoft's response to the fragmentation of enterprise analytics infrastructure — the problem that organisations running Power BI, Azure Data Factory, Azure Synapse Analytics, Azure Data Lake Storage, and Azure Machine Learning had five separate billing models, five separate management interfaces, and five separate licence agreements.
Fabric collapses all of these into a single "Software as a Service" platform with a single capacity unit (CU) billing model. Buy one Fabric capacity, and that capacity runs all analytical workloads: data engineering (Spark/notebooks), data integration (pipelines, Dataflows Gen2), data warehousing (T-SQL serverless), real-time intelligence (KQL, Eventstream), Power BI (reports, dashboards, semantic models), data science (ML experiments), and data governance (Purview integration).
The strategic implication for licensing: Fabric eliminates the "which Azure service do I use for this?" decision and replaces it with "how much Fabric capacity do I need?" This is simpler — but the capacity sizing decision is consequential. Every dollar of capacity you over-provision is wasted; every dollar of capacity you under-provision creates throttling and user frustration. Getting this right requires the measurement-first methodology covered in Chapter 5.
Chapter 2: The F SKU Licensing Model
Fabric capacity is measured in Compute Units (CU) and purchased as F SKUs ranging from F2 (2 CU) to F2048 (2,048 CU). F SKUs are Azure resources — provisioned through the Azure portal, billed to an Azure subscription, and eligible for MACC burn. They are available in two commercial models:
Pay-as-you-go (PAYG): Billed at an hourly rate. No commitment. Can be paused at any time — billing stops immediately when paused and resumes when restarted. An F64 at PAYG costs $11.52/hour. Running full-time (720 hours/month), that's $8,294/month. Most enterprises should not use PAYG for production analytics that run 24/7 — the premium vs reserved is too high.
Reserved (1-year or 3-year): Commitment to a fixed CU amount for 1 or 3 years. In exchange: 1-year reserves give approximately 31% discount vs PAYG; 3-year reserves give approximately 47% discount. An F64 on 3-year reserved costs approximately $6.08/hour equivalent ($4,378/month at full utilisation). You can still pause reserved capacity — pausing doesn't eliminate the reserved commitment cost, but it stops additional PAYG charges if you've exceeded your reservation.
See our Microsoft Fabric licensing complete guide for the full F SKU pricing table and the complete comparison with P SKUs.
Chapter 3: F SKU vs P SKU Migration
If you are currently on Power BI Premium P SKUs, you need to understand three things: (1) F SKUs provide equivalent Power BI compute at lower cost, (2) F SKUs add Fabric workloads at no additional charge, and (3) P SKUs are being phased out by Microsoft — the migration is when, not if.
The cost comparison is straightforward: P1 at $4,995/month (EA) vs F64 3-year reserved at approximately $3,200–$3,800/month depending on negotiated EA pricing. That's a 24–36% reduction on the capacity cost alone. Add the Fabric workloads (previously requiring separate Azure services at $3,000–$12,000/month) and the total platform cost reduction is typically 40–60% for organisations that were running the full separate-service architecture.
The migration timing framework: migrate at EA renewal. Do not renew P SKU terms — use the renewal as the commercial cut-over point to F SKU reserved pricing. If your P SKU EA has more than 18 months remaining, negotiate a P-to-F migration credit or begin Fabric workloads on new Azure F SKU capacity (additive, not replacement) while retaining P SKU for existing Power BI workloads until renewal.
For the complete migration decision framework see our Microsoft Fabric vs Power BI Premium migration guide.
Chapter 4: Per-User Licensing Requirements
Fabric capacity does not eliminate per-user licensing. The rules:
- Report consumers (view-only): No per-user licence required when consuming content from an F/P capacity workspace. This is Fabric's key economic advantage over Power BI Pro-only models — unlimited consumers at no marginal per-user cost.
- Content creators (build reports, edit semantic models): Require Power BI Pro ($10/user/month) or PPU ($20/user/month). Pro is sufficient for most creator scenarios on F/P SKU capacity.
- Data engineers, warehouse developers, notebook users: Require Pro or PPU. These roles write to Fabric workloads — they need a Fabric-enabled licence.
- M365 E5 users: Power BI Pro is included in M365 E5. Audit your E5 population before purchasing standalone Pro licences.
The creator vs consumer ratio determines your per-user licensing cost. A 1,000-user analytics organisation with 80 creators and 920 consumers pays 80 × $10 = $800/month in Pro licences plus capacity cost — vs $10,000/month in all-Pro licences if there were no capacity. The break-even: capacity saves money once creators + consumers exceed approximately 120–150 users (depending on SKU size).
Chapter 5: Capacity Sizing Methodology
The Fabric Capacity Metrics App (free, from AppSource) provides per-workload CU telemetry. Install it on the first day of any Fabric deployment. Run representative workloads for 30 days before committing to reserved pricing. The measurement protocol:
- Weeks 1–2: Normal operational workload baseline
- Week 3: Stress test — simulate peak scenarios (end-of-month reports, batch pipelines, concurrent user sessions)
- Week 4: Analyse P95 utilisation and throttling events
- Decision rule: P95 < 75% = right-sized. P95 75–85% = acceptable. P95 > 85% or any interactive throttling = upsize to next SKU
The smoothing model means short bursts above capacity are absorbed. The sizing constraint is sustained demand, not instantaneous peaks. See the full methodology in our Microsoft Fabric capacity planning guide.
Chapter 6: OneLake Storage
OneLake storage is billed at $0.023/GB/month (LRS) — identical to Azure Data Lake Storage Gen2. It is charged separately from Fabric capacity. The key cost optimisation levers:
- Shortcut architecture: Create OneLake shortcuts to existing data lakes (ADLS Gen2, S3, GCS) instead of copying data. Shortcuts consume zero OneLake storage.
- Delta VACUUM: Run VACUUM weekly to remove orphaned Delta table files from updates and deletes. Without VACUUM, storage grows 2–5x over 12 months.
- Artifact retention: Set appropriate retention policies for Fabric artifacts and caches.
- Redundancy right-sizing: Use LRS unless regulatory requirements mandate GRS. LRS is half the cost of GRS.
For the complete OneLake cost model see our OneLake licensing and storage cost guide.
Chapter 7: EA Negotiation Strategy
The four negotiation levers for Fabric capacity pricing:
Lever 1 — Reserved vs EA P SKU: Always request F SKU pricing. The 3-year reserved F64 vs P1 difference ($1,200–$1,800/month) is self-evidently in your favour. Microsoft will try to sell P SKU simplicity. Push back.
Lever 2 — MACC linkage: F SKU reserved capacity purchased via Azure counts toward MACC burn. If you have Azure MACC commitments, Fabric capacity converts analytics platform cost into committed Azure spend — effectively giving you Azure EA discount rates on analytics.
Lever 3 — Competitive alternatives: Databricks (Unity Catalog + Delta Lake) is the primary alternative for Fabric's data engineering and warehousing components. Document a credible Databricks path. Microsoft's discount ceiling for documented competitive situations is typically 25–35% below list price.
Lever 4 — Bundle Power BI Pro: Negotiate Power BI Pro licences (for creators) as a bundle with Fabric capacity. Combination of 200+ Pro licences + F64 capacity often unlocks 10–15% bundle discount on the Pro side.
Download the Full Guide
Get the complete guide with pricing tables, negotiation scripts, and capacity planning worksheets.
Download Free PDF →