Microsoft Fabric launched in 2023 as Microsoft's unified analytics platform, consolidating Data Factory, Synapse Analytics, Power BI, Azure Data Explorer, and Azure Machine Learning into a single integrated experience. For enterprise buyers, Fabric introduced a new licensing architecture — one that is significantly more complex than the Power BI Premium model it partly replaces, and one that Microsoft's sales teams have become adept at monetising aggressively. This guide provides the commercial clarity that Microsoft's own documentation does not.
The Fabric Licensing Architecture: Capacities and Compute Units
Microsoft Fabric is licensed entirely through a capacity model. There are no per-user licenses for Fabric workloads — instead, you purchase compute capacity measured in Capacity Units (CUs), and all users within your Fabric-enabled tenant can consume from that shared capacity pool. This is architecturally different from most Microsoft licensing, where per-user assignment is the norm.
Fabric capacity comes in two procurement pathways: F-SKUs (the native Fabric capacity offering) and P-SKUs (Power BI Premium Per Capacity, now rebranded as Fabric with backward compatibility). F-SKUs are the current Microsoft positioning and offer more flexibility; P-SKUs are legacy but remain commercially relevant for organisations already running Power BI Premium commitments.
F-SKU Capacity Tiers and Pricing
The F-SKU range spans from F2 (2 CUs, approximately $263/month) up to F2048 (2,048 CUs, approximately $269,000/month). The commercially relevant range for most enterprise buyers sits between F32 and F256, corresponding to roughly $4,200 to $33,800 per month. These are list prices — actual EA negotiated rates are typically 20-40% below list, depending on your Azure consumption commitment and Microsoft commercial relationship.
| SKU | Capacity Units | List Price/Month | Comparable P-SKU | Key Use Case |
|---|---|---|---|---|
| F2 | 2 CUs | ~$263 | None (dev/test) | Development and experimentation |
| F4 | 4 CUs | ~$526 | None | Small team analytics |
| F8 | 8 CUs | ~$1,052 | None | Department-level workloads |
| F16 | 16 CUs | ~$2,100 | ~P1 | Mid-market analytics |
| F32 | 32 CUs | ~$4,200 | ~P1+ | Enterprise reporting tier |
| F64 | 64 CUs | ~$8,400 | ~P2 | Enterprise analytics + Lakehouses |
| F128 | 128 CUs | ~$16,800 | ~P3 | Large-scale data platform |
| F256 | 256 CUs | ~$33,800 | ~P4 | Enterprise data platform + AI |
Insider Perspective: The F2 and F4 SKUs are Microsoft's foot-in-the-door strategy. We routinely see enterprises enter Fabric conversations with "just an F4 to experiment" and find themselves at F64 within 18 months as workload adoption grows. The architecture makes consolidating workloads onto Fabric easy — and the budget implications of that consolidation are rarely modelled upfront. Always project your likely capacity requirements 24 months forward before committing.
F-SKU vs P-SKU: The Migration Decision
If your organisation runs Power BI Premium Per Capacity (P-SKUs), Microsoft's positioning is that you should migrate to F-SKUs. The pitch is compelling: F-SKUs unlock all Fabric workloads (Lakehouses, Warehouses, Data Factory, Real-Time Analytics) at the same price point as legacy P-SKU capacity. P-SKUs, by contrast, only support Power BI workloads within the new Fabric model.
The commercial reality is more nuanced. P-SKUs negotiated prior to 2024 typically carry enterprise pricing that was benchmarked against the full Power BI Premium feature set — and many organisations secured significant discounts on those commitments. Before migrating to F-SKUs, you must verify that your enterprise pricing transfers correctly, understand the impact on any existing Embedded Analytics commitments, and assess whether the additional Fabric workloads actually represent value you will activate within your planning horizon.
Microsoft will use the F-SKU migration conversation as an opportunity to reset pricing relationships. Organisations that migrate without independent advisory support frequently find that their effective per-CU cost increases during the migration, even when headline SKU pricing appears comparable.
Per-User Licensing: Fabric Free vs Paid
While capacity is the primary Fabric licensing mechanism, there is a user-level dimension that creates compliance complexity. Fabric Free licenses allow users to create content in My Workspace (individual sandboxes) but cannot publish to shared workspaces or consume Premium or Fabric-capacity content. For shared workspace access and full collaboration, users need either a Power BI Pro license ($10/user/month) or Power BI Premium Per User ($20/user/month).
The important nuance is that Fabric capacity workloads — Lakehouses, Warehouses, Data Factory pipelines, Real-Time Analytics — do not require per-user licenses for consumers. Users consuming reports or dashboards backed by Fabric capacity do not need Pro licenses. However, content authors creating and publishing Power BI reports to Fabric workspaces do require Pro or PPU licenses, even when the underlying data sits in a Fabric Lakehouse.
This distinction creates a common compliance risk: organisations deploying Fabric find that their content creator population — typically a broader group than initially anticipated — requires Pro licenses that were not budgeted for, while the consumer population does not. The authoring-versus-consuming distinction is poorly communicated by Microsoft in early sales engagements.
OneLake Costs: The Hidden Consumption Layer
OneLake is Fabric's unified data lake layer — a single logical storage system that underpins all Fabric workloads. OneLake storage is separate from Fabric capacity and is billed at Azure Data Lake Storage Gen2 rates through your Azure subscription: approximately $0.023 per GB/month for hot storage, with tiered pricing for cool and archive tiers.
For most analytics workloads, OneLake storage costs are modest relative to capacity costs. The risk area is egress: moving data out of OneLake for non-Fabric consumption, particularly to external systems or cross-region transfers, incurs Azure egress charges that can accumulate significantly for large data platform deployments. Organisations with hybrid architecture — some workloads in Fabric, others in non-Microsoft platforms — need to model data movement costs carefully.
Advisory Insight: Independent advisory firms including Redress Compliance have found that enterprises routinely underestimate Fabric's total cost of ownership by 35–50% when they model only the headline capacity SKU cost. OneLake egress, Pro license requirements for authors, and compute burst scenarios from underprovisioned SKUs are the three categories that generate the most budget surprises in Fabric deployments.
Capacity Burst and Smoothing: The CU Economics
Fabric capacity operates on a burst-and-smooth model. Workloads can consume more CUs than your committed capacity for short periods — Microsoft refers to this as "bursting." However, sustained over-utilisation leads to throttling: queries queue, reports slow, and pipeline jobs delay. Microsoft's Capacity Metrics app provides visibility into this utilisation, but interpreting the data and right-sizing capacity requires commercial and technical judgement that is distinct from the data analysis skills most analytics teams possess.
The practical consequence is that right-sizing Fabric capacity is genuinely difficult at initial procurement. Workloads that appear light in testing can generate significant burst activity in production — particularly Power BI semantic models with complex DAX, large-scale data pipelines, and Spark-based Notebook workloads running in Fabric Data Engineering. The guidance from Microsoft is to start with a smaller SKU and scale up, but this assumes frictionless contractual upsize — which is achievable in Azure-billed Fabric but less clean for Fabric purchased through EA with fixed-term commitments.
Buying Fabric Through EA vs Azure Marketplace vs Direct Azure
Microsoft Fabric can be purchased through three channels: as a Microsoft Azure service billed directly to your Azure subscription, as a component of a Microsoft Azure Committed Use (MACC) spend commitment, or increasingly as a line item within Enterprise Agreement modifications. Each channel has different pricing, contractual, and compliance implications.
Azure-billed Fabric consumption counts toward your Azure MACC commitment, which is commercially significant for organisations managing large Azure spend commitments — particularly if you are approaching the end of a MACC period with under-consumed capacity. EA-negotiated Fabric pricing provides better unit economics but less flexibility to scale down; Azure Marketplace consumption provides more flexibility but typically at higher unit prices. The optimal channel depends on your existing Microsoft commercial relationship, your Azure MACC structure, and your anticipated Fabric growth trajectory.
For a comprehensive view of how Fabric fits into your Microsoft EA framework, see our Complete Microsoft EA Guide. For organisations with significant Azure consumption, our Azure EA Negotiation Guide covers how Fabric commitments interact with broader Azure commercial structures. Our Cloud Contract Negotiation practice covers Fabric capacity agreements as part of full Microsoft commercial advisory engagements.
Fabric Licensing for AI and Copilot Workloads
Microsoft's AI positioning for Fabric centres on Copilot features embedded within the platform — natural language query, code generation for notebooks, automated data insights. Copilot capabilities in Fabric require a minimum F64 capacity SKU, which Microsoft frames as an incentive to scale capacity. The commercial reality is that this creates a meaningful price cliff: the difference between an F32 and F64 is a doubling of capacity cost, and many enterprises find that Copilot's practical value in Fabric — versus its demo value — does not justify the capacity uplift at current maturity levels.
As Microsoft integrates Azure OpenAI capabilities more deeply into Fabric, expect additional consumption-based charges for AI-generated features. The model where Fabric AI capabilities are "included" in capacity pricing is a current market-development position that Microsoft will monetise more aggressively as usage scales. Building flexibility to exit or renegotiate AI feature pricing into current Fabric contracts is an underappreciated commercial protection.
Practical Recommendations Before Buying Fabric
Before committing to Fabric capacity, conduct a structured assessment that covers four dimensions: workload fit (which of your analytics workloads genuinely benefit from Fabric's unified architecture versus your existing tooling), user population analysis (how many authors require Pro licenses, how many consumers can be served from shared capacity), capacity modelling (projected CU consumption for each workload type at production scale, with headroom for burst), and commercial structure (whether EA, MACC, or pay-as-you-go best fits your governance and flexibility requirements).
Microsoft's Fabric sales motion tends to lead with the platform vision — unified analytics, OneLake simplicity, Copilot — rather than the commercial complexity. Engaging an independent advisor with experience in both Microsoft licensing and data platform architecture ensures that the commercial case is grounded in your actual workload economics rather than Microsoft's reference architectures. Access our Microsoft EA White Paper for a broader Microsoft negotiation framework that contextualises Fabric within your overall Microsoft spend portfolio.