Power BI's commercial model rewards mistakes. The tier ladder — Pro, Premium Per User, Premium Capacity, and now Fabric — is structured so that almost every enterprise drifts into a more expensive tier than the workload requires. Microsoft sales is incentivised to push the drift; renewal cycles compound it. The right-sizing conversation is worth $50–500 per user per year, multiplied across every analyst seat in the business.
Power BI is sold in four overlapping tiers. The differences are functional, not just commercial, and the boundaries are where most over-spend originates. Pro is the entry licence for collaboration. Premium Per User adds the larger model sizes, AI features and refresh frequency. Premium Capacity moves the spend from per-user to per-capacity. Fabric, introduced in 2023 and aggressively repositioned through 2025, rolls Power BI Premium into a broader analytics platform sold by capacity unit.
Pro is the licence most users actually need. It allows authoring, sharing inside a tenant, and consumption of Pro-published content. The model limits — 1GB dataset, 8 refreshes per day — bite at the analytical edges but rarely at the operational core. Pro is included in Microsoft 365 E5 and sold standalone at roughly $14 per user per month.
PPU is positioned as a midpoint between Pro and Premium Capacity. It unlocks 100GB models, 48 refreshes per day, AI features and XMLA endpoint write. The catch: every viewer of PPU-published content also needs a PPU licence. The tier is correctly sized for analyst teams; it is structurally over-priced for general consumption.
Premium Capacity moves spend from per-user to per-capacity. Viewers do not need an individual Pro licence, which inverts the economics above roughly 250 viewers. P-SKUs are sold in P1, P2, P3 and P4 sizes, with corresponding v-cores. The historic break-even is around 250 viewers; modern Fabric pricing has shifted it slightly upward.
Fabric is the umbrella SKU positioned to absorb Premium Capacity over time. Fabric F-SKUs are sold by capacity unit with pause/resume billing, and bundle Power BI Premium with Synapse, Data Factory, OneLake and Real-Time Intelligence. The migration from P-SKU to F-SKU is being driven from Redmond and timing is increasingly non-optional.
The Fabric overlay changes the right-sizing calculation. The buyers who get the best outcomes model the alternatives before Microsoft does.
PPU at $20 per user per month is roughly 1.4x Pro. The licence is correctly sized for an analyst — someone who authors PPU-grade content. It is wildly over-sized for a consumer. We routinely see PPU footprints where 80% of the licences are viewers who could be served by Pro or by Premium Capacity. The fix is workload-based licensing: who authors PPU content vs. who consumes it.
P-SKU capacity is sticky. A team that started on P1 four years ago has often grown to P2 or P3 without questioning whether the original P1 was correctly sized to begin with. The fix is a capacity utilisation review — measure CPU, memory and refresh queue utilisation over a representative 30-day window. We have moved capacity down a notch in 40%+ of the engagements where we have run this review.
Microsoft's Fabric pitch encourages buying a full F64 capacity to replace a P1 — F64 is roughly Fabric's stated equivalent of P1 — at a higher list price but with the workload diversification. The migration math favours customers who model Fabric capacity against the actual Power BI workload alone, not against the bundled future Fabric workloads Microsoft expects to materialise.
Includes the Power BI right-sizing worksheet, the Fabric migration model and the EA Power BI clause library.
Power BI sits inside the Microsoft Enterprise Agreement (EA) or, increasingly, the Microsoft Customer Agreement for Enterprise (MCA-E). The renewal cycle is the only point at which capacity can be repriced, licences right-sized and tier mix renegotiated without uplift carrying forward. Three workstreams matter.
We benchmark Power BI commercials against EA peer data across 340+ Microsoft engagements.
Microsoft has signalled that Premium P-SKUs will eventually be deprecated in favour of Fabric F-SKUs. The exact timeline has shifted multiple times, but the direction is set. Customers who migrate too early pay the early-Fabric premium with no rebate; customers who wait too long find themselves with no commercial negotiation room. The right window is typically 12 months ahead of the customer's EA renewal — late enough to have Fabric pricing visibility, early enough to negotiate. Folding the Fabric move into a broader Microsoft EA optimization exercise, rather than running it as an isolated SKU swap, is what preserves that negotiating room.
Fabric capacities support pause/resume billing, which materially changes the cost of low-utilisation environments — dev, test, infrequent reporting workloads. The capacity model also allows fractional capacities, so over-provisioning is less binary than under P-SKUs. The downside: the licence costs for Fabric F-SKUs at equivalent compute are higher than P-SKUs, and the discount track is less mature.
Fabric bundles in components that customers may already have purchased separately — Synapse, Data Factory, OneLake. The migration decision is not a like-for-like commercial swap; it is a bundle decision that should factor in displaced spend in Azure data services. The customers who model both sides of this — Power BI cost and Azure displacement — get a substantially clearer picture than those who model Power BI alone.
Our Microsoft practice rightsizes Power BI estates for a living. Independent, buyer-side, no Microsoft partner status.
Weekly compliance intelligence for IT leaders.