Power BI Licensing Explained (2026): Free vs Pro vs PPU vs Fabric Capacity
- Matt Lazarus
- 7 days ago
- 5 min read
Power BI licensing generates two reliable emotions: confusion at purchase and surprise at renewal. The tiers have changed names, Premium has been retired in favour of Fabric capacity, and entitlements hide inside Microsoft 365 bundles that nobody audits.
The result is predictable: businesses paying for Pro seats that only ever open reports, or carrying capacity sized for an organisation twice their size. The licensing model is actually coherent once you see the structure - it prices two things, authoring and hosting, and everything else follows.
This guide explains the structure rather than quoting prices that date, so the logic keeps working at your next renewal.
Key Takeaways
Licensing prices two things: who can build and share (per-user) and where content is hosted (workspace tier or capacity).
The builder-to-viewer ratio decides the model - per-user economics flip to capacity economics as audiences grow.
Most overspend is structural: Pro seats for viewers, unaudited E5 entitlements, and capacity bought before it was measured.
What Does Each Licence Tier Actually Do?
Free authors for yourself; Pro authors and shares; Premium Per User (PPU) adds the heavy features to a Pro seat; and Fabric capacity moves the cost from people to infrastructure, letting free-licence users consume content hosted on it. Every tier is a different answer to two questions: can this person publish, and where does the content live?
The mechanics that matter:
Free builds full reports in Desktop but cannot share through the service - fine for personal analysis, a dead end for collaboration.
Pro is the workhorse: publish to shared workspaces, share with other Pro users, refresh on schedule. Both author and audience need Pro - the detail that drives most costs.
PPU gives individual users the big-model, frequent-refresh, paginated-report features without buying organisation-wide capacity - the right bridge for small teams with heavy needs.
Fabric capacity (F-SKUs) hosts workspaces on dedicated compute; viewers then consume with free licences (at F64 and above), and the same capacity runs pipelines, warehousing and AI workloads.
Where Do Licences Hide Inside Microsoft 365?
Microsoft 365 E5 includes Power BI Pro - and in most organisations running E5, a meaningful share of separately purchased Pro seats duplicates entitlements already paid for. The licensing audit that takes an afternoon routinely recovers thousands per year.
The audit is unglamorous: export the licence assignments, match Power BI Pro purchases against E5 holders, and reclaim the duplicates. While there, check the inverse error - report consumers holding no licence at all, "sharing" via exported PDFs because nobody assigned them the Pro seat that would let them use the actual reports. Both errors are invisible until someone looks.
When Does Capacity Beat Per-User Licensing?
When the audience grows past the point where viewer seats cost more than shared infrastructure - the crossover arrives somewhere in the low hundreds of users, earlier if you also need the capacity-only features. The builder-to-viewer ratio is the whole calculation: builders always need Pro; the question is how you pay for viewers.
Three worked profiles make the maths concrete:
Organisation | Profile | Right-sized answer |
~20 staff, 4 builders | Everyone collaborates on reports | Pro seats throughout; PPU only if model sizes demand it |
~200 staff, 10 builders, 150 viewers | Broad consumption, modest building | The crossover zone: price viewer Pro seats against an entry F-SKU - measured refresh and model loads decide it |
~2,000 staff, 30 builders | Enterprise distribution | Fabric capacity (F64+) with free-licence viewers; Pro for builders only |
Capacity also changes behaviour, not just cost: it is finite, so inefficient models and refresh storms now crowd out other workloads. Right-sizing is therefore an engineering exercise as much as a procurement one - the reason capacity decisions are best made alongside a performance optimisation pass that shrinks what you are about to pay to host.
What Changed With the Retirement of Premium?
Power BI Premium P-SKUs have been retired in favour of Fabric F-SKUs - functionally similar hosting, but purchased as Fabric capacity and opening the wider workload family. For organisations that ran Premium, renewal is now a migration event with a date attached, not a rollover.
The transition deserves deliberation rather than like-for-like replacement: F-SKU sizing granularity is different, pay-as-you-go and reservation options change the cost profile, and the capacity now serves more than reporting. Treating the renewal as the first stage of a planned platform move - rather than an emergency repurchase - is the difference between a budget surprise and a strategy. Sizing on measured workload, with a quarter of observation before committing to reservations, is the discipline that protects the spend.
How Do You Keep Licensing Right-Sized Over Time?
Annual hygiene, three checks: reconcile entitlements (E5 versus purchased Pro), re-run the viewer maths against current headcount, and review capacity utilisation against its meter. Usage shifts, teams grow and shrink, and last year's correct answer quietly stops being correct.
The capacity meter deserves particular attention - sustained utilisation under a third suggests a smaller SKU or consolidation; regular throttling suggests either growth or, more often, a handful of inefficient models taxing everyone. Right-sizing recovers real money: reviews of mid-market tenants routinely return 20 to 40 per cent of Power BI spend, simply because nobody had looked since the original purchase. It is a standing agenda item in any serious Power BI managed services arrangement for exactly that reason.
What Does a Licensing Health Check Look Like?
An afternoon of evidence-gathering, repeated twice a year. Pull the admin portal's usage data and ask four questions: who holds a Pro licence but has not published or consumed anything in ninety days; which viewers open reports often enough to justify their tier; what percentage of paid capacity is actually consumed at peak; and which workspaces could consolidate.
The same patterns appear in almost every audit. Licences assigned to leavers and role-changers that nobody reclaimed. Whole departments licensed at Pro because one person once needed to publish. Viewers paying per-user rates in an organisation whose audience size crossed the capacity threshold a year ago. Each is invisible in the monthly invoice and obvious in the usage data.
The structural question matters more than the housekeeping: as viewer counts grow, per-user licensing quietly becomes the expensive option, and the crossover arrives earlier than most finance teams expect. Run the capacity-versus-per-user comparison annually with current headcounts - the answer changes as you grow, and the organisations that never re-run it are the ones funding Microsoft's margin.
Assign the health check an owner - usually whoever administers the tenant - and put both dates in the calendar now. Licensing drift is not a failure of diligence; it is the default behaviour of any subscription estate that nobody is paid to question. Thirty minutes twice a year is the entire cost of staying right-sized.
Buy the Structure, Not the SKU
Licensing questions feel like procurement but answer like architecture: who builds, who views, how big are the models, and where should hosting live as the audience grows. Get those four answers right and the SKU names take care of themselves - this year and at every renewal after.
The structural rules travel well: builders need Pro, viewers are a ratio decision, entitlements hide in E5, and capacity is bought on measurement. Everything else is a price list, and price lists change.
