Power BI Premium vs Fabric: UK Capacity Costs

For a new purchase, the practical choice is no longer Power BI Premium P-SKU versus Fabric F-SKU: Microsoft has stopped selling new P-SKUs, so new capacity is an F-SKU. The real decision is whether you need capacity at all and, if you do, whether to run it pay-as-you-go or reserve it.

Here is the current UK reference point. On 22 August 2026, Microsoft's Azure Retail Prices API returned £0.158 per capacity unit per hour for Power BI capacity usage in UK South. That makes an always-on F64 about £7,382 per month pay-as-you-go, using a 730-hour month. The same API returned a one-year reservation reference price equivalent to about £4,391 per month. Both figures exclude VAT, storage, networking and any negotiated agreement terms.

The answer in one table

These are a dated planning model, not a quote. Azure says non-US-dollar Retail Prices API figures are reference estimates, and your invoice can vary with exchange rates, tax, region, contract discounts and additional services. Confirm the final price in your Azure account or with your Microsoft representative before purchasing.

UK Fabric capacity cost by F-SKU

The table below derives each figure from the same UK South API values: £0.158 per CU-hour for pay-as-you-go Power BI capacity usage and £823.2682 per CU for a one-year reservation. The monthly pay-as-you-go column assumes the capacity runs for 730 hours. The reservation column divides the full one-year compute reference by 12.

At this dated rate, the one-year reservation is about 40.5 per cent below running the same capacity pay-as-you-go for a full year. That saving is useful only for capacity you will actually consume: reserved CUs are applied hourly, unused reserved capacity does not carry forward, and the reservation covers compute rather than storage or networking. Pay-as-you-go can be better for development, seasonal loads or working-hour schedules because pausing stops usage charges.

For an exact estimate, replace the 730 hours with your intended running hours and verify your Azure region:

monthly compute = capacity units x regional CU-hour rate x running hours

The F64 break-even versus Power BI Pro

F64 matters because it is the smallest Fabric capacity that permits people with a Fabric Free licence and the Viewer role to consume Power BI content. On F2 through F32, each viewer still needs Power BI Pro or Premium Per User. Authors and editors need Pro or PPU at every F-SKU size.

Microsoft's UK list price checked on 22 August 2026 is £10.80 per Power BI Pro user per month, paid yearly, excluding VAT. A deliberately simple view-only comparison is therefore:

• One-year reserved F64: £4,391 / £10.80 = about 407 viewers.
• Always-on pay-as-you-go F64: £7,382 / £10.80 = about 684 viewers.

This is not a recommendation to buy at user 407 or 684. It is the point where the avoidable Pro list price for view-only users approximately equals F64 compute in this model. Adjust it before making a decision:

• Subtract viewers who already receive Power BI Pro through Microsoft 365 E5 or another entitlement.
• Use your contracted licence and Azure prices, not list price, if you have them.
• Keep Pro or PPU for every author, editor and workspace contributor.
• Add storage, networking, support and operational costs.
• Model realistic running hours for pay-as-you-go and realistic utilisation for a reservation.
• Confirm that F64 can sustain the actual refresh, query, data engineering and Copilot workload.

At 250 otherwise unlicensed viewers, Pro list price is £2,700 a month and remains below either F64 example. At 500, it is £5,400: above the reserved F64 compute reference but below an always-on pay-as-you-go F64. This is why a viewer count without the billing model and workload is not enough to choose capacity.

For the per-user options below this threshold, see Power BI licensing explained.

P-SKU versus F-SKU: what actually changed

The P-to-F capacity-unit mapping is direct, but it is a starting point rather than a purchasing instruction:

Microsoft recommends examining 30 to 45 days in the Fabric Capacity Metrics app before right-sizing. Moving from P1 to F64 preserves the nominal capacity-unit level; it does not prove that F64 is correctly sized once you add Fabric pipelines, warehouses, notebooks or Copilot.

What smaller F-SKUs are for

F2 through F32 are useful when you need Fabric compute but do not need capacity to remove viewer licences. Typical uses include development and test, scheduled data pipelines, small semantic models and a controlled introduction of Fabric workloads. You still budget Pro or PPU for every Power BI viewer on those tiers.

Microsoft currently supports Copilot on paid F2 and larger capacities, as well as P1 and larger, but trial and free capacities are not supported. Capacity eligibility is only the first gate: tenant settings, regional availability, data residency and capacity headroom also matter. An F2 that technically enables Copilot can still be too small for a production workload.

A Fabric trial lasts 60 days and is represented as an F64-equivalent trial capacity. Use it to measure and validate; do not treat a trial result as proof of ongoing price, availability or production performance.

How to migrate from P-SKU without an outage

P-to-F migration is manual. Microsoft says a P-SKU enters a 30-day grace period after its subscription ends, then becomes throttled from day 31 and rejects operations from day 91. Do not use that grace period as the migration plan. If you would rather have the move planned and run for you, our fixed-price Premium to Fabric migration sizes the F-SKU on your measured usage and reassigns the workspaces in waves.

Use this sequence:

1. Inventory workspaces, capacity administrators, workload settings, refresh schedules, gateways and dependent integrations.
2. Review 30 to 45 days of Capacity Metrics data and choose a provisional F-SKU.
3. Buy the F-SKU in the same region where practical and reproduce the required capacity settings.
4. Move a representative pilot group, then test permissions, refreshes, pipelines, subscriptions, apps, embedding and performance.
5. Reassign the remaining workspaces outside active refresh windows and monitor throttling and failures.
6. Retire the P-SKU only after the F-SKU workloads and viewer access are proven.

Existing Enterprise Agreement customers may be able to keep and annually renew an existing P-SKU until their agreement term ends. Microsoft no longer sells new P-SKUs, and expiring Enterprise Agreements or Microsoft Cloud Agreements cannot add or renew them. Confirm the exact contract position with Microsoft rather than inferring it from a general migration article.

How to size Fabric capacity properly

Licence break-even answers whether capacity might make commercial sense. Workload testing answers which capacity can run the service. Measure:

• Peak concurrent report queries, not just total named users.
• Semantic-model size, DAX complexity and interactive latency.
• Refresh timing and overlap with business-hour consumption.
• Pipelines, warehouses, notebooks, AI and other workloads sharing the CUs.
• Throttling, carry-forward and smoothing behaviour in Capacity Metrics.
• The effect of planned growth, failure retries and month-end peaks.

Start with evidence, set performance and cost guardrails, and scale deliberately. A cheaper capacity that throttles critical reporting is not a saving; an always-on capacity that is idle every night may be avoidable spend.

Official sources and calculation basis

This page was checked against Microsoft's primary documentation on 22 August 2026:

• Power BI UK pricing
• Azure Retail Prices API documentation
• Microsoft Fabric pricing and billing
• Microsoft Fabric licences
• P-SKU to F-SKU migration FAQ
• Fabric capacity reservations
• Copilot in Fabric FAQ

The exact planning inputs used here were GBP, Azure region uksouth, service Microsoft Fabric, SKU Power BI Capacity Usage, and the current Fabric Capacity Reservation one-year entry. Re-run the official price lookup and replace the inputs before approval; regional reference prices and exchange rates change.

Get the capacity decision in writing

A capacity purchase should end with a user-and-workload model, a dated price source, an explicit break-even, a test result and a rollback plan. Our fixed-price Trusted Numbers Review covers that decision and credits the fee against follow-on work. Book a Trusted Numbers Review, have a UK Power BI consultancy size the capacity decision with you, or review our published BI build pricing.