Power BI and Fabric Pricing: Reference

Summary page: Pricing. Per-user prices from Microsoft's US and Canadian pricing pages. Capacity prices from the Azure Retail Prices API for US East and Canada Central. Companion pages: Licensing reference, Decision Tree reference.

This page holds every monetary figure and calculation. Feature eligibility is in the licensing reference, and the decision tree reference links here for every cost comparison so the numbers live in one place.

Assumptions for every figure on this page. Prices are Microsoft list as of 22 September 2026 and change without notice. Per-user prices are Microsoft list. Capacity prices are Azure retail list for the stated region and currency, assume 730 hours per month always-on, and exclude storage, networking, the gateway machine, support, and administration. Prices through a reseller differ from list, and we confirm current pricing before any quote. Every comparison assumes the alternatives being compared meet the same feature, model-memory, and workload requirements.

Per-user licences

Pro rose from USD 10 and PPU from USD 20 on April 1, 2025. No change since.

Annual cost by user count, list price, no existing entitlements

Microsoft Fabric capacity (F SKUs)

Billed through Azure per second with a one-minute minimum. A paused pay-as-you-go capacity does not bill compute. A reservation bills for its full term whether or not the capacity is running.

Base rates

Region and currency are separate effects. Azure prices each region in USD, then converts to CAD at a Microsoft-set rate.

Canada Central is about 11% more than US East in the same currency. CAD is about 1.386× USD for the same region. Combined, a Canada Central buyer paying CAD pays about 1.54× the US East USD figure most articles quote. The reservation discount is 40.5% against pay-as-you-go in every cell checked. At these rates and regions the 3-year reservation is exactly three times the 1-year price, so it fixes the price for three years without lowering it.

The tables below use US East in USD and Canada Central in CAD, the two cases relevant to a Canadian buyer.

Monthly figures are computed from the unrounded rates and rounded for display.

• F64 is the licensing threshold for Free viewers. Below F64 every viewer needs Pro or PPU. The following are available from F2: 48 scheduled refreshes per model, uncapped API refreshes, models above 1 GB up to the SKU's memory ceiling, XMLA read/write, deployment pipelines, all Fabric workloads, and Copilot. Report Server rights need an F64 or larger reservation.
• Pay-as-you-go can be paused. An F capacity running 10 hours a day, 5 days a week bills about 30% of the always-on PAYG figure. Sizing down below F64 during off-hours removes Free viewing during those hours.
• Reservations cannot be paused. They are bought in 1 CU increments, cover compute only, and bill for the term. Unused reserved hours are lost, so a reservation should not be combined with a pausing plan.
• Power BI Premium-style Autoscale does not exist on F SKUs. Resize manually or by script; capacity overage protection handles bursts. Spark workloads have a separate autoscale billing option that is unrelated.
• OneLake storage bills separately. Hot tier about USD 0.026 per GB per month in US East and CAD 0.0347 in Canada Central. Estimate it whenever you will land data in Fabric; it is small relative to compute for typical small-business volumes and material for large lakehouses.

Power BI Embedded (A SKUs)

Billed hourly through Azure. Can be paused, as can any pay-as-you-go F SKU. Microsoft directs new embedding builds to F SKUs, which include the same embedding rights plus Fabric workloads and per-second billing. Listed here for existing deployments.

A pay-as-you-go capacity of any type running 40 hours a week (about 173 hours a month) bills about 24% of the always-on figure. A1 on that schedule is about USD 175 or CAD 242 per month; F2 on that schedule is about USD 62 or CAD 96.

Power BI Premium per capacity (P SKUs)

Not purchasable. Listed only for organizations that already hold P capacity on an Enterprise Agreement.

Use the equivalent F SKU reserved price as the comparison baseline for an organization that holds P capacity. P1 maps to F64 by compute; F64 reserved in US East is USD 5,003. Your actual cost depends on region, currency, whether they reserve, whether they right-size after measuring with the Capacity Metrics app, and whether they split development onto a smaller pausable capacity.

Cost comparisons

To compare options, count people once, subtract existing licences, and price only the options that meet the requirements.

Formulas

Pro for everyone       = Pro price × (people without E5, Pro, or PPU already)

                Capacity + free view   = capacity monthly price
                                         + Pro or PPU price × (publishers without an existing licence)
                                         [viewers free only on F64+ or P]

                PPU for everyone       = PPU price × (people touching the PPU content who hold no licence)
                                         + PPU add-on price × (people touching it who already hold Pro or E5)
                                         + 0 × (people who already hold PPU)

                Embedded, app owns data = capacity hourly rate × active hours
                                         + Pro or PPU × human publishers
                                         + application costs
                

Compare full cost with full cost, or incremental cost with incremental cost.

Break-even points at list price

These are the seat counts at which one alternative's monthly cost equals another's. They describe cost only; whether a capacity can run the workload is a separate check. F64 has fixed compute, and the alternative must also satisfy model memory, refresh, and concurrency requirements. Model memory means measured peak, including refresh and queries, not the resident size: Microsoft states that refresh can more than double a model's footprint while it runs. PPU's ceiling is 100 GB; F64's is 25 GB.

Pro viewers versus F64 with Free viewers. Publishers hold Pro in both cases, so their cost cancels.

PPU for everyone versus F64 with Pro publishers. The author count matters because it sits on the capacity side.

Microsoft's own guidance that PPU is cost effective below about 250 users lines up with these figures.

What changes the break-even point. Existing E5 or Pro entitlements shrink the per-user side and push the break-even point higher. A capacity that is needed anyway, for a large model, for Copilot, or for Fabric workloads, has its cost partly justified before viewers are counted, which pushes the break-even point lower. Pausing lowers the PAYG capacity cost but not a reservation, and pausing or downsizing below F64 removes Free viewing during that time.

Example: 5 publishers and 350 viewers, standard reporting, no existing licences

Pro is cheaper in both currencies, so a headcount of 355 does not justify F64.

Example: 12 publishers and 600 viewers, imported models up to 20 GB resident, no existing licences

Pro alone is not eligible: the models exceed 1 GB. The eligible alternatives are PPU for everyone, or a capacity whose memory holds the measured peak.

Evaluate F64 only if peak memory and load testing qualify it. Otherwise F128 is the next candidate, subject to the same validation, and it still beats PPU for this audience if it qualifies.

Example: 10 publishers and 35 consumers, one 4 GB resident model, 30-minute refresh, XMLA

PPU is cheaper than either capacity alternative and meets the requirements. F16 is listed to show why the smallest SKU whose ceiling exceeds the resident size is not automatically eligible.

Example: 5 publishers and 80 viewers who all need the same PPU-hosted content

The F2 row is cheapest if the largest model's measured peak fits in 3 GB. A 4 GB model rules it out. For a model with a peak between 25 GB and 100 GB, the eligible options are PPU (100 GB), F128 (50 GB), or F256 (100 GB); F64 is not.

Worked examples, list price

All figures pre-tax, at Microsoft list. Thirteen fuller situations with reasoning are in Worked Examples.

Sources

• Power BI pricing, United States
• Power BI pricing, Canada
• Azure Retail Prices API, read 22 September 2026. Reproducible queries:
  • US East Fabric, USD
  • US East Fabric, CAD
  • Canada Central Fabric, USD
  • Canada Central Fabric, CAD
  • US East Embedded, USD
  • Canada Central Embedded, CAD
• Microsoft Fabric pricing
• Fabric capacity reservations
• Pause and resume a Fabric capacity
• Power BI pricing change, April 1 2025