Microsoft Fabric costs $0.18 per capacity-unit hour in US East at list prices accessed 18 November 2026, making F2 about $263 and F64 about $8,410 monthly. A one-year reservation is $938 per CU yearly. Price storage and the viewer licences required below F64 alongside capacity before deciding whether to reserve it.
Finance needs a monthly estimate, while Microsoft's price page starts with a region, a currency and a capacity size. An F SKU gives you the compute line. To finish it, count the report builders and readers, estimate stored data and set the hours the capacity will remain active.
The figures below use US East list prices in US dollars, accessed 18 November 2026. Monthly capacity calculations use Microsoft's 730-hour month. Apply your own active hours when the capacity will run for less than a full month.
Capacity units set the Fabric compute bill
A Fabric capacity serves the workloads assigned to it, with its size expressed in capacity units, or CUs. The pipeline loading records, the notebook preparing them and the report querying them all draw on that compute. Their combined demand determines the size the application needs.
The F SKU names its capacity size. Moving to a larger SKU gives those workloads more capacity, but the useful size depends on when the jobs run together. If finance refreshes the data while people query the report, both jobs use capacity together. Test that overlap as well as a schedule that separates them.
Pay-as-you-go F capacities bill by the second without a reservation commitment. While a capacity remains active, its compute charge continues. Pausing it stops ongoing provisioned compute billing and makes its content unavailable until it resumes. Microsoft also bills remaining smoothed usage and accumulated overages when a capacity is paused.
Fabric spreads some compute usage over time through smoothing. That helps absorb bursts, but sustained demand can still lead to throttling. Capacity overage provides a paid route for excess consumption, while surge protection lets administrators set limits on background or workspace activity. Decide which jobs can wait and which should consume paid overage before a busy reporting period reaches those limits.
Each larger F SKU doubles the price
Moving up the capacity ladder doubles both the CUs and the base compute price at each step shown below. The purchasing model stays the same within each column. Storage and licences remain separate lines even when the capacity you buy can handle more work.
For pay-as-you-go, multiply CUs by $0.18 and the hours the capacity will run. At 730 hours, an F32 costs $4,204.80. For a one-year reservation, multiply CUs by the annual $938 rate and divide by 12 to express the commitment as a monthly equivalent. The F32 equivalent is $2,501.33.
The reservation column spreads an annual commitment across 12 months. Use it for a capacity you expect to keep running, and compare pay-as-you-go against the actual active hours of a development capacity. Finance can then see which schedule supports the purchase decision.
These monthly amounts are calculated from the Microsoft Azure Retail Prices API, accessed 18 November 2026, for US East and a 730-hour month. Microsoft's Fabric pricing page lists larger capacities through F8192.
| SKU | Capacity units | Pay-as-you-go a month | 1-year reservation a month |
|---|---|---|---|
| F2 | 2 | $262.80 | $156.33 |
| F4 | 4 | $525.60 | $312.67 |
| F8 | 8 | $1,051.20 | $625.33 |
| F16 | 16 | $2,102.40 | $1,250.67 |
| F32 | 32 | $4,204.80 | $2,501.33 |
| F64 | 64 | $8,409.60 | $5,002.67 |
| F128 | 128 | $16,819.20 | $10,005.33 |
| F256 | 256 | $33,638.40 | $20,010.67 |
Power BI licences change at F64
Separate report builders from readers in the licence estimate. Their Power BI rights depend on their role and the capacity serving the report. Below F64, each Power BI viewer needs Pro, Premium Per User or an individual trial licence. At F64 or larger, users with a free licence and the viewer role can read Power BI content.
Creators still need the appropriate Power BI licence to create and share items outside My workspace. Microsoft's Power BI price list, accessed 18 November 2026, lists Pro at $14.00 and Premium Per User at $24.00 per user per month, paid yearly.
Check the licences your organisation already holds before adding new ones to the estimate. A report reader who already has Pro affects the incremental purchase differently from a reader who needs a new licence. List the licences each group needs, then subtract the ones already covered by existing agreements when estimating the new purchase.
Look closely at F64 when most of the report audience only reads. It can change which capacity produces the lower combined bill, even if a smaller capacity could handle the queries. It does not settle sizing by itself: the selected capacity must also support the refreshes and report activity you expect.
Storage and overage are billed on top
Add stored data and any excess compute to the base capacity estimate. The CUs cover compute. Records kept in OneLake, cache or SQL storage appear under the relevant storage category.
The Azure Retail Prices API lists OneLake hot storage at $0.026, cool at $0.019 and cold at $0.004 per GB per month. It lists OneLake cache at $0.20 and SQL storage at $0.25 per GB per month. These are US East rates accessed 18 November 2026. Split the application's stored records by category before multiplying GB by the monthly rate.
Microsoft includes mirroring replica storage up to the purchased SKU's size in TB. An F64 includes up to 64 TB for mirrored replicas. Ordinary files use separately billed storage. The mirroring allowance also changes when the provisioned capacity is paused, so a paused development environment needs its own storage estimate.
Capacity overage is $0.54 per CU hour in the same US East price list. Microsoft also offers zero-provisioned F0 with on-demand billing at regional pay-as-you-go rates. Reservation discounts do not apply to overage or on-demand usage. Show on-demand usage and overage as separate lines. Administrators can set limits on additional consumption, which helps define what the estimate is intended to cover.
Reservations and pausing lower the Fabric bill
Reserve capacity that you expect to use consistently, and consider pausing a pay-as-you-go development capacity when nobody needs it. An always-available production report and a development notebook can justify different purchasing choices even at the same SKU size.
Using the same US East rates, an F64 reservation has a monthly equivalent of $5,002.67, about $5,003. A continuously active pay-as-you-go F64 is $8,409.60, about $8,410. Both figures come from the Azure Retail Prices API accessed 18 November 2026. The reservation exchanges an ongoing commitment for a different compute rate.
A development capacity may have a different schedule from the production reports. Pausing can reduce its active compute hours, but stored data remains part of the estimate. Account for the work that must finish before the pause and the users who need the capacity available after it resumes.
Use Microsoft's Fabric capacity estimator for an initial sizing exercise, then inspect actual consumption in the Capacity Metrics app. Microsoft's trial lasts 60 days and can be configured as F4 or F64. Check the assigned size before interpreting trial results. Watch the jobs run together during a reporting period. A single successful refresh will not show the demand when several users open the reports while data is still being prepared.
The capacity is only part of the cost
Someone still needs to connect the source records, define the measures and build the reports the capacity will serve. Include that implementation and continuing support in the budget so the purchased capacity reaches the intended business use.
As our data platform cost guide puts it: “The licences are only part of the bill. When two engineers spend most of the week fixing pipelines, their time is usually the larger cost.” A useful estimate follows a report through its source connection, preparation and use, identifying who maintains each part.
Databasin can deliver data into Fabric and run over the environment your team keeps. We implement the source connections, prepared measures and applications together. Saint Louis University used Databasin in its move to Fabric and retired spending on Banner, Cognos, Alteryx and Tableau.
The budget should show finance what people will be able to use after the capacity is purchased. Put the Microsoft services and licences beside the implementation scope, with the source connections, reports and continuing support that make the application work.
Example: choosing F32 or F64 for 250 report readers
This illustrative bill covers 20 report builders, 230 readers and 2 TB of OneLake hot storage at US East list prices accessed 18 November 2026.
| Line | On F32 | On F64 |
|---|---|---|
| Capacity pay-as-you-go | $4,204.80 | $8,409.60 |
| Power BI Pro for builders | 20 × $14 = $280 | 20 × $14 = $280 |
| Power BI Pro for readers | 230 × $14 = $3,220 | $0 |
| OneLake hot storage, 2,048 GB | $53.25 | $53.25 |
| Monthly total pay-as-you-go | $7,758.05 | $8,742.85 |
| Monthly total with a 1-year reservation | $6,054.58 | $5,335.92 |
Under pay-as-you-go, this organisation spends less on F32. The one-year reservation reverses the comparison: F64 has the lower total because its readers can use free licences with the viewer role.
Five checks before you buy capacity
- Count builders and readersSeparate people who create reports from those who only view them, then check the Power BI licences already assigned.
- Pick your region and currencyUse one region and currency across capacity and storage estimates so every line describes the same proposed deployment.
- Run a trial on real workloadsTest source refreshes alongside report use, and inspect the trial's assigned capacity before using its consumption to size production.
- Decide on a reservation after a monthReview a month of active hours and report demand, including overlapping refreshes, before deciding how much capacity to reserve.
- Add the build and support workInclude the source connections, measure definitions, applications and continuing support needed to put the purchased capacity into business use.
Sources
Microsoft, Azure Retail Prices API, 2026 · prices.azure.com/api/retail/prices
Microsoft, Microsoft Fabric pricing, 2026 · azure.microsoft.com/en-us/pricing/details/microsoft-fabric
Microsoft Learn, Understand Microsoft Fabric licenses and capacity, 2026 · learn.microsoft.com/en-us/fabric/enterprise/licenses
Microsoft, Power BI pricing, 2026 · microsoft.com/en-us/power-platform/products/power-bi/pricing
Microsoft Learn, Pause and resume your Fabric capacity, 2026 · learn.microsoft.com/en-us/fabric/enterprise/pause-resume
Microsoft Learn, Fabric trial capacity, 2026 · learn.microsoft.com/en-us/fabric/fundamentals/fabric-trial
Plan the work on top of Fabric
We can discuss the reports and sources your Fabric capacity needs to serve and what the first use case would involve. A demo explores the application, the data behind it and how our team would support its use.