Reporting & analytics articles

Microsoft Fabric licensing cost for a small business: capacity units, F SKUs and the seats on top

By Edward Pinela — Microsoft-certified consultant · September 12, 2026

Microsoft Fabric licensing cost is two bills, not one: a capacity rented from Azure by the capacity-unit hour, plus the per-user Power BI seats you may still have to buy on top of it. Checked against Microsoft’s live rate card on 12 September 2026, Fabric capacity in East US runs at $0.18 per capacity-unit hour — $262.80 a month for the smallest F2 — or $938 per capacity unit on a one-year reservation, which is $156.33 a month for that same F2. Below F64, everyone who opens a report still needs a $14 Pro seat. That threshold, not the capacity rate, is what decides most small-business bills.

Two meters and a seat count

Fabric bills three things, and conflating them is where budgets go wrong. The first is capacity: a pool of compute measured in capacity units (CUs) that every Fabric workload draws from — Power BI queries, semantic model refreshes, Spark notebooks, warehouse queries, pipelines. The second is OneLake storage, charged per gigabyte per month and entirely separate from capacity. The third is per-user licenses, which Fabric does not abolish; it only changes who needs one, and only above a specific SKU.

The capacity is an Azure resource. You buy an F SKU in an Azure subscription, Azure bills per second with a one-minute minimum, and the rate is regional. Every Fabric meter in Microsoft’s East US rate card — Power BI capacity usage, data warehouse, Spark, data movement, eventhouse, Copilot and AI — prices at the same $0.18 per CU-hour, which is what makes SKU arithmetic simple: an F SKU costs its CU count times the hourly rate. Microsoft’s Fabric licenses and capacity documentation (opens in new tab) is the authoritative reference for which SKU carries how many CUs and which per-user license each activity needs.

One structural point worth absorbing before any numbers: Premium Per User is not a capacity. Microsoft states plainly that PPU does not provision a Fabric capacity, so a PPU seat will never run a lakehouse, a warehouse or a notebook. If you need those workloads, an F capacity is not optional — and if you do not need them, the capacity is the part you can skip.

What each F SKU costs, at today’s published rate

The table below applies Microsoft’s East US rates, read from the Azure retail price sheet on 12 September 2026, to the F SKUs a small or mid-market business would realistically consider. Monthly figures use Microsoft’s 730-hour month convention, and the reserved column divides the published one-year commitment of $938 per capacity unit across twelve months. Prices are regional and Microsoft’s to change — check the live rate card before you commit a purchase order.

Two numbers in that table are worth circling. The reservation saves a consistent 40 percent at every size, because it is priced per capacity unit rather than per SKU. And the jump from F32 to F64 doubles the bill — roughly $2,500 a month more on a reservation — which is precisely why the licensing rule attached to F64 matters so much.

Reserved, pay-as-you-go, or paused

A Fabric reservation is bought in one-CU increments for a chosen Azure region and a one- or three-year term, and the matching capacity usage then stops being charged at on-demand rates. It is a commitment, not a discount code, and four conditions in Microsoft’s Fabric capacity reservation documentation (opens in new tab) deserve a read before signing: the reservation covers capacity only and never storage or networking; unused reserved CU-hours expire hourly and do not carry forward; reservations do not renew automatically unless you switch renewal on; and returns are self-service but capped at $50,000 of cancelled commitment across a rolling twelve months.

The per-CU rate is also identical on both terms — $938 for one year and $2,814 for three — so a three-year commitment buys no extra discount over a one-year one, only a longer lock-in. For a first Fabric year, the one-year term is the obvious choice.

Pay-as-you-go is the right starting position for a capacity whose workload you cannot yet describe, and it has a lever a reservation does not: an F capacity can be paused from the Azure portal, which stops it billing, with Microsoft noting that resuming the capacity resumes billing too. Paused content is unavailable, so this works for a development capacity or an overnight batch capacity on a scheduled runbook, not for the one serving reports at 9am. Two more mechanics shape real bills: smoothing spreads background jobs across a 24-hour window and interactive queries across ten minutes, which means you size for average load rather than peak — and capacity overage is metered at $0.54 per CU-hour, three times the standard rate, which is what an undersized SKU actually costs you.

The F64 line, where per-user seats stop

This is the single rule that decides whether Fabric saves money or adds to the bill. On any F SKU below F64, every person who views Power BI content needs their own Pro, Premium Per User or trial license — the capacity buys compute, not viewing rights. At F64 and above, a user holding only a free Fabric license and the Viewer role on a workspace can read that content, which is the point at which capacity starts replacing seats rather than supplementing them.

The arithmetic is unforgiving. An F64 on a one-year reservation is $5,002.67 a month. At Microsoft’s current list prices of $14 for Pro and $24 for Premium Per User, that same money buys roughly 357 Pro seats or 208 PPU seats. So F64 only wins on viewer economics somewhere north of 350 report readers — a scale almost no small business reaches. Below that, buying F64 to avoid Pro seats is paying five thousand dollars a month to avoid a few hundred. The per-seat decision itself is a different question with a different answer, and we cover it separately in Power BI licensing for small business.

OneLake storage, and why it is rarely the problem

OneLake is billed per gigabyte per month, independent of the capacity SKU and not covered by a reservation. Microsoft’s East US rates on 12 September 2026 list hot storage at $0.026 per GB per month, cool at $0.019, cold at $0.004, and business-continuity (BCDR) hot storage at $0.0468. The OneLake cache is a separate and much dearer meter at $0.20 per GB per month.

Put those against a realistic small-business estate and the conclusion is quick: half a terabyte of hot data costs about $13 a month, and two terabytes about $53. Storage is a rounding error next to a capacity that starts at $156 a month reserved. Budget for compute, verify storage, and do not let a tiering exercise distract from the SKU decision — though the cold tier at a sixth of the hot rate is genuinely worth using for archival history you query once a quarter.

You may not need Fabric yet

Most companies who search for this cost are not ready to spend it, and saying so is more useful than selling a capacity. If your reporting need is dashboards over data that already lives in Microsoft 365, SQL Server or a line-of-business system, Power BI Pro seats cover the whole requirement. Fifty Pro seats are $700 a month. The smallest Fabric capacity adds $156 a month reserved on top of those seats and removes none of them — because you are below F64 — so Fabric is strictly additive until either your workload or your viewer count crosses a real line.

Three signals say the line has been crossed: you need genuine data engineering that Pro cannot run at all (a lakehouse, a warehouse, Spark notebooks, orchestrated pipelines); a semantic model has outgrown what Pro and PPU can hold or refresh often enough; or your viewer count has grown past the point where seats beat an F64. If none of those is true, the honest answer is that the money belongs in the reports themselves, not the platform under them. Start with the per-seat licensing guide instead, and note that Microsoft offers a 60-day Fabric trial capacity with F64-equivalent compute — the cheapest way to test whether the workload is real is to run it before you buy anything.

A worked comparison

Illustrative scenario — not a client, and the license mix is the only variable.

Take a 40-person distributor: six people build reports, thirty-four read them, and the data sits in an ERP database and a handful of spreadsheets. Option one is 40 Power BI Pro seats at $14 — $560 a month, no capacity, no Fabric. Option two adds an F2 reservation to run nightly pipelines into a lakehouse, at $156.33 a month; the 40 Pro seats stay, because F2 is below F64, so the total is $716.33. Option three buys an F64 reservation so the 34 readers can drop to free licenses: $5,002.67 for the capacity plus six Pro seats at $84, or $5,086.67 a month — seven times option one to serve the same 40 people.

Option one is right until the pipelines are genuinely needed, at which point option two is a $156 increment rather than a platform migration. Option three does not become rational at this size at all. Run the same arithmetic with your own headcount before anyone quotes you a SKU.

Where the licensing decision meets the work

Capacity is the recurring bill; the build is the one-time one. If you are moving an existing Power BI Premium estate onto an F SKU, that is a scoped project rather than a licensing choice, and our Power BI to Microsoft Fabric migration services cover sizing the replacement capacity against a real consumption baseline and reassigning workspaces — this page prices the SKU, that page runs the move. If what you actually need is the reports rather than the platform, what a Power BI dashboard costs sets out the build bands, and our fixed-fee dashboard service states the license position for your head count in the written scope.

Because Fabric capacity is an Azure resource, it lands on the same invoice as the rest of your subscription and responds to the same disciplines — right-size before you commit, then reserve what survives. The sequencing rules in Azure cost optimization quick wins apply to a Fabric SKU as squarely as to a virtual machine, and Azure migration and cost optimization consulting is where the whole subscription gets reviewed. Where the requirement turns out to be scheduled data movement rather than a full analytics platform, Azure Data Factory consulting services is often the smaller, cheaper answer. Fabric capacity is billed to you by Microsoft through Azure and never through us; what we quote is the fixed fee for the work.