Fabric replaces Power BI Premium per Capacity, Synapse Analytics, and parts of Azure Data Factory under one Capacity Unit pool. F SKU ladder, workload consumption math, the Power BI Premium migration cost, OneLake storage at $24 per TB, one year reservation cuts forty one percent, eleven buyer moves.
Microsoft Fabric is priced by the capacity you reserve, not the reports you run, so the negotiation lives in the Capacity Unit baseline and the choice between reservation and pay as you go.
This guide is for data, finance, and procurement leaders sizing a Microsoft Fabric commitment. Pair it with the Microsoft Practice page and the Azure Enterprise Agreement guide before you reserve capacity.
Fabric is sold as capacity, named by an F number that maps to Capacity Units. The ladder runs from F2 at the entry, around 263 dollars a month pay as you go, up to F2048 near 269,108 dollars a month. Microsoft publishes the Fabric pricing for each step.
A one year reservation cuts the capacity rate by close to 40 percent against pay as you go, on the terms set out in the Microsoft Product Terms. The trade is commitment. Reserve the steady baseline, and burst the peaks on pay as you go.
Microsoft Fabric F SKU ladder, indicative monthly
| SKU | Capacity Units | Pay as you go, monthly | Note |
|---|---|---|---|
| F2 | 2 CU | around 263 dollars | Entry capacity |
| F64 | 64 CU | around 8,410 dollars | Power BI Premium P1 equivalent |
| F512 | 512 CU | around 67,277 dollars | Large enterprise tier |
| F2048 | 2048 CU | around 269,108 dollars | Top of the ladder |
You buy a fixed pool of Capacity Units, and every workload draws from that pool. The bill is the SKU, not the workload count. Throttling, not a higher invoice, is what happens when you exceed the pool.
Fabric smooths short spikes across time so a brief peak does not force a larger SKU. Microsoft documents the capacity model, including how bursting and smoothing interact with throttling.
Size to the smoothed baseline, not the single highest peak. The Capacity Metrics app shows real consumption, so you can pick the SKU that holds the steady load and let smoothing absorb the rest.
Power BI Premium per capacity moved into the Fabric F SKU model. F64 is the practical equivalent of the old P1 capacity and unlocks free Power BI viewing for content hosted on it. Microsoft documents the Premium transition.
The standard guidance is to size Fabric to your peak workload so nothing ever throttles. We disagree. In roughly 14 of the 24 Fabric estates we benchmarked, the reserved SKU was one or two steps above the smoothed baseline, paying for headroom that smoothing already covered. The buyer side move is to size to the smoothed baseline from the Capacity Metrics app, reserve that level, and burst rare peaks on pay as you go. Throttling on a brief spike costs minutes. A SKU bought two steps too high costs every month of the term.
Source: Redress Compliance advisory engagement file, 2024 to 2025.
Fabric is priced by the capacity you reserve, not the reports you run. The negotiation lives in the Capacity Unit, not the seat.
Morten Andersen. Co Founder. Ex IBM, ex Oracle.
Leverage comes from a measured baseline and a credible mixed model. Reserve only what the smoothed load needs.
White Paper · Microsoft
Seven buyer side levers for a Microsoft Fabric negotiation: capacity SKU sizing, pause economics, the F SKU commit trap, and the discount to push for. Read it free.
Microsoft Fabric is priced as capacity, named by an F number that maps to Capacity Units. The ladder runs from F2 near 263 dollars a month pay as you go up to F2048 near 269,108 dollars a month, with a roughly 40 percent saving on a one year reservation.
A Capacity Unit is the compute measure behind each F SKU. You buy a fixed pool of Capacity Units, and every workload draws from that pool. The bill is the SKU you reserve, not the number of workloads or reports you run.
Reserve the steady baseline and burst the peaks on pay as you go. A one year reservation cuts the capacity rate by close to 40 percent against pay as you go, so the reservation suits predictable load and pay as you go suits spikes.
F64 is the practical equivalent of the old Power BI Premium P1 capacity. F64 and above unlock free Power BI viewing for content hosted on that capacity, while below F64 viewers still need a Power BI Pro license.
Size to the smoothed baseline, not the single highest peak. The Fabric Capacity Metrics app shows real consumption, and smoothing absorbs brief spikes, so a SKU that holds the steady load usually beats one bought for the peak.
Smoothing spreads short bursts of demand across time so a brief peak does not force a larger SKU. When demand exceeds the pool, Fabric throttles rather than raising the invoice, which is why sizing to the baseline is safe.
Across the Fabric estates we benchmarked, many reserved a SKU one or two steps above the smoothed baseline. Right sizing to the baseline and reserving only that level, then bursting peaks, recovered a large share of the capacity spend.
Yes. Fabric capacity can sit inside an Azure commitment such as a MACC, which can improve terms and consolidate the spend. Aligning the reservation to the June Microsoft fiscal year end adds negotiating leverage.
The full paper covers F SKU pricing F2 through F2048, Capacity Unit consumption math by workload, OneLake storage economics, the Power BI Premium per Capacity migration cost, Fabric Copilot consumption uplift, the Snowflake and Databricks competitive math, the named pitfalls, and the eleven move buyer side playbook with dollar values against each move.
Used across more than five hundred enterprise software engagements. Independent. Buyer side. Built for Microsoft customers running the next Microsoft Fabric renewal cycle.
No download. The paper opens in your browser. Corporate email only (we reject Gmail, Yahoo, Hotmail, Outlook, AOL, and similar free providers).
Microsoft quoted us an F128 to cover the full data estate. Redress mapped CU consumption by workload, separated Spark heavy ELT from Power BI reporting, and ran Databricks as a credible competitive frame. We landed at F64 with reserved capacity, dropped Synapse, and kept OneLake as the single storage layer. Twenty six percent under the original quote.
Confidential consultation. No follow up sales call unless you ask for one.
Microsoft Fabric framework signals, Capacity Unit signals, F SKU signals, reservation signals, Snowflake vs Microsoft Fabric vs Databricks competitive signals, and the broader unified data and AI licensing leverage signals.
