In Short: What Does Microsoft Fabric Cost in 2026?
Microsoft Fabric is priced on capacity: you buy an F SKU (F2 through F2048), each providing double the Capacity Units (CUs) of the one below it. Pay-as-you-go starts at roughly $263 per month for an always-on F2 in US regions, a 1-year reservation cuts roughly 41% off, and the price doubles with each SKU step. On top of capacity you pay separately for OneLake storage (commodity data-lake rates) and per-user Power BI licences where needed.
This guide gives you the full price list and three worked examples. Prices below are approximate for US regions as of mid-2026 - always confirm current rates for your region on the Azure pricing page. For how the licence model works (who needs which licence), see our Fabric licensing guide; for how usage is measured once you are running, see how Fabric cost measurement works.
The Three Things You Pay For
Every Fabric bill is made of three components:
- Capacity (F SKU) - the pool of compute all your Fabric workloads share. This is the big number.
- OneLake storage - billed separately per GB per month, at roughly $23 per TB. Cheap, but not zero.
- Per-user licences - Power BI Pro (around $14/user/month) for anyone publishing or sharing Power BI content; viewers are free on F64 and above.
The F SKU Price List (2026)
Approximate US-region pricing. Pay-as-you-go is billed hourly and can be paused; reserved requires a 1-year commitment and saves roughly 41%.
- F2 - 2 CUs - about $0.36/hour pay-as-you-go (about $263/month always-on) or about $156/month reserved
- F4 - 4 CUs - about $526/month pay-as-you-go or about $311/month reserved
- F8 - 8 CUs - about $1,051/month pay-as-you-go or about $623/month reserved
- F16 - 16 CUs - about $2,102/month pay-as-you-go or about $1,245/month reserved
- F32 - 32 CUs - about $4,205/month pay-as-you-go or about $2,491/month reserved
- F64 - 64 CUs - about $8,410/month pay-as-you-go or about $4,982/month reserved
- F128 - 128 CUs - about $16,819/month pay-as-you-go or about $9,964/month reserved
- F256 - 256 CUs - about $33,638/month pay-as-you-go or about $19,928/month reserved
- F512 to F2048 - continue doubling per step, reserved discounts apply throughout
Two practical notes. First, pay-as-you-go capacities can be paused - a development capacity running only working hours costs roughly a quarter of the always-on number. Second, prices vary by Azure region, sometimes meaningfully; the ratios between SKUs stay constant.
Pay-As-You-Go or Reserved?
The decision is about workload stability, not size:
- Pay-as-you-go for evaluation, development, and spiky workloads - you pay hourly and can pause
- Reserved (1-year) for steady production - roughly 41% cheaper for the same SKU
The common pattern is a reserved capacity for production plus a small pay-as-you-go capacity for development and testing that gets paused outside working hours.
The F64 Threshold: The Most Important Line in the Price List
On F64 and above, report viewers no longer need Power BI Pro licences - anyone with a free licence can consume Power BI content. Below F64, every viewer needs Pro at around $14/user/month.
That creates a break-even worth modelling: a few hundred viewers on Pro licences cost more per year than the jump from F32 reserved to F64 reserved. Organisations with wide report audiences almost always land on F64 for this reason, not for the compute.
Three Worked Examples
Small team getting started
F2 reserved (about $156/month) plus 5 Power BI Pro licences (about $70/month): about $226/month, plus a few dollars of storage. Enough for a lakehouse, pipelines, and a handful of governed reports.
Mid-market analytics platform
F16 reserved (about $1,245/month) plus 20 Pro licences (about $280/month): about $1,525/month. Runs a full medallion lakehouse with scheduled refreshes and department-level reporting for a few hundred employees.
Enterprise estate
F64 reserved (about $4,982/month) plus Pro for 40 creators (about $560/month): about $5,542/month - with unlimited free viewers. Compare that against 500 viewers on Pro below F64 (about $7,000/month in licences alone) and the F64 economics explain themselves.
How to Size Your Capacity
Data volume is the wrong sizing input. CU consumption follows refresh frequency, query concurrency, and workload efficiency - measured in 30-second windows. Two organisations with identical data sizes can need SKUs two tiers apart.
The reliable approach: start one tier lower than your instinct on pay-as-you-go, run real workloads for two to four weeks, watch the Capacity Metrics app, then reserve the SKU the data supports. Our regional cost guides cover full project budgets beyond the platform itself: UK, US and South Africa.
If you want the sizing done against your actual estate, our Microsoft Fabric consulting team runs capacity and licensing reviews as standard.
