Recommendation
All Fabric SKUs & monthly cost at $0.18/CU/h
| SKU | Capacity units | Max model (GB) | PAYG / mo | Reservation / mo |
|---|---|---|---|---|
| F2 | 2 | 3 | $263 | $155 |
| F4 | 4 | 3 | $526 | $310 |
| F8 | 8 | 3 | $1,051 | $620 |
| F16 | 16 | 5 | $2,102 | $1,240 |
| F32 | 32 | 10 | $4,205 | $2,481 |
| F64 | 64 | 25 | $8,410 | $4,962 |
| F128 | 128 | 50 | $16,819 | $9,923 |
| F256 | 256 | 100 | $33,638 | $19,847 |
| F512 | 512 | 200 | $67,277 | $39,693 |
| F1024 | 1,024 | 400 | $134,554 | $79,387 |
| F2048 | 2,048 | 400 | $269,107 | $158,773 |
How to choose the right Microsoft Fabric capacity
A Fabric F SKU is a shared pool of compute measured in capacity units (CUs). Sizing it comes down to three capacity questions — this calculator returns the smallest SKU that satisfies all of them.
1. Your largest model sets a hard memory floor
Each SKU caps the memory a single Power BI semantic model can use: F16 allows 5 GB, F32 10 GB, F64 25 GB, F128 50 GB, and so on. If your biggest model is 12 GB, anything below F64 simply won't load it — that constraint alone can decide your SKU.
2. Your user base drives query compute
Not everyone hits reports at once — we estimate peak concurrency at about 15% of your total users (adjustable in the calculator). The more concurrent load, the more CUs you burn on interactive queries. A small SKU comfortably serves a hundred-odd users of light reports; thousands of users need F32 or larger.
3. Fabric workloads add background compute
If you run data engineering, warehouse, pipelines or Spark, the volume of data you process pushes the floor up — those jobs consume CUs on top of your reporting load. Power BI-only capacities can usually run leaner.
Pay-as-you-go vs reservation — and pausing
Pay-as-you-go bills per minute and can be paused, ideal if you only need the capacity during business hours. A 1-year reservation is roughly 41% cheaper than 24/7 pay-as-you-go, but you commit to the full term. Rule of thumb: run it most of the day → reserve it; switch it off nights and weekends → pay-as-you-go with a pause schedule usually wins.
How this calculator estimates your SKU
It takes the highest of three floors — the memory your largest model needs, the compute your user base needs (via estimated peak concurrency), and (for full Fabric) the compute your data volume needs — then prices that SKU on pay-as-you-go and reservation. It's a starting point for planning; always confirm real usage with the Capacity Metrics App over a 2–4 week window.
Frequently asked questions
Which Microsoft Fabric F SKU do I need?
It depends on three things: your largest Power BI semantic model (it must fit the SKU's memory limit — F32 = 10 GB, F64 = 25 GB, F128 = 50 GB), how many users you have (and how many are active at peak), and — for full Fabric workloads — the volume of data your pipelines and warehouse process. This calculator returns the smallest SKU that satisfies all three.
How much does Microsoft Fabric capacity cost?
Capacity is billed per capacity unit (CU) at about $0.18/CU/hour pay-as-you-go — roughly $263/month for F2 and $8,410/month for F64 running 24/7. A 1-year reservation saves about 41%.
What is the difference between pay-as-you-go and a reservation?
Pay-as-you-go is billed per minute and can be paused, so you only pay for the hours the capacity runs. A 1-year reservation is about 41% cheaper but bills for the full term whether or not the capacity is running.
Can I pause Fabric capacity to save money?
Yes — pay-as-you-go capacities can be paused. If you only need the capacity during business hours, pausing nights and weekends can cut the pay-as-you-go cost by half or more. Reservations cannot be paused.
Get more out of your Fabric capacity
DataTako runs on any Fabric capacity from F2 up, with auto-pause and usage-aligned refresh windows to keep your capacity cost down. See how it fits your setup.
Plan a demo →