Microsoft Fabric offers a powerful analytics platform with entry-level pricing starting around CHF 300/month for Swiss users. This post explains how Fabric capacity billing works and details two key strategies for reducing operational costs: reservations and pausing idle capacities. It is intended for both newcomers evaluating Fabric and professionals seeking to optimise enterprise spending.
Fabric Pricing Basics
Working with Fabric creates cost which are divided into 4 cost types:
- OneLake Storage
- Mirroring
- Networking (not yet billed)
- Capacity
Storage and mirroring are relatively inexpensive. Each capacity includes 1 TB of free mirroring storage per Capacity Unit (CU). Additionally, cost for networking is not yet billed: Microsoft Fabric – Pricing | Microsoft Azure.
This makes capacity the primary cost driver in Microsoft Fabric and the most effective target for cost optimisation. The following section explains how capacities are structured as well as how usage is calculated and billed.
Fabric Capacities Explained
While unintuitive in the beginning, the cost of Fabric Capacities is pretty straight forward: Fabric uses Capacity Units (CUs) as its core billing metric.
When purchasing a Fabric capacity, you are effectively buying a specific number of Capacity Units (CUs). Each SKU defines the CU count: the number following the “F” in the SKU name corresponds directly to the number of CUs provided. For example, an F2 capacity includes 2 CUs, an F4 includes 4 CUs, and so on.
What is this capacity unit and what does it do? Each second a Capacity unit is running, it contributes a Capacity Unit Second also called “CU(s)”. You can think of this as the currency within Fabric. A F2 capacity which is providing 2 CUs yields a daily budget of 2 × 86’400 s = 172’800 CU(s).
Any operation within Fabric, such as executing a Data Warehouse query or running a Notebook, draws from this CU(s) budget. Lets make a simple example: If you have a copy operation in your F2 Capacity which consumes 8’640 CU(s) and you run it once per day, it is responsible for using 5% of your total CU(s) budget:
8’640 CU(s) / 172’800 CU(s) = 0.05 = 5%
All those operations can be seen within the Microsoft Fabric Capacity Metrics app. The screenshot below shows a copy job which consumed ~1’441 CU(s) and was therefore responsible for using ~1% of the total daily budget (1’441 CU(s) / 172’800 CU(s) = 0.00833).
Cost Savings
Cost effective Fabric usage requires an SKU which is sized according to your workloads. Check the utilization in the Fabrics Capacity Metrics app if you are not sure if you are utilizing a fitting SKU. Assuming you are already working with a adequately sized SKU, there are two options how to effectively decrease your Azure bill:
- Capacity reservations (Save costs with Microsoft Fabric Capacity reservations – Microsoft Cost Management)
- Capacity pausing
Capacity reservation is Azure’s standard mechanism for offering discounts in exchange for a time-based commitment. For Microsoft Fabric, a 1-year reservation yields a 41% cost reduction on capacity. This approach is highly effective for predictable, steady workloads. Have a look at the discount examples and decide if your regular usage patterns would benefit from reservations.
What if you expect very irregular usage or are just doing some proof of concepts? This is where the second way to decrease cost comes in to play: Capacity pausing when it is not in use. Running a capacity while you are not using it is essentially just burning money. Microsoft is not allowing us to store our accumulated CU(s) to use at a later point – use it or lose it. When you pause a Fabric capacity, all your used CU(s) are calculated and it creates a billing event. In the paused state, the capacity will not lead to any more cost until you choose to activate the capacity again.
Be aware that a Fabric capacity has the ability to burst its performance for a short amount of time to speed up your operations which uses more CU(s) than there are actually available in your SKU. This usage above your total capacity is then smoothed over future timepoints. If you pause your capacity, the accumulated smoothed usage and any overages are calculated and billed to you. The resulting spike can be seen in the utilization page of the Fabric Capacity Metrics app:
Summary
This post outlined cost-saving strategies targeting Fabric capacities, the primary cost driver in Microsoft Fabric.
| Strategy | Savings | Best Use |
|---|---|---|
| Reserved capacity | ~41% | Long-term, consistent usage |
| Pause when idle | Variable | Irregular or usage for proof of concepts |
More about Fabric: