Microsoft Fabric Pricing Scenarios

One of the most common questions we hear from customers about Microsoft Fabric is “how much is this going to cost?” Now that Fabric is generally available and more detail has been published, there is still not yet an easy answer. Microsoft published a blog post detailing the costs of capacities and OneLake storage along with official pricing, but it’s still challenging to understand exactly how much capacity you will need. The recommended strategy is to spin up a trial and build out some of your workloads, then monitor it through the Fabric Capacity Metrics report. Based on the usage you see there, you can get an idea of the capacity or capacities that you’ll need once the trial period is over. 

In this post, we’ll provide an overview of what goes into Fabric pricing, then go through a few sample scenarios to shed some light on what you can expect to see on your bill.  

Fabric Costing Components 

Your overall Fabric costs will be based on a few different components. The first is individual licenses for Power BI, the second is OneLake storage, and the third is capacities. The first two components will be the easiest to determine, so let’s tackle those first. 

Individual User Licenses 

For traditional Power BI functionalities like creating and viewing reports and dashboards, an individual Power BI Pro license may or may not be required. This has not changed with Fabric. If your users already have M365 E5 licenses, Power BI Pro licenses are included so no additional Power BI licenses are required. Without E5, all users (developers and consumers) will need Pro licenses if no Premium capacity at F64/P1 or above is purchased. If a Premium capacity is purchased, consumers do not need a Pro license, but developers still do. Finally, if a Premium Per User (PPU) model is chosen instead of a dedicated Premium capacity, all developers and consumers will need a PPU license for content published to workspaces assigned as PPU. PPU is the lowest cost entry point for Power BI Premium features without purchasing a dedicated capacity, but includes no other Fabric workloads. For non-Power BI Fabric activities, no individual user license is required. 

You can use the Power BI Licensing Calculator to determine which combination of Pro/Premium/PPU/Free licensing will meet your organization’s needs for Power BI content.

OneLake storage 

Storage costs will typically be far less expensive than compute costs. According to the blog post mentioned above and official documentation, OneLake storage pricing is comparable to Azure ADLS pricing. For US West 2, the cost will be $0.023/GB/month, or $23/TB/month. There may also be cross-region data transfer network charges (also known as egress), so be aware of those to avoid any surprises. Additionally, there is a OneLake cache cost (for KQL cache storage and retained Data Activator data) of $0.246/GB/month and BCDR (Business Continuity and Disaster Recovery – can be optionally enabled or disabled) storage of $0.0414/GB/month.

Capacities  

The third, and most complicated, aspect of Fabric licensing is determining how much capacity you need, and how many capacities. The table below shows the current published prices of Fabric dedicated capacities along with their Power BI Premium counterparts. The RI (reserved instance) price is generally 41% lower than pay-as-you go and requires an annual commitment.

One of the biggest differences between Power BI Premium dedicated capacities and Fabric capacities is the pay-as-you-go option for Fabric. With Power BI Premium, there was simply a monthly rate. With Fabric pay-as-you-go, we can scale up or down, and even pause. This allows us to have a lot more control over how much we use and pay for, but it also makes our determinations of what to buy more complicated. A few important considerations to help us decide are: 

Each Fabric workspace is assigned to a single capacity. 

• Power BI content in Fabric-assigned workspaces lower than F64 requires consumers to have Power BI Pro licenses. 
• Copilot in Fabric is only available in the F64 SKU and above (and is not supported in the trial SKUs) 
• Non-Power BI workloads do not require individual licenses. 
• The capacity needed is dependent on the compute required. Right now the best way to determine how much compute you need is to start a trial and build out workloads to monitor with the Fabric Capacity Metrics report. 
• If a Power BI report is using Direct Lake mode and the capacity for the workspace is paused, the report will display no data. 
• If you already have a Premium capacity and find that you’re not using it all, you can add some Fabric workloads and/or scale down as necessary. 
• Reserved instance pricing will save money if you’re using the capacity at least 60% of the time. 
• Power BI Premium per User (PPU) workspaces do not support non-Power BI Fabric workloads. 
• Fabric capacities take advantage of bursting and smoothing. This means that if you have a workload that is resource intensive Fabric will “burst” more resources to get the job done faster. Smoothing means that your capacity utilization is calculated as an average over a period of time rather than based on occasional spikes of high usage.  
• Coming soon are additional capabilities like autoscale and configurable surge protection. 

On March 14, 2024, Microsoft announced the phasing out of Power BI Premium per capacity SKUs (P-SKUs). The exact timing of this retirement will depend on your existing agreement but the bottom line is that only Fabric capacities will eventually be available for purchase and renewal. This concerns only the cloud Power BI service, Power BI Report Server, will only be available through SQL Server Enterprise Edition with Software Assurance. Power BI Pro and Power BI Premium Per User (PPU) remain unaffected.

Copilot

Development/Test/Production 

Currently with Power BI Premium capacities, it can be difficult to justify having a copy of the same large imported dataset in three different workspaces if you need it in a development, test, and production environment. There are workarounds such as using parameters to limit the amount of imported data within each workspace, but that requires configuration of parameters and rules in deployment pipelines. With Fabric capacities, one option is to have a different capacity for each phase of the lifecycle and pause development and test when they’re not needed, then use a reserved instance capacity for production since it will likely need to be running all the time. 

Power BI vs Non-Power BI Workloads 

With Power BI reports, we generally always want them available to our end users, so we should place those objects in a workspace assigned to a reserved instance capacity. For other workloads such as data pipelines that only run once a day, we can use a pay-as-you-go capacity to run it only when we need it, or we can keep it at a lower capacity most of the time for smaller workloads and scale it up for the daily run. It’s important to keep in mind, however, that Power BI reports using Direct Lake mode won’t work if the capacity being used for the underlying lakehouse is paused. 

Medallion Architecture 

If you’re following the medallion architecture model of bronze (raw data), silver (validated and deduplicated data), and gold (transformed, refined, and business-ready data) layers, you have one more opportunity to consider costs before assigning capacities. Because the unprocessed raw data in the bronze layer may not need to be accessed on demand, it could be placed in a pay-as-you-go capacity that is paused or scaled down when not needed. Gold layer data used as a source for Direct Lake mode Power BI reports will need to be always accessible so should be in a workspace assigned to a different capacity. 

Final Thoughts 

Microsoft Fabric’s pricing model provides a valuable opportunity for cost-savings but requires careful planning of all the moving parts to make the best choice. By combining both the pay-as-you-go option with the reserved instance pricing based on your unique scenarios, you can get the most return on your investment. While the free trial is available, now is the time to get started to have the best understanding of what your future Fabric costs will be. 

Please contact us for a free data strategy briefing to learn how Fabric can help you accelerate, de-risk, and streamline your data and analytics strategy in the most cost-effective way. Now that Fabric is generally available and more detail has been published, there is still not yet an easy answer.pilo