.png)
You implemented Microsoft Fabric six months ago. The platform is powerful, and all the teams you serve love having data engineering, business intelligence, and real-time analytics all in one place. Then one day, you get a throttling alert. You are running an F64, which should have plenty of CU - where did all that money go?
This happens to most organizations that deploy Fabric for the first time. The platform delivers incredible results, but the cost structure is complex enough that most companies overspend by 30-50% because they don't fully follow a few proactive steps. Today, we're going to run through an easy framework to audit your Fabric costs and walk through a 90-day plan to reduce spend without sacrificing performance and proactively control Fabric costs. But first, we need to understand why your Fabric bills are higher than expected.
Fabric is always licensed per capacity. The size of capacity you buy determines how many CU operations you can perform per second, which is a proxy metric to calculate shared compute. We need to start thinking of CU’s as money. If you need a deeper explanation of how this works, this article breaks down Fabric and Power BI pricing in detail.
Now let’s dive into the audit framework. The first place to look is where we are using Fabric unnecessarily.
The most common mistake is running Power BI on Fabric when you have viable shared capacity already available. This is double-paying for Power BI.
Organizations running F64 Fabric capacity or higher don't need individual Power BI Pro licenses for viewers, because the capacity covers everyone. But many organizations of this scale also have Microsoft 365 E5 licenses. In that case, all those users already have a Power BI Pro included in their subscription. If you run Power BI in Fabric, you're paying twice for the same thing.
A good rule of thumb: If you don’t use Fabric workloads in a workspace, don’t put that workspace on a Fabric capacity. Click here for a quick explanation of Power BI vs. Fabric workloads.
Shared capacity is often more scalable and cost-effective, especially for organizations that don’t need dedicated resources for every workload. It’s worth revisiting your licensing strategy to see if shared capacity can meet your needs (particularly for non-critical or exploratory workloads).
Cost Saving: Moving Power BI workloads off of Fabric → 100% Fabric savings
Most teams don’t know how many CUs their workloads consume until the capacity throttles. You can stay on top of these workloads with proactive monitoring.
Install the Capacity Metrics app to monitor usage in real time. This tool gives you visibility into how your workloads are performing and where your spend is actually going.
There’s a lot of information, but the quickest value is to open the first page, then sort the table at the bottom by CU(s) descending. It’s very common to see usage data in this shape, with a small number of items dominating CU consumption. Focus on optimizing those items first.
.png)
Once you’ve identified the dominating workload, contact the owner or creator of that workload and share with them the rough cost that one item is consuming. For example, if you spend $10k/mo on Fabric, then in the case above, you would inform them they are consuming 96% or $9,600 of that spend. Work with them to see if there is a best practice they need to follow, or if they really need the frequency or volume they are using.
For deeper analysis, you can use Microsoft’s Fabric Cost Analysis tool or the Fabric Unified Admin Monitoring tool on GitHub. Third-party services like Argus PBI can also provide great insights into Fabric activity. But install the Capacity Metrics app today. Identify your top 10 CU consumers, calculate what each workload costs per month, and you’ll immediately see where the money is going.
Cost Saving: Monitoring for large CU consumption and optimizing → 40-90% Fabric savings
Fabric lets you create multiple capacities, each with its own purpose. Most organizations use one capacity for everything. This may be slightly easier to manage, but usually expensive and unnecessary.
Think about how your workloads actually run. Batch jobs don't need the same resources as interactive workloads, so why treat them the same? Lower-cost capacities work fine for batch jobs, especially when you turn them off between runs. Interactive workloads like dashboards and notebooks need to be responsive at all times, which justify reserved capacity that stays on.
Go look at your capacity utilization right now. Are your capacities running 24/7 when they should be scheduled to run during specific hours? Can you turn them off the rest of the time and save thousands per month? Map your workloads to capacity types, identify opportunities to separate batch and interactive jobs.
This is our recommended capacity structure as of the time of writing this article:
Cost Saving: Putting the right workload on the right capacity → 20 - 80% Fabric savings
Fabric charges for storage, and most organizations treat it like a bottomless pit. That's a mistake that can cost thousands every month.
You start with a few gigabytes. Then a few terabytes. Before long, you're paying thousands per month for data that hasn't been opened in over a year, and the storage bills keep growing in the background until they're taking a serious bite out of your budget.
Data retention policies aren't optional anymore. Archive old data to cheaper storage tiers or move it to external systems where it belongs. Delete datasets nobody uses and set retention limits based on actual compliance and reporting requirements, not hypothetical "what if" scenarios that never materialize.
You can quickly see this in the Fabric Capacity Metrics app. Click the “Storage” page and get an overview of what workspaces are topping the storage list.
.png)
Contact the workspace owners who top this list. Find out if some or all of the data can be deleted. Look for datasets untouched for 90+ days.
Cost Saving: Eliminate unneeded data from storage → 5-20% Fabric savings.
When you do run Power BI on Fabric, data loads and DAX queries will drive your Fabric consumption. Semantic models can vary widely in how much compute they require with very simple adjustments. You never have to worry about this tuning when Power BI runs on shared capacity, but on a Fabric capacity, it matters.
DAX Studio, Vertipaq Analyzer, and Tabular Editor 3 give you the tools to find inefficient DAX expressions, cut memory usage, and speed up query performance without having to guess what's wrong.
Well-optimized semantic models give users faster insights while using fewer resources. Your users get better performance. Your budget gets relief. Everybody wins.
Run Vertipaq Analyzer on your top 5 most-used reports this week. Look for optimization opportunities and calculate savings from reducing memory footprint, then fix the worst offenders first. The lowest hanging fruit is almost always to get rid of unused columns.
Cost Saving: Eliminate unused columns, lowering cardinality, reducing model size, eliminating bi-directional filters and many-to-many relationships → 10-70% Fabric savings
You've audited your Fabric tenant. Time to make changes. First 30 days: quick wins. Next 30 days: bigger moves. Final 30 days: long-term improvements.
Fix double-licensing issues. Install monitoring tools, turn off unused capacities, and archive or delete unused datasets you don't need anymore.
Start a monthly review habit using the steps above to look for the highest consumers of compute and storage. Reach out to your Fabric users to begin to make improvements.
Most organizations cut 10-20% of their spend in the first month just by getting rid of waste. Low effort, fast results.
Create separate capacities for batch and interactive workloads. Set up automated capacity scheduling, optimize your top 10 semantic models, and test new workloads before full deployment so you know what they'll cost.
Build a simple CU-to-dollar converter for your team. Give analysts and developers a way to estimate workload costs before they deploy. This prevents budget surprises and keeps everyone accountable.
Review your reservation strategy. Reserve only what needs 24/7 availability. Move everything else to pay-as-you-go.
Put data retention policies in place. Set up governance for new workload deployment. Train your team on cost-conscious development. Make cost awareness part of your culture, not something you deal with after the fact. By day 90, you're looking at a 30-40% reduction in Fabric spend without giving up performance.
Most organizations never ask this question: Should this workload even be on Fabric?
Just because Fabric can run a workload doesn't mean it should. Fabric is your unified analytics platform, but if your needs are limited to Power BI dashboards, you may not need Fabric capacities at all. Or perhaps you can turn on a small capacity just to enable specific features like Copilot for Power BI without committing to a full-scale deployment.
Be strategic. Use Fabric where it adds value. Avoid it where it doesn't. Review every workload on Fabric and ask yourself: "Does this need to be here?" You might be surprised by how many workloads can live elsewhere and save you money. And remember that Azure can still handle the well-established patterns that don’t require the collaboration and flexibility that Fabric provides.
Microsoft Fabric is a powerful platform, but power comes at a price. To make the most of your investment, you need to monitor usage, optimize workloads, manage capacities like servers, and align spend with business value. Fabric is expensive, but when done right, totally worth it.
The big question of the hour - are you overspending on Fabric? Let's be honest, if you've made it this far, you already know the answer without checking the financials. Or maybe you're curious if you're running workloads that don't even belong on Fabric in the first place. Either way, the audit framework above will show you exactly where your money is going and how to fix it.
Whether you're already running Fabric or want to onboard this tool with a cost-effective strategy, FirstLight has done the hard work and worked with countless businesses to help you through the process. Our team speaks business, not tech jargon. We're easy to reach, so fill out this form, and we'll help you through this.
Click Here To Schedule Your Free Assessment
.webp){kind=small}