[HELP - URGENT] Auto Scaling Microsoft Fabric Power BI

[Gerardv2]

Hello Microsoft team, good morning,

We are evaluating the use of Microsoft Fabric in the company, and I had two questions.

- Does Microsoft Fabric scale? For example, if the company wants to acquire Microsoft Fabric 256 and needs more power, but we don't want to scale to the next version, which is Fabric 512. How does autoscaling work in Fabric? Does it scale by CU? If so, can a limit be set and is it per hour? Does it scale from 256 to 512 and is there no limit?

- I have a certain number of reports, users (developers and viewers), semantic models, and reloads in Power BI. How can I know which Microsoft Fabric I should need? If it's Fabric 64, Fabric 128, or Fabric 256. Is there any way to be certain or have a methodology for performing the analysis?

We look forward to your kind assistance.

Greetings

[Gerardv2]

Hello,
We're still in a bit of a bind. The company has Power Pro, but we need to migrate to Fabric, and the Microsoft estimation tool isn't very helpful. I have these questions.
How can I estimate correctly? Microsoft's own estimator (https://www.microsoft.com/en-us/microsoft-fabric/capacity-estimator) isn't very reliable.
How does Fabric truly scale? Can it scale by CU? And if it scales by CU, is it scaled by hour? Or do I have to wait 24 hours? How much would it cost to scale by CU?
I have a lot of questions.

[v-mdharahman]

Hi @Gerardv2,
I just wanted to follow up on your thread. If the issue is resolved, it would be great if you could mark the helpful reply as solution so other community members facing similar issues can benefit too. If not, don’t hesitate to reach out, we’re happy to keep working with you on this. 

Best Regards,

Hammad.

[v-mdharahman]

Hi @Gerardv2,
We noticed there hasn’t been any recent activity on this thread. If your issue is resolved, marking the correct reply as a solution would be a big help to other community members. If you still need support, just reply here and we’ll pick it up from where we left off.

Best Regards,

Hammad.

[lbendlin]

It scales, but not by single CU. It scales by SKU, effectively doubling the amount of CUs at your disposal (if you scale up). It scales up for a minimum of 24 hours, no matter what your actual usage is.  Except, of course, if you exceed that again, then it will scale up to the next SKU.

if you have consistently been below half the CUs for 24 hours AND autoscale is enabled then it will automatically scale down to the next smaller SKU. Without auto scale you have to do that manually, after at least 24 hours.

Scale up = (roughly) twice the cost.  Actual cost depends on your actual consumption. That's why the cost estimator is so useless.

[v-mdharahman]

Hi @Gerardv2,

Thanks for reaching out to the Microsoft fabric community forum.

Yes, Microsoft Fabric does support autoscaling, and it is based on Capacity Units (CUs). When you purchase a SKU like Fabric F256, it gives you a base of 256 CUs. If your workloads demand more resources than the base provides, autoscale can automatically add additional CUs temporarily, depending on usage and the limits you set.

Here are a few key points you need to know:

* Scaling happens in increments of 1 CU.

* You can set a maximum autoscale limit. For example, you can allow it to scale from F256 up to F320 (adding 64 CUs via autoscale).

* Billing for autoscale is hourly and only for the additional CUs consumed beyond the base. The minimum billing increment is 1 hour.

* If you do not set a limit, autoscale can go up to double the base capacity (e.g., from 256 to 512 CUs).

* Once usage drops, Fabric automatically scales back down to the base.

So yes, you can continue using F256 and allow autoscaling to fill the gap as needed, you do not have to commit to the next full tier (F512) unless consistent usage justifies it.

Now coming to choosing the Right Fabric SKU (64, 128, 256, etc.). There isn’t a one-size-fits-all formula, but here’s a structured approach to estimate the correct SKU. First understand your workload types like PowerBI only usage (report viewing, semantic model refreshes, etc.) or Lakehouse or Warehouse usage (Spark jobs, pipelines, notebooks) or Real-time analytics or data science workloads. Now evaluate the following factors like number of concurrent users (viewers & developers) or number and size of semantic models. The frequency and concurrency of data refreshes and volume and complexity of Lakehouse/Warehouse operations. You can use Capacity Metrics App (Power BI) to analyze the past resource consumption or peak usage times, Query durations and refresh bottlenecks

I would also take a moment to thank @lbendlin and @NandanHegde, for actively participating in the community forum and for the solutions you’ve been sharing in the community forum. Your contributions make a real difference.

If I misunderstand your needs or you still have problems on it, please feel free to let us know.  

Best Regards,
Hammad.
Community Support Team

[v-mdharahman]

Hi @Gerardv2,

As we haven’t heard back from you, so just following up to our previous message. I'd like to confirm if you've successfully resolved this issue or if you need further help.

If yes, you are welcome to share your workaround and mark it as a solution so that other users can benefit as well. If you find a reply particularly helpful to you, you can also mark it as a solution.
If so, it would be really helpful for the community if you could mark the answer that helped you the most. If you're still looking for guidance, feel free to give us an update, we’re here for you.

Best Regards,

Hammad.

[NandanHegde]

Below are the diff Fabric capacity tiers:

You can scale to max 2048 tier and not beyond that.

For auto scaling detail :

https://www.reddit.com/r/MicrosoftFabric/comments/1i81gt8/how_to_automatically_scale_fabric_capacity...

ideally there is no fixed way but usually it is suggested to start with F64 and based on the utilization aspect , scale up or down accordingly

[lbendlin]

Note that once you have scaled up you need to stay there for 24 hrs before you can scale down again.