
04 - Compute and Capacity Units (CUs) - The Microsoft Fabric POD
Welcome back to Episode 4! Today, let's talk about compute power and how pricing actually works in Fabric.
In traditional cloud setups, managing your bill is painful because you pay for every single tool separately:
• You pay for a data pipeline runner in Azure Data Factory.
• You pay for a Spark cluster that might be sitting idle.
• You pay for an always-on dedicated SQL database.
If no one uses that SQL database over the weekend, you still get charged for every single hour unless someone writes a script to shut it down on Friday evening.
Fabric fixes this by using something called Capacity Units, or CUs.
Instead of renting individual servers for every tool, you buy one shared bucket of computing power for your whole team.
Think of it like a shared office Wi-Fi connection:
• When your data pipeline runs at 2:00 AM, it uses power from that pool.
• When your data engineer writes Spark code at 10:00 AM, it draws from the same pool.
• When managers check Power BI reports after lunch, they use that exact same capacity.
Everyone shares the same pool, and you only pay one single compute bill.
Now, what happens when someone runs a massive job that takes a lot of processing power?
Fabric handles this with a feature called smoothing.
If a heavy query creates a sudden 5-minute spike in usage, Fabric doesn't immediately crash your system or throw an error. Instead, it spreads out and averages that
load over a wider window of time. Your reports keep running, your users don't get blocked, and you don't have to overpay for a massive server just to handle a
temporary 5-minute rush.
It turns computing power into a simple shared utility, rather than ten different servers you have to babysit.
In Episode 5, we’ll look at how this changes Power BI reporting using Direct Lake.
About the Host
Rajnish Pandey is a Senior Data Engineer with over decade of experience in the data industry. Through QueryZens, he shares practical insights, real-world experiences, and conversations around Data Engineering and modern data platforms.
Hosted by Rajnish Pandey | QueryZens
Making Data Engineering easier to understand, one conversation at a time.
More episodes from "The Microsoft Fabric Pod (by Queryzens)"



Don't miss an episode of “The Microsoft Fabric Pod (by Queryzens)” and subscribe to it in the GetPodcast app.








