Fabric Lab · Issue 02
A CIO forwarded me his Fabric renewal deck last month. One slide said moving to F64 capacity would “pay for itself in Power BI Pro license savings.” . BTW , we also made similar mistakes recommending our customers initially.
The quote next to that slide added roughly over $8,000 a month to his bill.
No-one from customer or vendor team had run the actual crossover math. So I did, on his real numbers. It did not pay for itself. It was not close.
What Actually Happens At F64
Microsoft Fabric capacity comes in SKUs: F2, F4, F8, all the way up to F64, F128, and beyond. Below F64, every person who views a Power BI report built on that capacity needs an individual Pro license, at $14 a month per viewer, or a Premium Per User license.
At F64 and above, viewers can use a free Fabric license instead. No per-user Pro cost.
That is a real saving. The question a lot of partners skip is: a saving big enough to justify what?
F64 capacity runs roughly $5,000 a month reserved, or around $8,400 a month pay-as-you-go. Run the actual crossover against $14-a-month Pro licenses and the breakeven lands around 600 regular viewers. Against Premium Per User pricing, it is closer to 210.
A mid-market manufacturer with 40 plant floor supervisors checking a shift dashboard is nowhere near either number. Paying $5,000 a month to save $560 a month in Pro licenses is not a licensing optimization. It is a bad trade with a good story attached.
Why This Pitch Works Anyway
The F64 story is technically true and directionally useless for most mid-market manufacturers, which is exactly why it survives in so many decks. It sounds like a savings pitch. It is usually a capacity pitch wearing a savings costume.
The real reason most organizations end up at F64 has nothing to do with viewer count. It is compute consumption: pipelines, dataflows, and Copilot features that eat capacity units regardless of how many people are looking at a report. When a partner needs to justify a capacity jump that is really about compute, “you will save on Pro licenses” is an easier sentence to say in a renewal meeting than “your workloads outgrew F32.”
Both reasons can be true at once. Only one of them is usually the real driver, and it is worth knowing which before you sign.
Run This Before You Renew
Three moves, none of them require touching your data model.
1. Count your actual viewers, not your license count.
Pull the real number of distinct people who open a Fabric-hosted report in a typical month. Compare it to 600 (Pro) or 210 (PPU). If you are not within shouting distance, viewer-license savings should not be in the justification at all.
2. Ask what is actually consuming capacity.
If the real driver is compute, not viewers, get the consumption report broken down by workload. That is the number that tells you whether F64 is a genuine need or a preemptive upsell.
3. Ask your partner for the crossover math, in writing.
Not “what capacity do we need.” Ask: “at what viewer count does F64 pay for itself in Pro savings, on our numbers.” If the answer does not show their work, they have not run it either.
What This Means For You
If your renewal cites Pro license savings as the main justification for F64, and your real viewer count is under a few hundred, ask for the compute breakdown instead. The savings story alone will not hold up.
If you are already well past 200-300 regular viewers, the F64 math may genuinely work in your favor, and it is worth having your partner show it on your real numbers rather than a generic slide.
Either way, this is a spreadsheet problem, not a data architecture problem. It is fixable in an afternoon once someone actually runs it.
The One Ask
If you want the real crossover math run against your actual Fabric bill and viewer count, I am opening five slots this month for an F-SKU Right-Sizing Audit.
One week. Fixed scope. Your real viewer count, your real capacity consumption, and the exact dollar crossover point, in writing.
No pitch at the end. You keep the number whether or not we ever talk again.
Reply here or DM. First five, first booked.
And if your renewal deck used this exact pitch, tell me in the comments: did anyone show you the crossover number, or just the slide?
I read every reply.
Next issue: The 10 Boring Projects. The unglamorous work that pays for the AI project you actually want to do, in the order that pays fastest.
Click here to read this article on Dave’s Demystify Data and AI LinkedIn newsletter.