And there you have it. It is now in GA.
Dipti Borkar announced it Tuesday on the Fabric Updates Blog: Planning in Microsoft Fabric IQ is generally available, one suite with PowerTable and Intelligence. One sentence in that post matters more than every feature bullet around it. “Planning in MS Fabric billing will start on August 1st, 2026.” That’s Saturday. Everything this series describes has been free to run through preview, and in two days the meter turns on.
The rates came through GA unchanged: 1.16 CU for a Planner, 0.23 for a Stakeholder, 0.05 for a Viewer, the 30-day session. The sizing work is the same work it was on Monday, it just costs money now.
For fifteen years I bought planning software the same way, a per-seat license paid every month for a room full of people who used it hard for four months and barely opened it the other eight. This is the first pricing I have seen built around the way planning actually runs.
Aggressive or expensive
Part 1 went out last week and the feedback split two ways. One group thinks the pricing is aggressive, and they mean it as a compliment: a Stakeholder holds 0.23 CU, about $30 for an active month at list and $18 reserved, and the role charge does not move inside the 30-day window no matter how much they do. (Part 1’s FAQ got updated). The other group thinks it is expensive. The number they keep pointing at is the Planner: 1.16 CU, 58% of an F2, the smallest capacity you can buy. Build a model and more than half your smallest SKU is committed for thirty days, whether that model moves five records or a million.
Same rule, read from opposite ends. Until this week you could have that argument for free. From Saturday, one of the two readings starts showing up on somebody’s invoice.
The expensive argument comes mostly from data engineering persona, where efficiency means output per unit consumed, so a charge that does not scale with the work reads as waste. I hear it and I understand where it comes from. But metering raw compute gets planning wrong from the other side, because a manager keying a few numbers barely moves a meter even though those numbers are what make it a plan. Fabric Planning is priced as an EPM tool, not as compute.
One correction, because the objection usually lands on the wrong role. The person who enters ten records and gets billed as a Planner is the person who created the plan, and building that plan is what made them one, not entering ten records. Once it exists, the next person entering ten records into it is a Stakeholder, around $30. The 58% only opens when somebody touches the model itself.
Sizing starts with your org chart
Before you size anything, sort your people by what they do.
Who changes the model, aka Planners
Who changes the numbers, aka Stakeholders
Who reads the result, aka Viewers
You can read more in Last week’s article on the roles and on sustained CU.
Your Planners are your FP&A or S&OP team, and there are never many of them. Stakeholders are the longer list, every manager who owns a budget or a forecast input plus salespeople if they enter their own numbers, and because that list grows with headcount it is the count to get right.
I sized this against a 2,200-person organization I knew from the inside, as its finance director. We needed about five Planners. Count the managers who own a budget or a forecast input and you land near 120 Stakeholders. Add sixty Viewers for leadership and reporting, and a cycle month comes to roughly 36 CU of role load before headroom.
A planning buffer of about 30% (my assumption, adjustable; it covers the database and processing overhead underneath the plan plus whatever else shares the capacity) puts you near 47 CU. That is an F64 during cycle months, running about three quarters full, with a quarter of the capacity still free for everything else on it.
Figure 1. My 2,200-person scenario in the live calculator: 5 Planners, 120 Stakeholders, 60 Viewers. 47.32 CU with the 30% buffer, an F64 at 74% full, which leaves 16.68 CU of headroom for everything else on the capacity.
The pattern holds at every size: Stakeholders are most of the bill, the Planners stay few, and that ratio is what you size against.
Figure 2. The three worked sizings: the CU each organization lands at with a 30% buffer, the F-SKU that covers it, and what that capacity runs per month at US pay-as-you-go and one-year reservation list rates. Capacity dollars are Microsoft’s published list prices and vary by region; the sizings are my planning assumptions, not Microsoft-published guidance.
One Planner alone occupies 58% of an F2, so the smallest SKU was never a production plan; the realistic floor for even a tiny FP&A-only deployment is F4. So, the pushback about running 58% of an F2, it was never meant for that.
Now put a dollar on it
Those F-SKUs have list prices, so we can leave CU behind and talk money. Now why do people think it’s aggressive?
Because a 500-person company could literally run an F16, even at pay-as-you-go, and pay about $25,000 a year for EPM capability. And that scale math holds at larger organizations.
Now the comparison a CFO will ask for, how does it compare to the market?
EPM vendors do not publish list prices, so I leveraged publicly available data rather than just speaking from my experience.
Vendr’s buyer data puts the median Anaplan contract in the low six figures, roughly $100,000 to $115,000 a year depending on the sample, with entry deployments in the $30,000 to $50,000 range.
For a mid-market 2,200-person company, run an F64 all year and you land near $60,000 on reservation, way under those medians. An SMB can start under $4,000 a year on reservation on an F4. The small enterprise on an F16, even holding capacity all year, lands around a quarter of a median deal.
These are procurement medians rather than published rates. Treat them as a range.
It’ll be interesting to see what people find out, once they have this calculator, about the gap between what they pay today and what Fabric Planning costs.
What GA actually added
The GA post changes what you are buying more than what it costs. Billing events now land in the Capacity Metrics app as production-ready metering, so from Saturday it shows actual charges, where a pilot in preview could only watch the CU draw.
It is the first time any of us can check this math against a bill. Planning, PowerTable and Intelligence are now formally one suite. And the automation surface got much wider: public APIs, CI/CD, deployment pipelines, service principal authentication, direct connectivity to Power BI semantic models and Fabric SQL, and Measure models and trees that build a planning model out of measures you already have.
That automation cluster is where I would slow down, because jobs bill per successful run. Automation volume used to be my honest shrug, since nobody outside your walls can guess it. In preview the automation surface was narrow. GA widened it, and the scheduled loads and recalcs you will wire up through those APIs are exactly the jobs that bill, so count them in the pilot the way you count Stakeholders.
Before the meter starts
Two days is enough time to sort your list. Not really, the main thing I would do before Saturday is to figure out the Fabric Planning artifacts. If you are tinkering with it, make sure you prepare to pay the bill going forward from August 1.
Healthy planning organizations have far more contributors and consumers than builders, so if your sizing comes out Planner-heavy, that is a process problem, and you found it before spending a dollar.
As Benjamin Franklin put it:
“Failing to plan is planning to fail.”
To make your planning easier, you can try the calculator at fabricplanning.io/calculator.
Bigger than the dollars: three to four weeks, down to about a week
Microsoft’s GA post quotes Rich Zielinski, Director of Finance & Treasury at Canlak Coatings, who says that moving off spreadsheets “transformed our forecasting process. What used to take three to four weeks of spreadsheets, emails, meetings and reconciliations can now be completed in about a week.” And now they have “simplified the process into a rolling forecast.”
Why is this personal? Canlak Coatings is a Data Crafters customer. We believe the proof is in the pudding, and that’s what we were busy with through the preview period.
What I want off the quote today is the cycle time, because it is the number this article has been missing. Three to four weeks of a finance team’s calendar, back to about one. The full study will be published by Data Crafters next week.
Final Take
I have done my analysis and made my decision, and I do not give advice where I have no skin in the game. My own company builds on this stack. I did not run the numbers in the abstract either. I ran them against the budget I used to sign as finance director of a 2,200-person organization. We were paying $120k a quarter for planning licenses and an F64 on reservation for a full year is about an eighth of that. So, from my finance background I agree the pricing is aggressive.
The way I read it, Microsoft is betting you will keep opening it and using it. I read that as a vote of confidence in how good they think the product is.
Which brings me back to Saturday. From August 1, the numbers above stop being a modeling exercise, so if you have a deployment running, sort your people by role and decide whether the capacity belongs on pay-as-you-go or on reservation, because Figure 2 prices the same deployment two ways.
Then stop looking at the cost, which is only part of the puzzle. I wrote it to give you real numbers, but what decides it is what the spend unlocks, and Canlak’s answer to that is a forecast cycle that used to eat weeks of a finance team’s year.
The seasonal case I opened with quietly assumes you plan a few months a year, the way finance always has. I doubt that survives. Volatility keeps rising, FP&A keeps moving toward rolling forecasts, and Rich’s team got there ahead of my argument, which is the turn showing up in a live deployment before I even finished making the case for it.
So, start with the process you already have, because that is the safe base case, and count past today’s budget owners to the people who would be in the plan once it runs every month. The version of always-on worth paying for is the one where the spend sits on something your people are inside every month, and the forecast in front of you is not three months old.
Size the tool by counting your managers, not your seats.
Planning in Microsoft Fabric IQ reached general availability on July 28, 2026, and Microsoft states that billing starts August 1, 2026. As of publication Microsoft’s billing document for Planning is still labeled Preview, so the rates here are live-billing rates published in a preview-labeled document; verify current terms with Microsoft before committing budget. Microsoft has published no region statement and no Copilot or AI billing statement for Planning. Fabric capacity dollar figures are Microsoft’s published US pay-as-you-go list prices and vary by region; the reservation figures are derived from Microsoft’s documented reserved discount, and the per-role dollar figures are derived from published capacity rates, because Microsoft publishes no per-role rate. EPM contract figures are third-party procurement medians, not published list prices. Capacity scenarios are illustrative planning assumptions, not Microsoft-published sizing. Canlak Coatings is a Data Crafters customer; the quoted remarks are from Microsoft’s GA announcement.
Sources: Bring Enterprise planning to the AI Era with Planning in Microsoft Fabric IQ, Dipti Borkar, Microsoft Fabric Updates Blog, July 28, 2026. Billing for Microsoft Fabric Planning (Preview), Microsoft Fabric Updates Blog, June 2026. Microsoft Fabric capacity pricing, Azure. EPM contract benchmarks via Vendr procurement data for Anaplan, vendr.com/marketplace/anaplan. Nassim Taleb, “Skin in the Game,” Stanford eCorner. Companion explainer: Understanding Fabric Planning cost: three roles, three costs, one capacity.
This is a great article. Thank you for sharing. I have been trying to figure this out for a few days now. Do you know if there is anywhere a user can go to get an inventory of active Planning sessions? The Metric App shows how many and what kind of sessions are active, but not when they will expire.