Fabric Planning capacity & cost
Size your Microsoft Fabric capacity by user role, then see the F-SKU that fits and what it costs each month.
Who is using Fabric Planning?
Rates below are sustained CU, the average capacity each user keeps occupied.
* CU-equivalent: total CU-hours over the 30-day period divided by the hours in that period, the same conversion that defines the sustained CU rates above.
Your plan
Region only changes the dollar price of one CU, not how much capacity you need.
Pricing as of June 16, 2026 · figures in USD.
How Fabric Planning is billed
For people, the cost follows a 30-day pattern: the first time a user takes an action, that action sets their role, and that role's capacity is billed for the next 30 days. Background automation jobs bill differently, per successful run; see theAutomation Jobs section below. Five things set the bill for people.
- Sustained CU is a seat license, not a meter. The first time a user acts in a role, they consume a 30-day license for that role, and that license is what sustained CU represents: a Planner's license holds 1.16 CU, a Stakeholder's 0.23 CU, a Viewer's 0.05 CU, occupying a fixed share of your capacity for the full 30-day period whether they log in every day or once. It is not CU per hour and it is not CU-seconds, so it lines up directly with the SKU number. One Planner alone holds 1.16 of an F2's 2 CU, about 58% of the whole tier, which is why F2 cannot host a real deployment.
- Your action sets your role, and your role sets the bill. You do not pick a role; your first action picks it for you.View only and you are a Viewer.Enter or write back data and you are a Stakeholder.Build or model the plan and you are a Planner. A user who holds several roles is billed once, at the highest role only.Learn more about roles.
- A session is a 30-day commitment. Once an action opens a session, its full 30 days are billed. The sustained draw runs for the whole window from the moment it starts, and closing the session early does not reduce the cost. Sizing is about who is active in the same 30 days, not how many people have access.
- Add up the sustained CU and pick an F-SKU above it. Your total sustained CU across everyone active, plus the CU-equivalent of any automation jobs, is the capacity your plan needs. Microsoft meters this per second underneath, but from a capacity view it reads cleanly as the share of your F-SKU each role occupies.
- Region changes price, not capacity. The number of CUs each role consumes is the same in every region. What changes by region is the dollar price of one CU, which is what the region selector adjusts.
This calculator is kept deliberately simple: one consistent unit and one set of numbers. For the fuller, more detailed capacity estimator, see Lumel, Microsoft's development partner for Fabric Planning, atlumel.com. For the current dollar cost of a CU in your region, see Microsoft'sFabric pricing page.
One open question: whether Copilot and AI features draw from the same CU pool or bill separately. See the costing question below for what we know so far.
Prices are per-CU figures derived from published F2 capacity pricing (F2 = 2 CU), in USD, as of June 16, 2026. Reservation reflects a 1-year monthly equivalent. Confirm current rates with Microsoft, as pricing changes.
What a single role costs on its own
The common case: someone opens a session and stays at one role for the full 30 days, no upgrade.
What do Viewer, Stakeholder, and Planner mean? →
What happens when you change roles
The section above shows the common case: one role, flat for the full 30 days, no upgrade. This section shows the other case, a role change mid-session, using one sample scenario: the same user moving from looking, to entering data, to building. Roles are earned by action. Watch what an upgrade does to the bill, day by day. The chart's vertical scale adjusts to each scenario, so compare the CU labels on the upgrade markers, not the bar heights, across tabs.
Microsoft states a session as 30 days (730 hours), which makes the official per-session figures ~37 (Viewer), ~168 (Stakeholder), and ~847 (Planner) CU-hours. The daily chart uses clean 24-hour days (720 hours per session), so its totals run slightly under the official figures. Billing is metered per second underneath; budget with the official 730-hour figures.
"Automation Jobs" billing
An automation job is work Fabric Planning does without a user in session: a background process runs, not a person clicking through the app. Two examples show up in practice:
Planning-sheet sync
A detailed-level plan is saved. A backend automation job re-runs the full P&L plan to keep the two plans in sync. No human runs it; the job is expected to bill an automation fee.
PowerTable intake flow
A customer intake form built with PowerTable gets a submission approved. Downstream creates and updates fire automatically. Each triggered run counts as one automation job.
This is a PowerTable- and planning-sheet-native capability, not a Power Automate feature.
Automation jobs are like workflows inSmartsheetSmartsheet includes 250 automations/month on Pro; Business and up are unlimited.orAirtableAirtable includes 100 (Free) to 500,000 (Enterprise) automation runs/month; when you hit the cap, automations stop until you upgrade., except instead of a per-seat tier cap, there is no bundled allowance and no monthly run cap to upgrade past: each job draws its own CU, at a small, disclosed rate, against the same F-SKU capacity you already own (2 CU-hours per successful job: ≈ $0.36 pay-as-you-go, ≈ $0.21 with a 1-year reservation*). See the calculator abovefor your region's figures.
* Central US pricing, as of June 16, 2026. Other regions price a CU-hour higher or lower; confirm current rates with Microsoft.
Why "CU-equivalent," not sustained CU: automation jobs bill per successful run, not as a continuously-held seat the way roles do. To size your F-SKU with one number, the calculator converts their total CU-hours into an average CU (divided by 730, the same conversion behind the role rates), which is why automation appears in the same tank and breakdown even though it bills differently. Full explanation in the FAQ below.
Microsoft has confirmed automation jobs draw CU on a per-job basis, distinct from the continuous sustained-CU model roles use, at a fixed cost per successful job. The exact per-job CU draw and what counts as a trigger are still being confirmed with the product team; the calculator above uses 2 CU-hours per job as a working estimate. We will update this section when we have the mechanics in writing.
Common costing questions
The pricing questions we hear most. For everything else about Fabric Planning, see the full FAQ; for who should hold which role, see the roles page.
What is a sustained CU?
A sustained CU is a seat license. The first time a user acts in a role, they consume a 30-day license for that role, and that license occupies a fixed share of your F-SKU for the 30-day period, whether they log in every day or once. It is not a running meter: no CU per hour, no accumulating CU-seconds to add up. The number maps directly to the F-SKU scale. A concrete example: one Planner holds 1.16 CU, which is 58% of an F2's total 2 CU. That is why one Planner alone nearly fills the smallest SKU, and why F4 is the realistic floor for production planning.
The calculator adds up sustained CU across your users and recommends the F-SKU that fits.
Why does the calculator show automation jobs as a CU-equivalent instead of sustained CU?
Because they are billed differently. Automation jobs bill transactionally, per successful job (2 CU-hours each), the same way other Fabric capabilities such as notebooks, pipelines, and Power BI refreshes are metered. Sustained CU is different: it is a continuously-held seat license for the full 30-day session, held by Viewer, Stakeholder, and Planner roles regardless of how often they act.
To size your F-SKU, the calculator still needs one number, so it converts total automation CU-hours over the 30-day period into a CU-equivalent: total CU-hours divided by the hours in the period, the same conversion that defines the sustained CU rates themselves. That equivalent is what shows up in the capacity tank and dollar breakdown alongside the roles, even though automation jobs are not held continuously the way a role's session is.
See the calculator's Automation Jobs section for what counts as a job and two concrete examples.
When does billing start, and how is my role set?
Billing starts when a session opens, and your role is set by what you do in it. Users start at the minimum, Viewer, when a session opens. The system upgrades them automatically as they take higher-privilege actions: data entry or write-back moves them to Stakeholder, and authoring, modeling, or admin actions move them to Planner. A user who holds multiple roles is billed once, at the highest role only. A role upgrade starts a new 30-day window at the higher role, and once a window is triggered it is a billing commitment: the full 30 days are incurred even if an admin closes the session early.
For what each role can do and who should hold it, see the roles page.
Someone opened a plan just to look. What did that cost?
A Viewer session: 0.05 CU held for 30 days, under $7 at US list rates. The peek is cheap. The upgrade is what to govern: entering one number turns that Viewer into a roughly $30 Stakeholder session, and touching the model opens a roughly $150 Planner session. Microsoft's own wording, "opens or meaningfully engages," leaves a bare peek slightly ambiguous, so treat any open as billable until told otherwise.
I'm a Stakeholder. Can I do unlimited Stakeholder things?
No cap. Within the 30-day session the role charge is fixed no matter how much you do: dozens of PowerTables, a hundred scenarios, thousands of edits, and the rate doesn't move. What scales with usage is the compute underneath, drawn from the Fabric capacity you already own; the seat itself stays fixed. That flat charge covers every artifact and unlimited activity across Planning Sheets, PowerTable Sheets, and Intelligence Sheets, for whatever role you hold.
I only entered a handful of numbers all month. Why did I pay for a full Stakeholder session?
You're billed for the window, not the volume. Once you engage, that 30-day window is committed at your role's rate, and ten entries cost the same as ten thousand. The same fixed charge that removes the ceiling also puts in a floor. The useful question isn't the bill, it's whether that person belongs in the plan at all: if they do, the session cost is a rounding error next to what their time costs; if they don't, that's a governance question, not a billing one.
Can we lower a bill by closing sessions or downgrading roles?
No. The window runs out either way, at the rate it opened at. Cost control happens before the click, by governing who gets access to which role in the first place, not by managing sessions after they've started.
Do I pay when no one is using Fabric Planning?
No. Billing is based entirely on sustained CU, and sustained CU only starts when there's activity. The first time someone acts, that sets their role and starts a 30-day billing window for it. If nobody opens a plan item, no role is triggered and there is no charge. No activity, no cost.
The one nuance is automation: a background automation job that fires in an otherwise idle month still bills its small per-job draw (see the Automation Jobs section). No sessions and no jobs means no charge.
What happens if we exceed our reserved capacity?
The same thing that happens on pay-as-you-go: both capacity types hard-cap at the CU limit, and both offer overage billing as a toggleable setting. With overage on, additional CU-seconds are charged at 3x the pay-as-you-go rate. With overage off, the capacity throttles. Reserved and PAYG behave identically here; the difference between them is price and commitment, and the way to avoid the question entirely is to size with headroom (see the calculator's buffer setting).
We sometimes end up restarting the capacity. What happens to the sustained CU we already paid for?
Pausing, resizing, or deleting a capacity closes the books on every open session: the remaining days of each 30-day window bill out immediately instead of burning down over time. A restart is a pause plus a start, so the true-up rides along, but it is not a second charge. When the capacity comes back, users burn down the hours already billed, and no new session opens until the original window would have ended. A restart accelerates the bill; it does not grow it. Confirmed directly by the product team; this true-up billing behavior is rolling out in early August 2026.
We run more than one Fabric capacity. Does the same user bill on both?
Yes. Sessions key on the user plus the capacity, so one person active on two capacities opens two separate sessions, one on each. Confirmed directly by the product team.
Will Fabric Planning starve the capacity our reports already run on?
They share the same pool, so size for the planning peak on top of your existing peak rather than as a separate allocation. The calculator's buffer setting is one way to build that headroom in deliberately.
How does Fabric Planning pricing compare to traditional EPM tools?
Traditional EPM tools charge annual per-seat licenses, and a viewer seat often costs nearly as much as a builder seat, regardless of how much anyone uses the tool. Fabric Planning is consumption-based: you pay for active use only, and the Viewer and Stakeholder tiers are priced deliberately low to make broad participation cheap. Think of it as SaaS within SaaS: the planning workload draws from the same Fabric capacity you already run analytics on, and unused capacity stays available to the rest of Fabric. The comparison favors Fabric Planning most for seasonal cycles with many occasional participants, which is what most planning processes look like. Put your own numbers in the calculator to see where you land.
Are there additional costs for AI or Copilot features?
A fair question with an honest answer: we have not confirmed whether Copilot and AI features in Fabric Planning draw from the same CU pool or bill separately. Microsoft's billing post does not address it. Check Lumel's pricing page for the latest before committing, and treat AI usage as a line item to verify during a POC.
Have a question?
Ask it on Substack or find me on LinkedIn — I read and answer both.