Copilot in Power BI requirements: F2 or P1 capacity, licenses and regions
Copilot in Power BI runs only when the work is tied to paid organizational capacity: a Microsoft Fabric capacity of F2 or higher, a Power BI Premium capacity of P1 or higher, or a Fabric Copilot capacity that an administrator assigns to a group of users. Microsoft's Power BI documentation, checked on November 8, 2026, states that a Power BI Pro or Premium Per User (PPU) license on its own is not enough, and that trial capacities and free SKUs are not supported.
The per-user side still applies on top of the capacity. Report authors need a Pro or PPU license to publish, and on capacities smaller than F64 every person who views Power BI content also needs Pro, PPU or an individual trial. From F64 upward, users with a Free license and a viewer role can open the content. For a small team, the cheapest path to Copilot is therefore an F2 capacity plus Pro licenses for everyone, and the illustrative example below puts that at $1,382.80 a month for 80 users at East US pay-as-you-go rates.
Region and tenant settings decide the rest. The capacity has to sit in a Fabric region that supports Copilot, the tenant setting for Azure OpenAI features must be on at tenant level, and tenants outside the US or EU data boundary need an extra cross-geo processing switch before Copilot works at all.
··········
COPILOT IN POWER BI CAPACITY AND LICENSE REQUIREMENTS.
Microsoft's "Copilot for Power BI overview" page, last updated in August 2026, lists the requirements that apply to every Copilot experience. The capacity rule is the one most teams trip over: the report or semantic model has to live in a workspace assigned to paid Fabric capacity (F2 or higher) or Power BI Premium (P1 or higher). The page adds that a Pro or PPU license alone is not sufficient because Copilot requires organizational capacity, and the companion page on enabling Copilot states that Copilot in Microsoft Fabric is not supported on trial SKUs or trial capacities.
The table below summarizes what each piece contributes, as documented by Microsoft on November 8, 2026.
........
Item | Role for Copilot in Power BI | Documented requirement |
Fabric capacity | Hosts content and runs Copilot | F2 or higher, paid, supported region |
Power BI Premium capacity | Hosts content and runs Copilot | P1 or higher |
Fabric Copilot capacity | Bills Copilot usage of assigned users | At least F2 or P1, tenant home region |
Power BI Pro | Publish and share content | $14.00 user/month, paid yearly |
Premium Per User | Premium features per user | $24.00 user/month, paid yearly |
Free license | View content on F64 or larger | Viewer role in the workspace |
Trial capacity | Not supported for Copilot | Paid SKUs only |
........
Prices in the table are the ones listed on Microsoft's Power BI pricing page on November 8, 2026; Microsoft notes there that prices are shown for marketing purposes and that checkout prices may differ. PPU is also available as a $14.00 user/month step-up for users who already have Pro, Microsoft 365 E5 or Office 365 E5.
The Fabric licensing page sets the viewer rule that drives most of the cost. On F SKUs smaller than F64, each user viewing Power BI content must have Pro, PPU or an individual trial. On F64 or larger, users with only a Free license and a viewer role can view Power BI content. The same page lists the capacity units (CU) of each SKU: F2 has 2 CU, F8 has 8 CU and F64 has 64 CU, which Microsoft compares in compute terms to P1. A guide to the licensing choice itself, with the Pro, PPU and F64 trade-offs, is in our article on Power BI Pro vs Premium Per User vs Fabric F64 capacity.
··········
WHICH COPILOT IN POWER BI EXPERIENCES NEED WHAT ACCESS.
Microsoft splits Copilot in Power BI into several experiences with different status and access rules. The Copilot pane inside a report is generally available. The standalone, full-screen Power BI Copilot reached from the left navigation, the app-scoped Copilot inside Power BI apps and Copilot in web modeling are in preview, according to the overview page.
........
Copilot experience | Status | Access the user needs |
Report Copilot pane (service) | Generally available | Read access for summaries, edit or build access to create content |
Report Copilot in Power BI Desktop | Documented | Write access to a workspace on paid Fabric or Premium capacity |
Standalone Power BI Copilot | Preview | Fabric Copilot capacity, or an eligible F2+ workspace chosen by the service |
App-scoped Copilot | Preview | Tenant-level setting on, app navigation option enabled |
Copilot in web modeling | Preview | Write access to the model, Editing mode |
........
In the Power BI service, summarizing a report needs at least read access, while adding a narrative visual or generating a Copilot report page needs edit access to the workspace or report, or build permission on the semantic model. In Power BI Desktop, the author needs admin, member or contributor access to at least one workspace on a paid Fabric capacity (F2 or higher) or Premium capacity (P1 or higher) with Copilot enabled, unless a Fabric Copilot capacity covers that user. Desktop asks the author to pick a Copilot-compatible workspace on first use, and that workspace is where usage is tracked.
The standalone experience behaves differently. If the organization has a Fabric Copilot capacity, it is used. Otherwise the service selects a workspace on an F2 or larger capacity in a supported region with Copilot enabled, and the user can change it later. If neither exists, the documentation says the user cannot use standalone Copilot. Microsoft also documents a prompt limit of 10,000 characters on all surfaces, including Power BI Mobile and Embedded, and a rolling 24-hour cache that returns the same answer for an identical prompt on an unchanged semantic model. Practical examples of the report pane at work, including DAX measures that Copilot drafts for a variance report, are in our guide to a Power BI budget vs actual dashboard with Copilot.
··········
FABRIC COPILOT CAPACITY FOR PRO AND PPU WORKSPACES.
The Fabric Copilot capacity is the documented route for teams whose content sits in Pro or PPU workspaces. Microsoft's page on Fabric Copilot capacity, updated on October 20, 2026, describes it as a Fabric capacity designated to collect and bill the Copilot usage of a specific group of users. When those users run Copilot or data agents, Fabric charges the usage to the Copilot capacity instead of to the capacity that holds their content.
Supported scenarios include Copilot in Power BI Desktop, Copilot in the Power BI service on a Pro or PPU workspace, and Copilot on a Fabric capacity workspace. Microsoft lists Pro, Trial, PPU, Premium capacity and Fabric capacity as workspace license modes that work with it, while capacities with the Embedded license mode are not supported. For other Fabric workloads and data agents, the Copilot capacity applies when the workspace capacity is smaller than F64.
The setup needs two roles. A Fabric administrator enables Copilot for the organization and authorizes capacity administrators to designate their capacity as a Copilot capacity. The capacity administrator then designates it and assigns a group of users. The rules are short: the capacity must be at least an F2 or P1 SKU, it is supported only in the tenant's home region, one Copilot capacity applies per user (the most recently created one wins if a user is assigned to several), and an assigned Copilot capacity takes precedence over the workspace's own capacity. Billing covers prompt processing and response generation only; downstream operations such as querying the semantic model are charged to the semantic model's capacity.
··········
COPILOT IN POWER BI TENANT SETTINGS AND REGIONS.
Copilot is controlled in the Fabric admin portal by the setting "Users can use Copilot and other features powered by Azure OpenAI." Microsoft documents it as enabled by default, and administrators can disable it or restrict it to security groups. The setting is automatically delegated to capacity administrators, who can override the tenant value for their capacity, but the overview page states that the standalone experience needs the setting on at tenant level and that enabling it only at capacity level is not sufficient.
Data location is the second gate. If the tenant or capacity is outside the US or EU data boundary, Copilot is disabled by default until an administrator enables the setting that allows data sent to Azure OpenAI to be processed outside the geographic region, compliance boundary or national cloud instance. The cross-geo setting is not delegated to capacity administrators. Microsoft's consumption page explains that Copilot runs on Azure OpenAI models deployed in a limited set of data centers and that billing follows the capacity unit price of the customer's billing region, wherever the GPU work happens.
........
Region or environment | Documented status |
Capacity in a supported Fabric region | Copilot available |
Tenant or capacity outside US or EU data boundary | Off until cross-geo setting is enabled |
India West, Indonesia Central, Korea South, Malaysia West | No standalone or app-scoped Copilot yet |
New Zealand North, Qatar Central, Taiwan North, Taiwan North West | No standalone or app-scoped Copilot yet |
UAE Central, France South, Germany North, Norway West | No standalone or app-scoped Copilot yet |
Sovereign clouds | Not supported |
Private Link or closed network environments | Not supported |
........
Microsoft also notes that prompts in languages other than English may work but multilingual use is not officially supported, and that after scaling up or buying capacity it can take up to 24 hours before Copilot recognizes the change.
··········
HOW COPILOT IN POWER BI CONSUMES FABRIC CAPACITY UNITS.
Copilot does not carry a separate per-user fee in Power BI; it consumes the capacity units of the capacity it bills to. Microsoft's page on Copilot consumption in Fabric lists the rates per 1,000 tokens: 100 CU seconds for input, 10 CU seconds for cached input and 400 CU seconds for output. Copilot operations are background jobs, so Fabric smooths their usage over 24 hours, and the first throttling phase starts when a capacity has consumed all its CU for the next 10 minutes.
The Azure retail price list for Microsoft Fabric showed $0.18 per CU hour, pay-as-you-go, in the East US region on November 8, 2026. Microsoft's Power BI pricing page states that an annual capacity reservation saves 40.5% over pay-as-you-go prices. Downstream work started by Copilot, such as DAX queries, refreshes and email subscriptions, is billed separately through the normal Fabric operations, and capacity administrators can see 30 days of Copilot usage in the Microsoft Fabric Capacity Metrics app by filtering on the Copilot in Fabric operation.
··········
WORKED COST EXAMPLE FOR COPILOT IN POWER BI ON F2 AND F64.
The following example is illustrative. It assumes a team of 20 report authors and 60 report viewers, a 730-hour month, the East US pay-as-you-go rate of $0.18 per CU hour, Pro at $14.00 per user per month, and Copilot requests of the size Microsoft uses in its own documentation example: 2,000 input tokens and 500 output tokens. Each author sends 15 Copilot requests per working day.
CU seconds per request = (2,000 x 100 + 500 x 400) / 1,000 = 400
Daily Copilot load = 20 authors x 15 requests x 400 = 120,000 CU seconds = 33.33 CU hours
F2 daily capacity = 2 CU x 24 hours = 48 CU hours
Share of F2 used = 33.33 / 48 = 69.4%........
Item | F2 option | F64 option |
Capacity, pay-as-you-go per month | $262.80 | $8,409.60 |
Pro licenses needed | 80 (authors and viewers) | 20 (authors only) |
Pro cost per month | $1,120.00 | $280.00 |
Total per month | $1,382.80 | $8,689.60 |
Copilot requests per day at 400 CU seconds, full capacity | 432 | 13,824 |
........
The F2 option is far cheaper for 80 users, but Copilot alone would use about 69.4% of its daily capacity units in this scenario, before any refreshes, report queries or other Fabric workloads on the same capacity. An F4 capacity at $525.60 per month would halve that share to 34.7%. Moving the authors' Copilot usage to a separate Fabric Copilot capacity keeps that load away from the capacity that serves reports. F64 becomes reasonable when the number of viewers is large enough that removing their Pro licenses covers the difference in capacity price; with 20 authors at the prices above, the two options cost the same at about 582 viewers, far above the 60 in this example. Microsoft's own documentation example reaches the same 13,824 requests per day for a fully dedicated F64.
··········
COMMON MISREADINGS OF COPILOT IN POWER BI LICENSING.
Several assumptions are contradicted by Microsoft's documentation. A PPU license does not by itself enable Copilot: the overview page says Pro or PPU alone is not sufficient, and the licensing page adds that PPU does not provision a Fabric capacity. A Fabric trial capacity does not cover Copilot either, since trial SKUs and trial capacities are excluded. Copilot in Power BI is also separate from Microsoft 365 Copilot licensing; the Power BI pages tie it to capacity, while the Microsoft 365 plans and seat licenses are covered in our overview of Microsoft 365 Copilot licenses and business access.
Turning Copilot on at capacity level does not switch on the standalone experience if the tenant-level setting is off, and buying a larger capacity does not take effect for Copilot immediately, since Microsoft allows up to 24 hours. Finally, a Fabric Copilot capacity in a region other than the tenant's home region is not supported, so organizations with capacities in several geographies need to plan where that capacity lives.
··········
COPILOT IN POWER BI: WHAT DEPENDS ON YOUR CAPACITY AND TENANT.
Microsoft's documentation confirms the core requirements: paid F2 or P1 capacity or a Fabric Copilot capacity, no trial capacities, the Azure OpenAI tenant setting turned on, a supported region, and the cross-geo setting for tenants outside the US and EU data boundary. The report Copilot pane is generally available, while the standalone, app-scoped and web modeling experiences remain in preview.
What a team pays depends on choices it controls. Below F64, every viewer needs Pro or PPU, so a small F2 capacity plus Pro licenses is the lowest documented entry point, as long as Copilot and the other workloads fit inside its capacity units. Above F64, Free viewers remove most per-user cost but the capacity price rises sharply. The actual CU consumption depends on prompt and answer length, which Microsoft does not fix per request, so the Capacity Metrics app is the place to measure it after a pilot. Microsoft does not publish a list of which countries fall outside its Copilot data boundary beyond the US and EU rule, so administrators outside those regions should check the cross-geo setting in their own admin portal before rolling Copilot out.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org
