Fabric and Azure

Does Copilot Work on a Microsoft Fabric Trial?

What actually works on a Microsoft Fabric trial, what it holds back, and what it costs to turn Copilot on. Tested in a live tenant.

Austin Levine

·

Updated

Summary of the Microsoft Fabric trial. Works: all five Fabric workloads, lakehouses and warehouses, pipelines and notebooks, 1 TB of OneLake storage, free-license report viewers. Blocked: Copilot in every workload, Fabric data agent, AI functions, AI services, trusted workspace access. The trial is 64 capacity units for 60 days on SKU FTL64.

What a Fabric trial gives you, and what it holds back.

Quick answer

Copilot does not work on a Fabric trial capacity. Not in Power BI, not in pipelines, not in notebooks. It requires a paid capacity, and Microsoft's documentation is explicit that "trial capacities and free SKUs aren't supported." SKU is the product code Microsoft gives each capacity size, such as F64.

Three things follow from that:

  • It is not a setting you missed. Every Copilot switch in this tenant was already on. The block is the capacity type, and nothing in the admin portal changes it.

  • The trial is still worth running. Everything in Fabric that is not AI works, and that is most of it.

  • There is a way to get Copilot without leaving the trial. One small paid capacity, used only for billing. Covered below.

Tested on a live Fabric trial capacity, SKU FTL64, with every Copilot tenant setting read through the Fabric admin API rather than by eye.

Why a trial blocks it, when a Pro license does not

If you come from Power BI, you are used to software sold per person. You buy a Pro license for each person who needs one, and that is the whole model.

Fabric is sold differently. You rent an engine, called a capacity, and everything your organization does in Fabric runs on it. Capacities come in sizes: F2 is the smallest, then F4, F8, and up through F64 and beyond. You pay by the second for the size you pick, whether one person uses it or a hundred.

Copilot's cost is charged to that engine, not to a person. Every question you ask Copilot burns a measurable amount of the capacity you are renting. That is why no per-person license gets you Copilot. A Power BI Pro license is not enough. Premium Per User is not enough either. Microsoft says so directly: "A Power BI Pro or Premium Per User (PPU) license alone isn't sufficient."

The free trial gives you an engine, and a good one, typically equivalent to a 64-unit capacity, though Microsoft now provisions some tenants at only 4 units. Microsoft simply excludes it from the AI. You are renting the engine for free, so they do not let you spend it on the expensive part.

Microsoft 365 Copilot is a different product

Worth clearing up before anything else, because it causes more confusion than the rest of this article combined. Microsoft 365 Copilot answering questions from Power BI data is not Copilot in Fabric. It runs on a per-person Microsoft 365 Copilot seat, needs no Fabric capacity at all, and is on by default. Two products, two bills, one name.

It is not your settings, and it is not your admin

The first thing anyone does when Copilot does not appear is ask IT whether it has been turned off.

In the tenant these screenshots come from, every relevant setting is on.

Fabric admin portal tenant settings, Copilot and AI section, with the main Copilot setting highlighted and reading Enabled for the entire organization

Fabric admin portal > Tenant settings > Copilot and AI. The main Copilot setting reads "Enabled for the entire organization."

With that setting on, opening Copilot from the Fabric navigation gives you this.

Power BI standalone Copilot showing a padlock icon and the message Copilot isn't available in your workspaces, with the explanation highlighted

Power BI standalone Copilot on a trial capacity. "Copilot isn't available in your workspaces. None of your workspaces have the required capacity for Copilot. To get started, contact an admin."

It tells you to contact an admin. In this tenant the person reading that message is the Fabric administrator, the capacity administrator and the global administrator, and every switch is already on. There is no admin to contact and nothing left to enable.

The same block appears three different ways

Here is the same single cause, presented three ways in three parts of the product.

In a Power BI report, it blames your workspace.

Power BI report ribbon with the grayed Copilot button highlighted and a tooltip reading Copilot isn't supported in this workspace

Power BI service, report in a trial workspace. Tooltip: "Copilot isn't supported in this workspace. Try using it in another workspace or check Power BI documentation for details on compatible workspaces."

The advice is to try a different workspace. On a trial, every workspace you can create sits on the same trial capacity, so every one of them fails identically. Nothing mentions the capacity or the trial, so the natural reading is that you set this workspace up wrong.

In a Data Factory pipeline, it says nothing at all.

Fabric data pipeline start screen with the Need help? Ask Copilot link highlighted

Fabric Data Factory, new data pipeline. The "Need help? Ask Copilot" prompt appears on the start screen. Clicking it, and clicking the Copilot button in the ribbon, produces nothing.

Fabric offers Copilot in the middle of the screen, on a capacity where it cannot run. Click it and nothing happens. No pane, no message, nothing logged. That silence is by design rather than a local fault: Microsoft's own Data Factory Copilot page states that when it cannot run, "it doesn't give an error message either."

In a notebook, it finally names the cause.

Fabric notebook ribbon with the grayed Copilot button highlighted and a tooltip reading Your Fabric license SKU does not allow access to copilot

Fabric notebook. Tooltip: "Your Fabric license SKU does not allow access to copilot (preview)."

Two AI features let you start, then fail later

Microsoft's trial documentation says AI experiences "such as Data agent, AI functions and AI services aren't supported" on a trial. The product does not stop you from starting them.

Data agents appear in the create menu like any other item.

Fabric New item panel filtered to agent, showing Data agent and Operations agent as creatable items, with Data agent highlighted

Fabric New item panel on a trial capacity. Data agent and Operations agent are both offered, described normally, with no notice that a trial capacity cannot run them.

You can click straight through, name it, and only then find out.

Fabric Create data agent dialog showing a red error reading Can't create data agents in this workspace, an admin needs to change the SKU type for your organization's Fabric capacity, with the error highlighted

Create data agent dialog on a trial capacity. "Can't create data agents in this workspace: An admin needs to change the SKU type for your organization's Fabric capacity."

"An admin needs to change the SKU type" means "somebody has to buy a paid capacity." If you are the admin, as you usually are on a trial, there is nothing for you to change.

Creating the same item through the Fabric REST API drops the euphemism:

{
  "errorCode": "UnsupportedCapacitySKU",
  "message": "FTL64 SKU Not Supported"
}

AI functions open a full setup pane.

Fabric notebook AI tools pane offering custom prompt, detect sentiment, categorize text and summarize text, all highlighted

Fabric notebook, AI tools > Set up an AI function, on a trial capacity. The pane opens and offers four AI functions. The code it generates needs a paid capacity to run.

In both cases the feature is discoverable, configurable, and will let you invest time in it. The failure is deferred until you try to run it. Evaluating against a deadline, that is the part that costs you a day.

What the trial is genuinely good for

None of the above makes the trial a waste. It is 60 days of a 64-unit capacity, free, and it covers almost everything that is not AI.

A data pipeline, a notebook and a lakehouse all create and open on a trial capacity without complaint.

Fabric workspace list showing a lakehouse, SQL analytics endpoint, notebook and pipeline created on the trial capacity, all highlighted

A workspace on a trial capacity after creating three items: a lakehouse with its SQL analytics endpoint, a notebook and a pipeline, all working.

One benefit is easy to miss, and it matters when you decide what to buy. A trial capacity behaves like an F64 for viewing, so colleagues holding only a free license can open your Power BI reports. On any paid capacity below F64 they cannot: every viewer needs Pro or PPU.

Microsoft now provisions trials at either 4 or 64 capacity units depending on the tenant, and a 4 can often be raised under Admin portal > Capacity settings > Trial > Change size.

Inviting a colleague onto the trial added 7 days to it. The counter read 59 days left, we invited someone onto the capacity, and it read 66. No Microsoft page documents this. The trial page names a sales representative as the only route to an extension, and it explicitly states that resizing the capacity does not extend the clock, so this is not that. Undocumented behavior can be withdrawn without notice, so take the extra week and do not plan around getting it.

Fabric admin portal Trial tab, with the 60-day trial description highlighted and the Days left column reading 66 also highlighted

Fabric admin portal > Capacity settings > Trial. The description says "a 60-day free Fabric trial". The Days left column says 66, on a capacity one day old.

What you can and cannot do on a Fabric trial

AI features

Feature

On a trial

Notes

Copilot in Power BI reports, DAX and summaries

No

Button visible, permanently grayed

Copilot in Data Factory pipelines and dataflows

No

Offered on screen, does nothing when clicked

Copilot in notebooks

No

Names the license SKU as the cause

Copilot in Data Warehouse and SQL database

No

Same capacity requirement

Copilot in Real-Time Intelligence and KQL

No

Same capacity requirement

Fabric data agent

No

Create fails at the API

AI functions

No

Setup pane opens, generated code cannot run

AI services and Foundry tools

No

Excluded from trial capacity

Microsoft 365 Copilot over Power BI data

Yes

Different product, per-user seat, no capacity needed

Building and shipping

Feature

On a trial

Notes

Lakehouses, warehouses, notebooks, pipelines, eventhouses

Yes

All five workloads included

OneLake storage

Yes

Up to 1 TB

Workspace collaboration and sharing

Yes

Invite colleagues into the trial capacity

Capacity Metrics app

Yes

Useful for sizing the capacity you will actually buy

Deployment pipelines and Git integration

Yes

Full application lifecycle, dev to production

Direct Lake semantic models

Yes

Query lakehouse data without importing it

Large models, 48 refreshes a day, XMLA endpoint

Yes

Via the Power BI individual trial included with it

Free-license colleagues viewing your reports

Yes

Trial behaves like an F64. A paid F2 does not

Platform and security

Feature

On a trial

Notes

Managed private endpoints

Yes

Outbound connections from Fabric to your own network

Trusted workspace access

No

Blocks the usual pattern for reaching firewalled storage

Private Link

No

Disabled on trial capacity

Pause and resume the capacity

No

Paid F SKUs only

Autoscale and Spark autoscale billing

No

Fixed at the trial's capacity units

ARM APIs and Terraform

No

No infrastructure-as-code against a trial capacity

Customer-managed keys

No

Microsoft-managed encryption only

Production workloads

No

Microsoft states the trial is for evaluation and testing only

What happens when the 60 days run out

Worth knowing before you build anything substantial on it, because the answer is not "nothing".

When the trial expires, access to the trial capacity is revoked and any workspace assigned to it reverts to Power BI Pro. Your Power BI content survives that. Everything that is not a Power BI item does not. Lakehouses, warehouses, notebooks and pipelines remain in the tenant but become inactive. You then have seven days to recover them by assigning the workspace to a paid F or P capacity. After that window, Microsoft's retention policy applies.

Do not assume you can simply start again either. Microsoft states that eligibility for a second trial is not guaranteed and that repeated or successive trial usage may be restricted.

So a trial is a good place to prove an architecture, and a bad place to build anything you will still need after it expires.

How to get Copilot without leaving the trial

This is the part most people miss, and it is the reason the answer to the title question is not simply "pay up".

The smallest capacity that supports Copilot is an F2. Azure's retail pricing API returns €0.1889 per capacity unit per hour for Germany West Central in EUR, which puts an F2 at €0.38 an hour, or about €276 a month left running. Check your own region and currency, because both change the number. Paid capacities can be paused and a paused capacity costs only storage, so a few days of evaluation is single-digit euros rather than a monthly commitment.

An F2 sounds far too small to serve AI, and that is the objection worth answering, because Copilot has a reputation for needing an F64. Copilot requests are billed as background jobs and smoothed over 24 hours, not over the five-minute window interactive operations use. Microsoft's own worked example puts an F64 at "over 13,824 Copilot requests per day". An F2 is a thirty-second of that capacity, which still leaves a few hundred requests a day. For one person evaluating whether Copilot is any good, that is not the binding constraint. Sustained team use is a different question and a different SKU.

One number to keep an eye on: capacity overage bills at €0.5667 per CU-hour in the same region, three times the base rate. A small capacity is cheap until you burst past it.

The obvious move is to put your workspace on that F2. The obvious move is also a downgrade. Your trial behaves like an F64 for viewing; an F2 does not. The moment your workspace moves to an F2, every person who opens those reports needs a Pro or PPU license. The cheapest way to switch Copilot on is also the thing that stops your colleagues reading your reports.

There is an alternative. A Fabric Copilot capacity is one you designate purely to collect Copilot usage for a named group of people, wherever their content actually lives. Microsoft's specification lists Trial among the workspace types those people can then use Copilot on in Power BI. Read straight, that means one small F2 doing nothing but paying for Copilot, while your work stays on the free trial capacity.

Four things to know before you plan around it.

  1. Two Microsoft pages disagree, and you will find that in one click. The Fabric Copilot capacity page lists Trial among the supported Power BI license modes. The Fabric trial page says flatly that "Copilot and Trusted Workspace Access aren't supported." Neither page acknowledges the other. The Copilot capacity page is the more specific one and is the reason this route is worth trying, but nobody should pretend the documentation settles it.

  2. Trial is listed for Power BI only. Not for pipelines, notebooks or the warehouse. Do not assume those come with it.

  3. AI functions are excluded outright, whatever else you set up.

  4. The capacity must sit in your tenant's home region, and a user can belong to only one Copilot capacity at a time.

This is the one recommendation in this article that comes from Microsoft's documentation rather than from our own tenant. Everything else here was clicked. Given that two Learn pages contradict each other on exactly this point, test it on your own tenant before you build a plan on it.

One more thing catches people out. Buying a capacity does not give you Copilot immediately. Microsoft states that when you purchase or resize a capacity, "it can take up to 24 hours for Copilot to recognize the change." Do not book the demo for the same afternoon.

The bottom line

A Fabric trial answers "is this the right platform for us" very well, and cannot answer "will Copilot make my team faster" at all. Those are different questions and the trial only serves one of them.

If Copilot is the actual question, do not spend two weeks of trial time trying to make the button light up. Buy an F2 and designate it as a Fabric Copilot capacity. Assign yourself to it, wait out the 24-hour recognition window, and keep your work where it is on the trial. That is a few euros and an afternoon, and it is the only way to find out whether the thing behind the button is worth a monthly capacity bill.

If the platform is the question, use the 60 days properly: build the lakehouse, wire up the pipelines, put a Direct Lake model in front of real data, and watch the Capacity Metrics app while you do it. That tells you which F SKU you would actually need, which is the number the whole decision turns on.

Sources

Microsoft ships changes to Fabric monthly and settings pages move. Everything here was observed in a live tenant when this was published, and is worth re-checking before you rely on it.

Summary of the Microsoft Fabric trial. Works: all five Fabric workloads, lakehouses and warehouses, pipelines and notebooks, 1 TB of OneLake storage, free-license report viewers. Blocked: Copilot in every workload, Fabric data agent, AI functions, AI services, trusted workspace access. The trial is 64 capacity units for 60 days on SKU FTL64.

What a Fabric trial gives you, and what it holds back.

Quick answer

Copilot does not work on a Fabric trial capacity. Not in Power BI, not in pipelines, not in notebooks. It requires a paid capacity, and Microsoft's documentation is explicit that "trial capacities and free SKUs aren't supported." SKU is the product code Microsoft gives each capacity size, such as F64.

Three things follow from that:

  • It is not a setting you missed. Every Copilot switch in this tenant was already on. The block is the capacity type, and nothing in the admin portal changes it.

  • The trial is still worth running. Everything in Fabric that is not AI works, and that is most of it.

  • There is a way to get Copilot without leaving the trial. One small paid capacity, used only for billing. Covered below.

Tested on a live Fabric trial capacity, SKU FTL64, with every Copilot tenant setting read through the Fabric admin API rather than by eye.

Why a trial blocks it, when a Pro license does not

If you come from Power BI, you are used to software sold per person. You buy a Pro license for each person who needs one, and that is the whole model.

Fabric is sold differently. You rent an engine, called a capacity, and everything your organization does in Fabric runs on it. Capacities come in sizes: F2 is the smallest, then F4, F8, and up through F64 and beyond. You pay by the second for the size you pick, whether one person uses it or a hundred.

Copilot's cost is charged to that engine, not to a person. Every question you ask Copilot burns a measurable amount of the capacity you are renting. That is why no per-person license gets you Copilot. A Power BI Pro license is not enough. Premium Per User is not enough either. Microsoft says so directly: "A Power BI Pro or Premium Per User (PPU) license alone isn't sufficient."

The free trial gives you an engine, and a good one, typically equivalent to a 64-unit capacity, though Microsoft now provisions some tenants at only 4 units. Microsoft simply excludes it from the AI. You are renting the engine for free, so they do not let you spend it on the expensive part.

Microsoft 365 Copilot is a different product

Worth clearing up before anything else, because it causes more confusion than the rest of this article combined. Microsoft 365 Copilot answering questions from Power BI data is not Copilot in Fabric. It runs on a per-person Microsoft 365 Copilot seat, needs no Fabric capacity at all, and is on by default. Two products, two bills, one name.

It is not your settings, and it is not your admin

The first thing anyone does when Copilot does not appear is ask IT whether it has been turned off.

In the tenant these screenshots come from, every relevant setting is on.

Fabric admin portal tenant settings, Copilot and AI section, with the main Copilot setting highlighted and reading Enabled for the entire organization

Fabric admin portal > Tenant settings > Copilot and AI. The main Copilot setting reads "Enabled for the entire organization."

With that setting on, opening Copilot from the Fabric navigation gives you this.

Power BI standalone Copilot showing a padlock icon and the message Copilot isn't available in your workspaces, with the explanation highlighted

Power BI standalone Copilot on a trial capacity. "Copilot isn't available in your workspaces. None of your workspaces have the required capacity for Copilot. To get started, contact an admin."

It tells you to contact an admin. In this tenant the person reading that message is the Fabric administrator, the capacity administrator and the global administrator, and every switch is already on. There is no admin to contact and nothing left to enable.

The same block appears three different ways

Here is the same single cause, presented three ways in three parts of the product.

In a Power BI report, it blames your workspace.

Power BI report ribbon with the grayed Copilot button highlighted and a tooltip reading Copilot isn't supported in this workspace

Power BI service, report in a trial workspace. Tooltip: "Copilot isn't supported in this workspace. Try using it in another workspace or check Power BI documentation for details on compatible workspaces."

The advice is to try a different workspace. On a trial, every workspace you can create sits on the same trial capacity, so every one of them fails identically. Nothing mentions the capacity or the trial, so the natural reading is that you set this workspace up wrong.

In a Data Factory pipeline, it says nothing at all.

Fabric data pipeline start screen with the Need help? Ask Copilot link highlighted

Fabric Data Factory, new data pipeline. The "Need help? Ask Copilot" prompt appears on the start screen. Clicking it, and clicking the Copilot button in the ribbon, produces nothing.

Fabric offers Copilot in the middle of the screen, on a capacity where it cannot run. Click it and nothing happens. No pane, no message, nothing logged. That silence is by design rather than a local fault: Microsoft's own Data Factory Copilot page states that when it cannot run, "it doesn't give an error message either."

In a notebook, it finally names the cause.

Fabric notebook ribbon with the grayed Copilot button highlighted and a tooltip reading Your Fabric license SKU does not allow access to copilot

Fabric notebook. Tooltip: "Your Fabric license SKU does not allow access to copilot (preview)."

Two AI features let you start, then fail later

Microsoft's trial documentation says AI experiences "such as Data agent, AI functions and AI services aren't supported" on a trial. The product does not stop you from starting them.

Data agents appear in the create menu like any other item.

Fabric New item panel filtered to agent, showing Data agent and Operations agent as creatable items, with Data agent highlighted

Fabric New item panel on a trial capacity. Data agent and Operations agent are both offered, described normally, with no notice that a trial capacity cannot run them.

You can click straight through, name it, and only then find out.

Fabric Create data agent dialog showing a red error reading Can't create data agents in this workspace, an admin needs to change the SKU type for your organization's Fabric capacity, with the error highlighted

Create data agent dialog on a trial capacity. "Can't create data agents in this workspace: An admin needs to change the SKU type for your organization's Fabric capacity."

"An admin needs to change the SKU type" means "somebody has to buy a paid capacity." If you are the admin, as you usually are on a trial, there is nothing for you to change.

Creating the same item through the Fabric REST API drops the euphemism:

{
  "errorCode": "UnsupportedCapacitySKU",
  "message": "FTL64 SKU Not Supported"
}

AI functions open a full setup pane.

Fabric notebook AI tools pane offering custom prompt, detect sentiment, categorize text and summarize text, all highlighted

Fabric notebook, AI tools > Set up an AI function, on a trial capacity. The pane opens and offers four AI functions. The code it generates needs a paid capacity to run.

In both cases the feature is discoverable, configurable, and will let you invest time in it. The failure is deferred until you try to run it. Evaluating against a deadline, that is the part that costs you a day.

What the trial is genuinely good for

None of the above makes the trial a waste. It is 60 days of a 64-unit capacity, free, and it covers almost everything that is not AI.

A data pipeline, a notebook and a lakehouse all create and open on a trial capacity without complaint.

Fabric workspace list showing a lakehouse, SQL analytics endpoint, notebook and pipeline created on the trial capacity, all highlighted

A workspace on a trial capacity after creating three items: a lakehouse with its SQL analytics endpoint, a notebook and a pipeline, all working.

One benefit is easy to miss, and it matters when you decide what to buy. A trial capacity behaves like an F64 for viewing, so colleagues holding only a free license can open your Power BI reports. On any paid capacity below F64 they cannot: every viewer needs Pro or PPU.

Microsoft now provisions trials at either 4 or 64 capacity units depending on the tenant, and a 4 can often be raised under Admin portal > Capacity settings > Trial > Change size.

Inviting a colleague onto the trial added 7 days to it. The counter read 59 days left, we invited someone onto the capacity, and it read 66. No Microsoft page documents this. The trial page names a sales representative as the only route to an extension, and it explicitly states that resizing the capacity does not extend the clock, so this is not that. Undocumented behavior can be withdrawn without notice, so take the extra week and do not plan around getting it.

Fabric admin portal Trial tab, with the 60-day trial description highlighted and the Days left column reading 66 also highlighted

Fabric admin portal > Capacity settings > Trial. The description says "a 60-day free Fabric trial". The Days left column says 66, on a capacity one day old.

What you can and cannot do on a Fabric trial

AI features

Feature

On a trial

Notes

Copilot in Power BI reports, DAX and summaries

No

Button visible, permanently grayed

Copilot in Data Factory pipelines and dataflows

No

Offered on screen, does nothing when clicked

Copilot in notebooks

No

Names the license SKU as the cause

Copilot in Data Warehouse and SQL database

No

Same capacity requirement

Copilot in Real-Time Intelligence and KQL

No

Same capacity requirement

Fabric data agent

No

Create fails at the API

AI functions

No

Setup pane opens, generated code cannot run

AI services and Foundry tools

No

Excluded from trial capacity

Microsoft 365 Copilot over Power BI data

Yes

Different product, per-user seat, no capacity needed

Building and shipping

Feature

On a trial

Notes

Lakehouses, warehouses, notebooks, pipelines, eventhouses

Yes

All five workloads included

OneLake storage

Yes

Up to 1 TB

Workspace collaboration and sharing

Yes

Invite colleagues into the trial capacity

Capacity Metrics app

Yes

Useful for sizing the capacity you will actually buy

Deployment pipelines and Git integration

Yes

Full application lifecycle, dev to production

Direct Lake semantic models

Yes

Query lakehouse data without importing it

Large models, 48 refreshes a day, XMLA endpoint

Yes

Via the Power BI individual trial included with it

Free-license colleagues viewing your reports

Yes

Trial behaves like an F64. A paid F2 does not

Platform and security

Feature

On a trial

Notes

Managed private endpoints

Yes

Outbound connections from Fabric to your own network

Trusted workspace access

No

Blocks the usual pattern for reaching firewalled storage

Private Link

No

Disabled on trial capacity

Pause and resume the capacity

No

Paid F SKUs only

Autoscale and Spark autoscale billing

No

Fixed at the trial's capacity units

ARM APIs and Terraform

No

No infrastructure-as-code against a trial capacity

Customer-managed keys

No

Microsoft-managed encryption only

Production workloads

No

Microsoft states the trial is for evaluation and testing only

What happens when the 60 days run out

Worth knowing before you build anything substantial on it, because the answer is not "nothing".

When the trial expires, access to the trial capacity is revoked and any workspace assigned to it reverts to Power BI Pro. Your Power BI content survives that. Everything that is not a Power BI item does not. Lakehouses, warehouses, notebooks and pipelines remain in the tenant but become inactive. You then have seven days to recover them by assigning the workspace to a paid F or P capacity. After that window, Microsoft's retention policy applies.

Do not assume you can simply start again either. Microsoft states that eligibility for a second trial is not guaranteed and that repeated or successive trial usage may be restricted.

So a trial is a good place to prove an architecture, and a bad place to build anything you will still need after it expires.

How to get Copilot without leaving the trial

This is the part most people miss, and it is the reason the answer to the title question is not simply "pay up".

The smallest capacity that supports Copilot is an F2. Azure's retail pricing API returns €0.1889 per capacity unit per hour for Germany West Central in EUR, which puts an F2 at €0.38 an hour, or about €276 a month left running. Check your own region and currency, because both change the number. Paid capacities can be paused and a paused capacity costs only storage, so a few days of evaluation is single-digit euros rather than a monthly commitment.

An F2 sounds far too small to serve AI, and that is the objection worth answering, because Copilot has a reputation for needing an F64. Copilot requests are billed as background jobs and smoothed over 24 hours, not over the five-minute window interactive operations use. Microsoft's own worked example puts an F64 at "over 13,824 Copilot requests per day". An F2 is a thirty-second of that capacity, which still leaves a few hundred requests a day. For one person evaluating whether Copilot is any good, that is not the binding constraint. Sustained team use is a different question and a different SKU.

One number to keep an eye on: capacity overage bills at €0.5667 per CU-hour in the same region, three times the base rate. A small capacity is cheap until you burst past it.

The obvious move is to put your workspace on that F2. The obvious move is also a downgrade. Your trial behaves like an F64 for viewing; an F2 does not. The moment your workspace moves to an F2, every person who opens those reports needs a Pro or PPU license. The cheapest way to switch Copilot on is also the thing that stops your colleagues reading your reports.

There is an alternative. A Fabric Copilot capacity is one you designate purely to collect Copilot usage for a named group of people, wherever their content actually lives. Microsoft's specification lists Trial among the workspace types those people can then use Copilot on in Power BI. Read straight, that means one small F2 doing nothing but paying for Copilot, while your work stays on the free trial capacity.

Four things to know before you plan around it.

  1. Two Microsoft pages disagree, and you will find that in one click. The Fabric Copilot capacity page lists Trial among the supported Power BI license modes. The Fabric trial page says flatly that "Copilot and Trusted Workspace Access aren't supported." Neither page acknowledges the other. The Copilot capacity page is the more specific one and is the reason this route is worth trying, but nobody should pretend the documentation settles it.

  2. Trial is listed for Power BI only. Not for pipelines, notebooks or the warehouse. Do not assume those come with it.

  3. AI functions are excluded outright, whatever else you set up.

  4. The capacity must sit in your tenant's home region, and a user can belong to only one Copilot capacity at a time.

This is the one recommendation in this article that comes from Microsoft's documentation rather than from our own tenant. Everything else here was clicked. Given that two Learn pages contradict each other on exactly this point, test it on your own tenant before you build a plan on it.

One more thing catches people out. Buying a capacity does not give you Copilot immediately. Microsoft states that when you purchase or resize a capacity, "it can take up to 24 hours for Copilot to recognize the change." Do not book the demo for the same afternoon.

The bottom line

A Fabric trial answers "is this the right platform for us" very well, and cannot answer "will Copilot make my team faster" at all. Those are different questions and the trial only serves one of them.

If Copilot is the actual question, do not spend two weeks of trial time trying to make the button light up. Buy an F2 and designate it as a Fabric Copilot capacity. Assign yourself to it, wait out the 24-hour recognition window, and keep your work where it is on the trial. That is a few euros and an afternoon, and it is the only way to find out whether the thing behind the button is worth a monthly capacity bill.

If the platform is the question, use the 60 days properly: build the lakehouse, wire up the pipelines, put a Direct Lake model in front of real data, and watch the Capacity Metrics app while you do it. That tells you which F SKU you would actually need, which is the number the whole decision turns on.

Sources

Microsoft ships changes to Fabric monthly and settings pages move. Everything here was observed in a live tenant when this was published, and is worth re-checking before you rely on it.

Show us your report nobody trusts.

1 · A 30-minute call.

2 · We look at your current reports together.

3 · We tell you what we would do.

Show us your report nobody trusts.

1 · A 30-minute call.

2 · We look at your current reports together.

3 · We tell you what we would do.

Show us your report nobody trusts.

1 · A 30-minute call.

2 · We look at your current reports together.

3 · We tell you what we would do.

CaseWhen is a Berlin BI consultancy that builds reporting that leaders can trust, on the Microsoft stack: Power BI, Fabric and Azure.

Berlin, Germany

© CaseWhen Consulting GmbH

English

CaseWhen is a Berlin BI consultancy that builds reporting that leaders can trust, on the Microsoft stack: Power BI, Fabric and Azure.

Berlin, Germany

© CaseWhen Consulting GmbH

English

CaseWhen is a Berlin BI consultancy that builds reporting that leaders can trust, on the Microsoft stack: Power BI, Fabric and Azure.

Berlin, Germany

© CaseWhen Consulting GmbH

English