Fabric vs Power BI Premium - Simply Explained

Welcome to the M365 FM podcast, where we explore the people, ideas and technologies shaping the future of work. The two names everyone mixes up. One of the two names in this title is being retired by Microsoft Power BI premium was the shared engine, an organization bought so hundreds of people could open the same report, and it came in sizes called P1 through P5. Microsoft Fabric is that same engine rebuilt as one platform for all of your data and its sizes. Now start with the letter F. Welcome to the M365 FM podcast, where we explore the people, ideas and technologies shaping the future of work. What trips people up is that a capacity and a license are two separate purchases, and one number decides whether viewing a report costs money or comes free. By the end, you'll be able to explain the whole thing to your team without the jargon. What power BI premium was built to do? Begin with the problem it solved. Everyone in a company wanted to look at the same handful of reports, and power BI Pro was priced per person, so a dashboard with four hundred regular readers turned into four hundred monthly bills. Power BI premium was Microsoft's answer, and it was a simple one. Instead of paying a license fee for every single reader, an organization bought one shared pool of computing power. The whole company then drew on that pool at once. That pool is called a capacity. It's shared computing power rather than a per person product. The sizes ran from P1 up to P5, with P1 as the smallest and P5 as the largest. Buying one worked differently from buying a license. A company committed to that capacity monthly or yearly. The bill arrived on a schedule whether the team used all of that power or not, and that trade off suited large organizations while it punished small ones. Premium was built for scale rather than for beginners. A company bought it when the number of readers made per person pricing painful, or when a data model had grown too large for the standard Pro limits. A bigger size also unlocked the heavier parts of power by larger data models could fit in memory. Scheduled refreshes could run many more times a day, and advanced third party tools could connect through something called the Xmla endpoint. There was an optional extra called Autoscale as well. It topped up the compute automatically whenever a capacity ran hot, then charged that extra usage back to the company's Azure subscription. Every bit of that power stayed inside power BI. So there was no place in it at all for pipelines, lakehouses, or data warehouses. One tool did one job and it did that job well. A second, smaller family called M existed for embedding, which is the job of placing a power BI report inside somebody else's application. That family sold separately and it never covered reporting for a company's own staff. Picture a sales dashboard that four hundred people open every Monday morning under a per person model that is four hundred licenses and under premium, it was one capacity plus a much smaller number of licenses for the people who actually built the report. Premium also changed. Who could read a report for free. A person holding only a free license could view content hosted on that capacity, which is exactly the rule that comes back later in this episode. What fabric is and why power BI now lives inside it. Fabric is the newer platform. It covers the whole journey. A piece of data takes from bringing it in to cleaning it up, storing it, analyzing it, and finally reporting on it inside a single product. Inside fabric, there are separate tools, one for each kind of work. Power BI handles. Reports. Data factory. Moves data in from other systems. Data engineering. Runs spark and notebooks. Data warehouse handles SQL data science. Builds models, and real time intelligence. Watches data as it arrives. Databases is the newest member of that group. It can mirror data continuously from systems like Azure, SQL database, Cosmosdb, Databricks, and Snowflake, which means an existing database can feed fabric without being rebuilt first. Underneath all of those tools sits one shared store called One Lake. Its job is to hold data in one place instead of copying it into every tool, and Microsoft compares the arrangement to the OneDrive that your office apps already use. One Lake also has a shape that is easy to picture. A tenant sits at the top. Workspaces sit inside the tenant, much like folders and items such as a lake house. Sit inside a workspace. Shortcuts let one lake reach into storage that already exists, including Amazon S3 and Google Cloud. Storage data doesn't have to be moved in first, which is what keeps the platform from becoming one more place to duplicate everything. Nobody has to set anything up before using it. A file is simply there when you open it. And that is the same idea one lake borrows for data across a whole company. Picture the same sales numbers moving from a raw export to a warehouse to a finished dashboard. Nobody makes a second copy along the way, and that is the main reason Microsoft calls fabric a platform rather than a single product. More than two hundred connectors come built in that range is the reason a team can pull data from the many systems it already runs, without writing any custom connection code at all. Copilot is built into the platform as well, and it helps with writing queries, building pipelines, and summarizing what the data says. It respects the same permissions a person already has, so it doesn't open a side door into anything. The limit that catches beginners comes in two halves. Fabric itself doesn't ask for an Azure account, but the capacity behind the heavy workloads is bought there, and those two facts sound like they contradict each other until you hold them side by side. That covers the platform, and a platform is only half the story. How anyone pays for it comes down to two purchases that people constantly blend into. One capacity and licenses are two separate purchases. This is the part that causes the most confusion. A capacity and a license are two different purchases, and people talk about them in the same breath as if they were the same thing. A capacity is the shared computing power, measured in units and named with an F and a number. So F2 is tiny and f sixty four is sixty four times its size. A license is a different animal. It describes what one person is allowed to do inside the platform, which is why the two ideas get tangled together so easily in conversation. Three licenses matter for this audience. Free is handed out automatically the first time a person signs into fabric if the tenant has it switched on. Pro is the one that lets you publish and share power BI reports with other people. The third is premium per user, usually shortened to poo. It hands one person most of the premium power BI capabilities without an organization wide capacity, and Microsoft's own guidance says it is the cheaper route below two hundred and fifty people. There is one catch, and it is easy to miss. Premium per user doesn't come with a capacity, so it can't run lakehouses warehouses or notebooks, and anyone who buys it expecting all of fabric will be disappointed. Inside a workspace, there are roles and the viewer role is the one that matters most for the rule. Coming up next. Admin member and contributor roles. Let a person change content and holding one of those still requires a paid license. Take a five person team that only builds and reads reports that team needs five licenses and no capacity at all, and the moment it wants to build a data pipeline, the answer changes completely. A working setup usually needs both purchases, which is the sentence worth remembering. The capacity provides the power and the license provides the permission, and neither one replaces the other. The two purchases also land on different desks. A capacity is bought by the organization, usually by whoever watches the cloud bill while a license is assigned to an individual. And those two jobs often belong to different people. The Ef64 line where viewing stops costing money. Here is the number that decides everything on a fabric capacity of ef64 or larger. A person with only a free license and the viewer role can open power BI reports, and the organization pays nothing extra for that reader. Below Ef64, the rule flips on F2 through F32. Every single person who wants to look at a report needs a pro or premium per user license, and that includes the people who never build anything building is separate from viewing and the distinction matters. Authors and editors need pro or premium per user, no matter how large the capacity is. So the Ef64 line only changes the rules for the audience, not for the people who make the reports. Capacity size also sets limits that have nothing to do with viewing. A model on F32 can use about ten gigabytes of memory, while the same model on Ef64 can reach twenty five. The amount of data in play can push a team across that line on its own. The mapping between the old and new sizes runs by computing power P1 maps to Ef64, P2 to F128, P3 to F two, fifty six, P4 to F five and twelve, and P5 to F ten twenty four, which means a team that already runs on P1 is looking squarely at f sixty four. Premium per user sits in a strange middle place. It raises one person's limits, including forty eight refreshes a day, without giving the organization a shared engine to build on. Here is the same rule in money. Using Microsoft's published list prices, three hundred viewers on a small capacity would need three hundred Pro licenses at fourteen dollars a person each month. That comes to four thousand two hundred dollars a month before any capacity is bought. Crossing that line has a price of its own. F sixty four lists at eight thousand four hundred and nine dollars sixty cents a month on pay as You go. A one or three year reservation brings that to about five thousand dollars a month, so free viewing only pays off once the audience is genuinely large. So there is the number, the opening held back. F sixty four is the point where viewers stop costing money one at a time, and below it, every reader turns into a monthly line item on the bill. Why power BI premium is being retired. Microsoft is simplifying how all of this gets bought. The result is that power BI premium per capacity is going away, and the P1 through P5 subscriptions can no longer be purchased. New existing customers can't add to them or renew them either, and that is the part that forces a decision. Each subscription retires at the end of its own agreement term, so there is no single date when every company has to move at once. Ignoring this has a real cost, and it arrives in three stages for the first thirty days after the subscription ends, everything still works. From day thirty one, interactive operations are delayed and on day ninety one and beyond. All operations are rejected. Microsoft does soften the landing a little for those same first thirty days. The company gets free premium capacity, matching the size it had before, which cuts down the odds of paying for two capacities at the same time. The data is not deleted, which is the reassuring half of that story. It stays in place but becomes unreachable until someone moves the workspaces onto a fabric capacity or deletes the old capacity entirely. The move itself is not automatic. Someone has to buy an F capacity and then reassign each workspace by hand or with a tool. Microsoft provides Microsoft's own advice is to buy the new capacity first move and test the workspaces and cancel the old one last. Large organizations don't have to click through every workspace one at a time. Microsoft ships a migration tool that works through its programming interfaces, which is what makes a move with thousands of workspaces realistic. One practical warning sits inside that sequence. A workspace reassignment can interrupt an active refresh or query, so the safe habit is to move workspaces outside the hours when reports refresh themselves. Most tenant settings carry over without change, which removes one common worry. Capacity level settings, such as memory limits and the capacity administrators do have to be set up again on the new capacity. Two things are not affected, and both are worth saying. Plainly premium per user is not being retired at all, and customers on sovereign clouds are untouched for the simple reason that fabric is not available there yet. What actually changes when you move to fabric? Now the upside because there is a real one. F capacities are bought through Azure build per second with a one minute minimum, and they can be paused. So a team that stops working over the weekend stops paying. Pausing is not a small change. Power BI premium capacity couldn't be paused at all, and a team could only scale it by adding more capacity in twenty four hour blocks of time. Fabric lets you scale the capacity up and down whenever you need to through the Azure portal, which is something the old capacity model never allowed. Commit for one year or three, and Microsoft puts the saving at about forty one percent against pay as you go. That is a real discount. Pausing does have a catch of its own, since a paused capacity can't serve reports. The skill is knowing which hours of the day. Nothing needs to run. Two more things come along with the move, and both were missing before. Power BI embedded is included with every F capacity, so separate embedded SKUs are no longer needed, and Azure native security features like managed private endpoints come with it, too. One finance detail carries real weight here. Fabric spend counts towards the Microsoft Azure Consumption Commitment and Power BI premium capacity spend never counted toward it at all. Power BI Report Server The on premises option still has a path forward. It comes with a reserved F sixty four purchase or with SQL Server Enterprise Edition, so teams that publish reports inside their own network are not stranded. None of that comes without trade offs, and two of them are worth knowing. Autoscale does not exist on fabric capacities. So scaling becomes something you do by hand and in a cross-region move, large data models and every fabric item have to be rebuilt rather than carried across. How to choose without getting it wrong. Knowledge only matters if it turns into a decision. Start with the free trial, which runs for sixty days on a trial capacity sized at F four or F sixty four because it costs nothing to look around before committing to anything. An individual learner can set all of this aside for a while. A free account plus a trial capacity is enough to build a report, open a lake house, and decide whether fabric is worth pursuing at all. Premium per user usually wins when premium capabilities are needed for fewer than two hundred and fifty people above that number, or once. A lot of people are only reading reports rather than building them. The math shifts toward buying a capacity. If the goal is lake houses, warehouses, or pipelines, a capacity stops being optional. Premium per user can't run those workloads at all, so a small F capacity or a trial capacity is the minimum you need for anything beyond power BI. Three mistakes show up again and again in this decision. The first is buying a small capacity to save money and then paying for a pro license for every viewer anyway, which defeats the entire point of the capacity. The second mistake is buying premium per user and expecting the whole of fabric, then discovering it is power buy only. The third mistake is smaller than it sounds. The retirement is real, but the deadline that matters is your agreement renewal date, not a date you saw in a headline. Take a ten person reporting team that grows to two hundred viewers. The day those viewers outnumber the builders is the day the F sixty four conversation starts, and it is usually cheaper than two hundred Pro licenses. Where this sits next to Microsoft three hundred sixty five and Azure, one question is left, and it's usually the first one. A beginner asks, where does all of this live? Inside the Microsoft world? They already know. And does it change the everyday tools they already open at work? The buying side sits in Azure. F capacities are purchased there, which means there is an Azure subscription and a cloud bill behind them. Even though nobody has to manage a server. One lake is built on Azure Data Lake storage, and Microsoft hides the infrastructure details such as resource groups, regions, and redundancy. The experience stays close to opening an everyday app. The everyday side is Microsoft three hundred sixty five. Fabric. Data can be analyzed in Excel and shared through teams, so the front door is a tool the audience already uses, which makes the whole platform feel far less foreign. Governance is centralized rather than scattered. The One Lake catalog is where items are found, secured and tracked, and Microsoft Purview layers protection, audit, and compliance on top of the data that it supports. Data can even be shared across company boundaries. Governance rules stay enforced at the source instead of being copied into the other tenant, which keeps one version of the truth instead of several. Picture a finance analyst exploring a fabric model in Excel. Then they drop the result into a team's channel and no data engineer is involved at all. Because the work happens in the apps that were already open. One line is worth repeating here because it trips up almost everyone. You don't need an Azure account to use fabric, but you do need one to buy the capacity. And that gap is where most beginners get stuck. The short version and a question for you. So here is the short version. Fabric and power BI premium are the same lineage one generation apart. A capacity is shared computing power. A license is one person's permission, and Ef64 is the line where viewing a report stops costing money for every reader. One of those two names is retiring at the end of your agreement, and the replacement is already here. Which one is your team paying for right now, the capacity or the licenses? If this made the topic feel less intimidating, subscribe to M365 FM to master the tools and the mindset shaping the future of work.