Microsoft Fabric is a bundle. Microsoft took the data tools it used to sell separately (the warehouse, the data engineering tools, the pipelines that move data around, the real-time streaming bits, and Power BI itself) and packaged them into one product with one storage layer underneath and one bill.
The bill is the part that changes how you buy. Power BI on its own is sold per person, at £10.80 per user per month for Pro on Microsoft's published UK prices as of September 2026. Fabric is sold as capacity: you rent a block of compute power, billed through Azure, and everything running on it draws from that block.
That is genuinely most of what a buyer needs to know. The rest of this is about who needs it, who does not yet, and what it does not fix.
What is actually in Fabric?
| Part of Fabric | What it does | Did you already have this? |
|---|---|---|
| Power BI | Reports and dashboards | Yes, this is the part you know |
| OneLake | One storage layer everything else reads from | This is the genuinely new idea |
| Data Factory | Moves and transforms data on a schedule | Existed separately before |
| Data Warehouse | SQL warehouse for structured data | Existed separately before |
| Lakehouse | Storage for large or messy data, queried directly | Existed separately before |
| Notebooks and Spark | Code-based data engineering, usually Python | Existed separately before |
| Real-Time Intelligence | Streaming data, event handling | For genuinely live data needs |
| Data Activator | Triggers actions when data hits a condition | Newer, still finding its use cases |
Most businesses reading this will use two of those: Power BI, and possibly Data Factory to move data around. The rest are there for organisations with data engineering teams, and paying for the bundle does not mean you have to use the bundle.
What is the actually new idea?
OneLake, and it is worth understanding because it is the honest argument for Fabric.
The old pattern was that each tool kept its own copy of the data. Your warehouse had a copy, your reporting layer had another, your data science team had a third. Copies drift, cost storage, and you spend real effort keeping them in step.
OneLake is one storage layer that all the tools read from directly. The warehouse, the notebooks and the reports look at the same physical data rather than at three copies of it. When that matters, it matters a lot, because "which copy is right" is a question that consumes an enormous amount of time in larger organisations.
If you have one accounting system and a spreadsheet, you do not have that problem, and the solution to it is not worth paying for.
Who genuinely needs Fabric?
Three groups, in my experience.
Organisations with hundreds of report readers. This is the clearest case and it is arithmetic rather than architecture. On a Fabric capacity of F64 or larger, people with a free licence and a viewer role can read Power BI content, per Microsoft's licensing documentation. Below F64, every viewer needs a Pro licence. Multiply your reader count by £10.80 a month and at some point the capacity is cheaper. That point is a real number you can calculate.
Organisations doing genuine data engineering. If you have people writing Python to process data, running pipelines nightly across multiple systems, or handling volumes that make a plain model impractical, Fabric gives them somewhere to work that connects directly to the reporting layer.
Organisations already committed to Azure. If your data estate is already Microsoft and your team already knows these tools separately, consolidating has real administrative value.
Who does not need it yet?
Most UK small and mid-sized businesses, and I would rather say that plainly than sell an assessment to someone who does not need one.
If your reporting runs on one or two source systems, has tens rather than hundreds of readers, and nobody on your team writes code to process data, Fabric is headroom you will pay for and not use. Power BI Pro licences will cost you a fraction of a capacity and do everything you need.
The pressure to buy it anyway is real, because Fabric is where Microsoft is heading and nobody wants to be told they are behind. Being behind on a platform you do not need is not a problem. The licensing arithmetic is worth doing before anyone makes the strategic argument to you.
What Fabric does not fix
This is the section I care most about, because it is where money gets wasted.
Reports that disagree with each other. If your sales report and your finance report show different revenue, that is because they define revenue differently. Fabric does not adjudicate definitions. Moving both reports onto a capacity gives you two reports that still disagree, now on more expensive infrastructure. The fix is a shared model with agreed definitions, described in six reports, six different numbers.
Bad data at source. Fabric moves and stores data faster. If what arrives is wrong, it now arrives wrong more efficiently.
Slow reports. A report that is slow because the model is badly shaped will be slow on a capacity too. Capacity buys compute, not design. The real causes are in why your Power BI report is slow.
Nobody using the dashboards. An adoption problem is a design and trust problem. No platform has ever solved it.
If any of those four are your actual problem, Fabric is an expensive way not to solve it.
What does it cost?
Fabric capacity is billed through Azure, per second, with regional pricing and a discount for reserving a year up front. I am deliberately not printing a pound figure, because it varies by region and changes, and a stale number here would mislead someone. Get yours from the Azure pricing calculator with your region selected.
What is worth knowing is the shape. Capacity is a fixed monthly cost whether you use it or not, unless you pause it. That is a different budgeting mindset from per-user licensing, and it produces bills that surprise people. The specific ways that happens are covered in why your Fabric capacity bill is so high.
One more thing worth knowing: if you are on an older Power BI Premium P SKU, Microsoft is retiring those and directing customers to Fabric F SKUs instead. That makes your next EA renewal a Fabric conversation whether or not you were planning one.
How should you decide?
Answer four questions honestly.
How many people need to read reports, and what do their Pro licences cost annually? Does anyone in your business write code to process data, or is that a role you are hiring? Do you have several systems that must reconcile with each other, or one system and some spreadsheets? And is your current problem about infrastructure at all, or about definitions, design and trust?
If the answers point to Fabric, the pre-purchase checks are worth doing properly, because buying capacity is easy and sizing it correctly is not. That is what a readiness assessment is for, at £5,950 fixed, and the checks worth doing regardless of who does them are in Fabric readiness: what to check before you buy a capacity. If the answers point away from Fabric, the honest recommendation is to spend the money on the model instead, and revisit in a year.
Common questions
What is Microsoft Fabric in simple terms?
It is Microsoft's data tools (Power BI, warehousing, data pipelines, data engineering, real-time analytics) sold as one product with one shared storage layer, bought as a block of compute rather than per person. For most existing Power BI users, it is the same Power BI plus a lot of things they may never open.
Is Microsoft Fabric replacing Power BI?
No. Power BI is part of Fabric and continues to be sold separately per user. Nothing about your Power BI licences stops working because Fabric exists. What is being retired is the older Power BI Premium per-capacity P SKUs, which Microsoft is consolidating into Fabric F SKUs.
Do I have to move to Fabric?
Not if you buy Power BI per user, which continues to be supported and sold. The main group facing a forced decision is organisations on legacy Premium P capacities, since those are being retired in favour of F SKUs. Everyone else can decide on the merits.
How much does Microsoft Fabric cost?
Capacity is billed per second through Azure with regional pricing and a reservation discount, so the figure depends on your region and commitment. Get it from the Azure pricing calculator rather than from a blog post. The important shape is that it is a fixed monthly cost regardless of usage, unlike per-user licensing which scales with headcount.
What is the difference between a lakehouse and a warehouse?
A warehouse holds structured, well-organised data and is queried with SQL, which suits finance-style reporting. A lakehouse holds data in a rawer form, including large or semi-structured files, and suits data engineering and machine learning work. In Fabric both sit on the same OneLake storage, which is much of the point.
Will Fabric make my reports faster?
Only if the reason they are slow is genuinely a lack of compute, which is uncommon. Most slow reports are slow because of how the data model is shaped, and that follows the model onto whatever it runs on. Diagnose the cause before buying capacity to fix it.