Fabric vs Synapse vs "just Power BI": which Microsoft data stack?

For most UK small businesses the right Microsoft data stack is "just Power BI": Pro licences and a well-built model. Choose Fabric when you have real engineering work, such as pipelines, a lakehouse or a warehouse, alongside your reports. Stay on Synapse only if you already run it or need something Fabric still lacks.

That is our view, not Microsoft's marketing line. Below: the reasoning, entry prices from Microsoft's price list as of October 2026, and what to do if you already run Synapse.

What are the three options, really?

They are three answers to one question: where should your data be prepared before anyone looks at a report?

• Just Power BI. Power BI Desktop connects to your sources, Power Query cleans the data, a model sits on top and reports publish to the Power BI service. You pay per user. There is no separate platform to run.
• Microsoft Fabric. A software-as-a-service analytics platform with one storage layer, OneLake, and workloads for data movement, Spark, warehousing, real-time data and Power BI. You pay for a capacity (an F SKU) by the hour or on a reservation. Our plain-English Fabric explainer covers what is inside it.
• Azure Synapse Analytics. An Azure platform-as-a-service that, in Microsoft's own description, brings together SQL data warehousing, Spark, pipelines for data integration and integration with Power BI. You provision and pay for each engine separately: dedicated SQL pools, a serverless SQL endpoint, Spark pools.

Fabric and Synapse overlap heavily; Power BI is the reporting layer either way.

How do they compare?

	Just Power BI	Microsoft Fabric	Azure Synapse Analytics
What you buy	Per-user licences	One capacity (F2 to F2048) shared by every workload	Separate engines: dedicated SQL pool, serverless SQL, Spark pools, pipelines
Entry price, UK, as of Oct 2026 (ex VAT)	Pro £10.80 per user per month, paid yearly	F2 about £231 a month pay-as-you-go, or about £138 on a 1-year reservation	Smallest dedicated SQL pool (DW100c) about £915 a month if left running; serverless SQL £4.72 per TB processed
Storage	Inside your semantic models	OneLake, billed separately	Azure Data Lake Storage Gen2 and SQL pool storage, billed separately
Who runs it	An analyst	An analyst plus someone who owns pipelines and capacity	A data engineer comfortable with Azure resources
Built-in deployment and CI/CD	Deployment pipelines need a Premium, PPU or Fabric workspace	Git integration, deployment pipelines, Fabric CLI	Azure DevOps integration; Microsoft lists no built-in CI/CD
Best fit	Up to a handful of sources and one reporting team	Many sources, scheduled transformation, several teams reusing the same data	Existing Synapse estates, or the few needs Fabric does not yet cover

Sources for the table: Microsoft's UK Power BI pricing page, the Azure Retail Prices API (UK South, GBP, queried 7 October 2026), and Microsoft Learn's Fabric vs Azure Synapse Spark comparison. The deployment-pipelines detail is covered in our guide to Power BI deployment pipelines.

What does each cost to start?

All figures are list prices in UK South, ex VAT, as of October 2026. The monthly figures are our arithmetic on Microsoft's hourly rates, using 730 hours a month.

Just Power BI. Pro is £10.80 per user per month, paid yearly, and Premium Per User is £18.50 (Microsoft UK pricing). Ten Pro users is £108 a month. Nothing else is required.

Fabric. The Azure Retail Prices API lists Fabric capacity at £0.1585 per capacity unit per hour on pay-as-you-go, and £825.6915 per capacity unit per year on a one-year reservation. An F2 has two capacity units, so:

• pay-as-you-go: 2 × £0.1585 × 730 = £231.41 a month if it runs all month (you can pause it);
• one-year reservation: 2 × £825.6915 ÷ 12 = £137.62 a month.

OneLake hot storage is a further £0.0181 per GB per month. One catch matters more than the capacity price: below F64, people viewing Power BI content still need Pro or PPU licences. An F2 adds to your Pro bill rather than replacing it. Our Power BI licensing guide explains the F64 threshold, and the Fabric pricing guide walks through the full F SKU range.

Synapse. Each engine is priced separately:

• dedicated SQL pool DW100c, the smallest size: £1.2529 an hour, so £914.62 a month if left running. It can be paused, and that is how most small estates keep the bill down;
• serverless SQL: £4.7172 per TB of data processed, with no charge when idle;
• Spark (memory optimised): £0.1156 per vCore-hour. Microsoft's comparison page gives a three-node minimum for a Synapse Spark pool, so three small four-vCore nodes come to 12 × £0.1156 = about £1.39 an hour while running;
• SQL pool storage: £17.3591 per TB per month.

These units do not convert: a DW100c is not "the same size" as an F2. Compare your own workload's monthly bill on each, not headline rates.

Is Synapse being retired?

We found no retirement date in Microsoft's Synapse documentation as of October 2026, so existing workloads are not on a deadline. But Microsoft's migration guidance and tooling all point one way, from Synapse to Fabric, which is why we would not start a new project on Synapse today.

The evidence for "one way" is in Microsoft Learn. There is a Fabric Migration Assistant for Data Warehouse for dedicated SQL pools, and a Spark Migration Assistant plus a four-phase best-practice series for Synapse Spark (migration overview). We found no equivalent guidance for moving the other way.

Synapse still has a few things Fabric does not. Microsoft's own comparison names GPU-accelerated Spark pools, an external Hive Metastore and .NET for Spark (C#) as Synapse-only. Fabric Data Warehouse also does not currently support the datetimeoffset data type, so time-zone offsets have to move into a separate column on migration. If your estate depends on one of those, staying on Synapse for now is a reasonable engineering decision rather than inertia.

When is "just Power BI" enough?

More often than vendors would like. Our view is that the problem behind many "we need a data platform" conversations is a weak model, not a missing platform: no clear fact and dimension tables, calculations duplicated across reports, refreshes that time out. A properly built star schema fixes those, and costs nothing extra in licences.

Our rule of thumb is to stay on Power BI alone while all of these hold:

• you have fewer than about six or seven sources, and Power Query can reach them all;
• one team owns the reports and nobody else needs the cleaned data;
• scheduled refresh finishes comfortably inside the window;
• you cannot name a pipeline you would build in Fabric this quarter.

The last test matters most: unused capacity is money spent on a platform rather than the reporting problem.

When does Fabric earn its place?

Fabric pays its way when the data preparation becomes a job in its own right. Signs to look for:

• the same cleaned data is needed by several reports, teams or tools, and is being rebuilt in each;
• sources include files landing in a lake, APIs or large tables that Power Query struggles to refresh;
• you need a warehouse or lakehouse that other tools (Python, Excel, an app) can query;
• you want proper development, test and production stages with Git behind them.

At that point the comparison with Synapse is not close for a new build. Fabric is one capacity bill and one admin surface. Microsoft's comparison lists Git integration, deployment pipelines and the Fabric CLI on the Fabric side and no built-in CI/CD on the Synapse side. It also gives Fabric a starter pool of pre-warmed Spark and one-node minimum jobs, where Synapse needs three nodes. Our data engineering service covers this kind of build.

We already run Synapse. What should we do?

Nothing rushed. With no announced retirement date, the sensible move is to plan the migration around your next significant change: a new source system, a reporting overhaul or a contract renewal. Avoid doing it as a project of its own.

When you do plan it, Microsoft is clear that it is "a copy-and-adapt process rather than a direct in-place move". Its planning guidance describes two shapes:

1. Lift and shift. Microsoft suggests this for a small number of warehouses with data already in a well-designed star or snowflake schema, which describes many SME estates. Expect data type changes (money becomes decimal(19,4), datetime becomes datetime2) and some T-SQL rewrites.
2. Phased modernisation. This suits a warehouse that has grown over years and needs re-engineering anyway, or one you want redesigned around lakehouse features.

Either way, Microsoft's runbook includes running old and new in parallel and comparing results before cutting reports over. Budget for that overlap, because it is the period when you pay for both platforms.

Not sure which camp you are in? Our Power BI consultancy and wider data practice do exactly that assessment. The pricing estimator prices the licence side.

Frequently asked questions

Is Azure Synapse being retired?

We found no retirement date in Microsoft's Synapse documentation as of October 2026, so existing workloads are not on a deadline. But Microsoft's migration guidance and tooling all point one way, from Synapse to Fabric, which is why we would not start a new project on Synapse today.

Is Fabric cheaper than Synapse?

For a small, steady workload it usually is, because the smallest Fabric capacity costs far less than a dedicated SQL pool left running. As of October 2026 in UK South, an F2 is about £231 a month pay-as-you-go, while the smallest dedicated pool is about £915 if left on all month. The units are not equivalent, so test with your own workload.

Does a Fabric capacity replace Power BI Pro licences?

Not on the smaller sizes. Below F64, anyone viewing Power BI content still needs a Pro or Premium Per User licence, so an F2 is an addition to your Pro bill rather than a substitute for it. Only from F64 can users with a free licence view content in that capacity.

Do we need Fabric or Synapse to use Power BI properly?

No. Most UK small businesses get everything they need from Power BI Pro licences and a well-built data model. A platform earns its place when you have engineering work to do: many sources to land, transformations to run on a schedule, or data volumes Power BI should not be wrangling on its own.

---

Easy Insight is a UK consultancy for AI, web, apps and data — senior specialists only, no juniors.