Back to Blog

Microsoft Fabric Pricing and Licensing Explained for Australia - 2026 Edition

May 29, 2026•10 min read•Michael Ridland

If you've spent any time on Microsoft's pricing page trying to work out what Fabric will actually cost your business, you have my sympathy. The list prices are in US dollars, the capacity model is unfamiliar if you're coming from Synapse or standalone Power BI, and the discounts depend on whether you pre-commit, which region you pick, and what licences you already own.

We've sized Fabric deployments for Australian organisations ranging from a 30-person accounting firm in North Sydney to a national logistics business running about $80,000 a month of Azure spend. The pricing patterns are remarkably consistent once you understand the model. This guide walks through what you'll actually pay in AUD, the licensing traps we keep seeing, and how to decide whether Fabric is the right commercial fit before you commit.

The capacity model in plain English

Fabric is sold by capacity, not by user or by query. You buy a capacity SKU (F2, F4, F8, F16, all the way up to F2048), and every workload you run on Fabric draws from that capacity. Data Factory pipelines, Power BI reports, notebooks, KQL queries, warehouses, real-time intelligence - they all consume Capacity Units (CUs) from the same pool.

The number after the F is the number of Capacity Units the SKU provides. An F64, for example, gives you 64 CUs of compute, and Microsoft prices the SKUs roughly linearly: an F128 costs about twice an F64.

There are two things that catch people out here. First, Fabric uses "smoothing" and "bursting", which means short spikes above your capacity don't throttle you immediately. Instead, the overage gets carried forward and smoothed over a period (currently 24 hours for interactive operations and up to 24 hours for background jobs). Run too many big jobs and you hit throttling later. Second, the same SKU number means very different things for different workloads. An F8 is fine for a small Power BI tenant with a handful of reports. It is not fine for serious data engineering on millions of rows.

What you actually pay in Australian dollars

Microsoft prices Fabric per CU-hour, and the Australia East region adds a small regional premium over US East. The list pay-as-you-go prices for F-SKUs in Australia East at the time of writing convert roughly as follows (these change, so always check the calculator before committing):

SKU CUs Pay-as-you-go (AUD/month, approx) 1-year reserved (AUD/month, approx)
F2 2 $410 $250
F4 4 $820 $490
F8 8 $1,640 $980
F16 16 $3,280 $1,960
F32 32 $6,560 $3,930
F64 64 $13,100 $7,850
F128 128 $26,200 $15,700
F256 256 $52,400 $31,400

A few things to call out about that table. The pay-as-you-go price is billed per second, so you can turn capacity off overnight and on weekends if your workload allows. The 1-year reserved instance discount is roughly 40 percent if you commit upfront. The 3-year reservation is around 45 percent off, but we rarely recommend a 3-year commit for Fabric because the product is still maturing and SKU mixes change.

The F64 is the magic number for most mid-market Australian businesses. Why? Because at F64 you get free Power BI Pro for any user who only consumes content from that capacity. That is the single biggest licensing lever in Fabric, and it changes the maths for any organisation with more than about 60 Power BI viewers.

When the F64 break-even happens

Power BI Pro is around $20 AUD per user per month if you buy it through a Microsoft partner. If your organisation has 200 Power BI consumers (people who only view reports, not authors), you are paying around $4,000 a month for Pro licences. Move to an F64 with reserved pricing and you pay about $7,850 a month but those 200 consumers no longer need Pro.

In practice the break-even sits somewhere between 350 and 500 viewers depending on what discounting you've negotiated. Below that, paying for Pro is cheaper. Above that, F64 wins commercially even before you count any Fabric data engineering you do on top. We've had clients where the Fabric capacity was effectively free because of Pro savings alone.

The trap: authors (people who build and publish reports) still need a Power BI Pro licence to publish. The F-SKU free viewer model only covers consumption. We've seen organisations buy an F64 thinking it would replace all licences, then have to back-buy Pro for their 12 report developers.

P-SKUs, A-SKUs, EM-SKUs and why F-SKUs won

Older Power BI Premium customers will remember P-SKUs, sold per month with no pay-as-you-go option. Microsoft has been quietly steering customers off P-SKUs and onto F-SKUs since late 2024. As of early 2026 you can still renew an existing P-SKU but you cannot start a new one through most reseller channels. The functional equivalence is P1 = F64, P2 = F128, P3 = F256, though F-SKUs include all the Fabric data engineering workloads that P-SKUs do not.

If you are still running a P-SKU, the move to F-SKU usually saves money because of the pay-as-you-go flexibility, but it requires reassigning workspaces and re-checking your row-level security setup. We've done this migration for several Australian organisations and it is rarely a one-day job.

A-SKUs (embedded) and EM-SKUs (embedded for ISVs) still exist for specific scenarios. If you are embedding Power BI in a SaaS product, A-SKUs are still the right choice, and they remain billed by the hour through Azure with no commitment. Most readers of this article will not need them.

What gets billed on top of capacity

The capacity SKU is not the whole bill. There are three other line items that catch Australian buyers out:

Storage. Fabric uses OneLake, which is built on Azure storage. You pay separately for the data sitting at rest. OneLake storage in Australia East is currently around $0.04 AUD per GB per month. For most organisations this is a rounding error. For a manufacturing client of ours holding 28 TB of historical machine telemetry it is about $1,100 a month.

Egress. If you move data out of Azure (back to on-premises, to AWS, or to a third-party SaaS), you pay egress fees. These start at about $0.12 AUD per GB and tier down as volume increases. We had one client surprised by a $4,000 egress bill because their daily Power BI exports to a partner FTP server added up.

Mirroring. Fabric Mirroring (which replicates Cosmos DB, Snowflake, or other sources into OneLake) is currently free of charge for the replication itself, but you pay capacity CUs for the queries that run against the mirrored data, and storage costs apply.

If your data factory pipelines pull from on-premises sources, you also pay for the Self-Hosted Integration Runtime VM (often a Windows server in your environment) and the bandwidth, neither of which Microsoft bills directly but both of which are real costs.

Reserved vs pay-as-you-go - the actual decision

Reserved instances save 40 percent, which is significant. But the pay-as-you-go option lets you pause capacity outside business hours, and an Australian business that genuinely turns capacity off on weekends and overnight can pay less than the reserved price even at full list rate.

We typically recommend pay-as-you-go for:

  • Pilot projects and proofs of concept (always)
  • Development and test capacities (always)
  • Capacities used heavily during business hours but idle overnight
  • Workloads still being right-sized

We recommend reserved for:

  • Production capacities running 24/7 (which is most Power BI workloads)
  • Capacities at F64 or above where the saving is material in absolute dollars
  • Stable workloads where the team has at least three months of usage data

A common pattern we use with clients is an F64 reserved for the production Power BI workload plus a separate F8 pay-as-you-go for ad-hoc data engineering and notebooks. That keeps the developer playground cheap and isolates capacity contention.

The hidden costs nobody tells you about

A few costs that don't show up in any pricing calculator but absolutely show up in invoices:

Skill ramp-up. Fabric is a new platform, even for teams that have been doing Power BI and Synapse for years. Plan for either training or consulting time. We typically see Australian organisations spend $25,000 to $80,000 in their first six months getting people productive on Fabric concepts they didn't have before, like lakehouses, shortcuts, and direct lake mode.

Tenant clean-up. If you've been on Power BI Premium for a few years, you almost certainly have legacy workspaces, datasets, and gateway configurations that need rationalising before Fabric will perform the way Microsoft demos suggest. Allow for a tenant audit.

Capacity right-sizing. Most organisations buy too big initially because they're worried about throttling. After three months of real usage data we have helped clients downsize from F128 to F64 and from F64 to F32. The Capacity Metrics App is your friend here.

Premium Capacity Utilisation app. Free, built-in, and absolutely essential. If your Fabric admin is not reviewing this weekly, you are flying blind on cost.

Comparing Fabric to the alternatives

We get asked this all the time. The honest answer is that Fabric is the right choice for most Australian mid-market organisations that already use Microsoft 365 and Power BI heavily. The platform pulls together what used to be five separate products into one bill and one workspace.

For organisations with serious data engineering on AWS or Google Cloud, Snowflake or Databricks are usually a better technical fit even if the licensing is more complex. For organisations doing pure data science without much BI, Databricks remains the gold standard.

For pure Power BI shops with no need for warehouses, lakehouses, or notebooks, an F-SKU is overkill. Stick with per-user Pro licensing until your viewer count gets above the break-even.

We've helped Australian clients pick Fabric, pick Databricks, pick Snowflake, and pick "stay where you are". The right answer depends on what you actually run, not on what Microsoft's marketing says.

A practical sizing checklist

Before you commit to a capacity tier, run through this checklist:

  1. Count your Power BI viewers (consume only). If above 400, F64 likely wins.
  2. Count your Power BI authors. Multiply by Pro licence cost. Add to the bill.
  3. List your data factory pipelines and estimate daily volume in GB.
  4. List your warehouses or lakehouses and estimate concurrent users.
  5. Identify any real-time intelligence or KQL workloads (these are CU-hungry).
  6. Pick a starting SKU one tier above the calculator estimate.
  7. Run pay-as-you-go for 60 to 90 days.
  8. Review the Capacity Metrics App weekly.
  9. Right-size and convert to reserved.

This is the process we walk every Fabric client through, and it almost always lands them on a better SKU than the original Microsoft account team recommendation.

When to get help

If you are a small business with one Power BI tenant and a handful of reports, you do not need a Fabric consultant. The default settings will get you most of the way there.

If you are migrating from Synapse or a P-SKU, integrating Fabric with on-premises SAP or SQL Server, or running serious data engineering across multiple workspaces, the licensing decisions interact with the architecture decisions in ways that are hard to undo. Spending two weeks with someone who has done it before saves months of expensive trial and error.

We help Australian organisations with Microsoft Fabric consulting, including capacity sizing, licensing reviews, and full migrations. If you want a second opinion on a Microsoft quote, we are happy to do an honest review.

You can read more about our broader Microsoft data and AI services, our Power BI consulting work, or our Microsoft Data Factory practice. For organisations evaluating multiple platforms, our AI strategy consultants can help you compare options objectively.

Get in touch if you want to talk through your Fabric sizing or licensing options. We do free 30-minute reviews for Australian businesses considering the platform.