Power BI Embedded vs Power BI Service - Which Licensing Model
The question I get asked more than any other Power BI question, by a wide margin, is some variation of: "Should we use Power BI Embedded or Power BI Service?"
It sounds like a simple question. It isn't. The wrong answer can cost you tens of thousands of dollars a year, or worse, lock you into a model that doesn't scale with your business. I've watched Australian companies pay for capacity they didn't need because someone told them Embedded was the "enterprise" choice. I've watched SaaS startups try to do customer-facing analytics on Power BI Pro and discover six months in that the licensing model literally doesn't let them.
This guide is the version of that conversation I usually have over the phone, written down. If you're trying to pick between Power BI Embedded and Power BI Service in 2026, here's how to actually make that decision.
The Distinction Most People Get Wrong
Most articles tell you that Power BI Service is for internal users and Power BI Embedded is for external users. That's mostly true but it's not really the right frame.
The actual question is who pays for the licence. With Power BI Service, every user who looks at the reports needs either a paid licence themselves (Pro or Premium Per User) or to be inside a Premium capacity that grants them free viewer rights. With Power BI Embedded, the users don't have any Microsoft licence at all. Your application sits in front of Power BI and uses an embedded token to render reports inside your own UI.
That single difference cascades into everything: who builds the experience, what it looks like, who handles authentication, and how it scales.
| Question | Power BI Service | Power BI Embedded |
|---|---|---|
| Who are the users? | People with work email at your company | People who use your app, regardless of where they work |
| Where do they view reports? | app.powerbi.com, Teams, SharePoint | Inside your application's UI |
| What do they log in with? | Their Microsoft 365 account | Your application's auth |
| Who pays per user? | You, every month | Nobody, it's bundled into your capacity cost |
| What about branding? | Looks like Power BI | Looks like your application |
| Who builds the embedding? | Microsoft did it for you | Your developers |
If you find yourself in a situation where the people viewing reports don't have a Microsoft 365 account at your company, you are almost certainly looking at an Embedded scenario, regardless of whether they're internal or external in some other sense.
Power BI Service in Plain Language
Power BI Service is the standard Microsoft offering. You publish reports to app.powerbi.com, users sign in with their work account, and they view reports in the browser or in Teams. Pricing for Australian businesses in 2026 looks like this:
Power BI Pro: ~$15.10 AUD per user per month. Every report builder needs one, and every viewer needs one unless content is hosted on Premium capacity. Dataset size capped at 1 GB. Eight refreshes per day.
Power BI Premium Per User (PPU): ~$30.20 AUD per user per month. Same as Pro plus access to Premium features like 100 GB datasets, more refreshes, paginated reports, and AI features. The catch is that PPU content can only be shared with other PPU users. If you have ten authors on PPU but 200 viewers, all 200 viewers also need PPU. This makes PPU expensive once you scale past a small team.
Power BI Premium Per Capacity (P-SKUs). You buy a block of capacity instead of per-user licences. Viewers only need a free Power BI account. P1 is around $7,200 AUD per month for entry-level enterprise capacity. The break-even versus Pro is roughly 475 users.
Microsoft Fabric capacity (F-SKUs). This is where most new Power BI Premium customers are landing in 2026. Fabric capacity provides Premium-equivalent Power BI features plus the rest of the Fabric data platform. F64 is approximately $7,200 AUD per month and is the equivalent of P1 for Power BI use. Smaller F-SKUs are available for smaller workloads, but you need at least F64 to grant free viewer rights to non-licensed users.
If your story is "we have a few hundred staff, they all have Microsoft 365 accounts, they want to see internal reports," you are firmly in Power BI Service territory. The decision is really just which tier you buy.
Power BI Embedded in Plain Language
Power BI Embedded is a different product with a different commercial model. It's designed for ISVs (independent software vendors) and for businesses that want to put Power BI reports inside their own applications, branded and authenticated their way.
The licensing is capacity-based and billed through Azure. You buy A SKUs that scale with the number of concurrent users hitting your reports. Approximate AUD pricing in 2026:
| SKU | Per Hour (AUD) | Per Month (AUD, 24/7) | v-Cores | Typical use |
|---|---|---|---|---|
| A1 | ~$1.50 | ~$1,100 | 1 | Dev, low-volume test |
| A2 | ~$3.00 | ~$2,200 | 2 | Small SaaS, dozens of concurrent users |
| A3 | ~$6.00 | ~$4,400 | 4 | Mid-size SaaS, 100+ concurrent |
| A4 | ~$12.00 | ~$8,800 | 8 | Mid-large SaaS, equivalent to P1 |
| A5 | ~$24.00 | ~$17,600 | 16 | Large SaaS, 1000+ concurrent |
| A6 | ~$48.00 | ~$35,200 | 32 | Very large workloads |
One of the underrated features of A-SKUs is that you can pause them. If your customers only use your analytics during business hours, you can have your capacity online 60 hours a week instead of 168. That's a 64% saving. We've worked with B2B SaaS companies that run A3 capacity from 7am to 7pm weekdays only and save roughly $30,000 a year doing it.
For the embedding itself, your developers need to call the Power BI REST API, generate an embed token, and render reports inside your application using the Power BI JavaScript SDK. It's not a huge amount of work, but it is real work, and it has to be done well. Row-level security is your responsibility to enforce in the embed token, not Power BI's.
The Decision Framework
Use this to figure out which model you actually need. Run through it in order.
1. Do your users have a Microsoft 365 account at your company?
If yes, default to Power BI Service. If no, you almost certainly need Embedded.
2. Are you putting reports inside your own application?
If yes, Embedded is the natural fit. The Service experience lives on app.powerbi.com and looks like Microsoft. Embedded reports live in your app and look like yours.
3. How many viewers will you have, and how does that compare to your authors?
For Service: if you have more than around 475 viewers, you should be on Premium capacity (Fabric F64 or P1) rather than Pro licences. The maths is simple. 475 x $15.10 = ~$7,173. Premium capacity = ~$7,200. Past that point, capacity gets cheaper per viewer.
For Embedded: capacity scales with concurrent users, not total users. You can support thousands of total users on a small capacity if only a handful of them are looking at reports at any given moment.
4. Are your viewers internal or external?
The Power BI Service licence is a per-user agreement with Microsoft. External users need either a guest account (which has its own complexity) or a B2B arrangement that gets messy fast. Embedded was built for the external scenario from day one.
5. How much development capacity do you have?
Service is essentially zero development. You build reports, you publish them, you share them, you're done. Embedded requires real engineering work, ongoing maintenance, and someone who understands the Power BI JavaScript SDK and embed token flows.
If your development team can't reliably ship a small application feature in two weeks, Embedded is going to be painful. We've been called in to rescue Embedded projects where the implementation team underestimated what "embed Power BI" actually meant. It's not a one-week job for a typical web app, it's a four-to-six-week job.
Scenarios from Real Engagements
These are anonymised but real situations we've worked through. Names changed.
Scenario 1: Mid-size accounting firm, 180 staff, all internal reporting.
This was a straightforward Service engagement. They had 12 report authors and 180 potential viewers. The maths said Pro licences for everyone, not capacity. Total monthly Power BI cost: around $2,800 AUD. They later moved to Fabric F8 for some pipeline work but kept Pro licences for users because the user count never grew enough to justify F64. Lesson: don't buy capacity you don't need.
Scenario 2: Australian property tech startup, 4,000 end customers viewing analytics.
This had to be Embedded. The customers were property investors, not staff. They didn't have email accounts at the startup. Power BI Service couldn't licence them at any reasonable cost. The startup landed on A2 capacity (~$2,200/month) with autoscale to A3 during peak reporting periods. Embedding work took about five weeks of one senior developer's time. Ongoing cost: roughly $30,000/year all-in, far less than the licensing cost would have been on Service.
Scenario 3: Industrial company with a customer portal and 1,200 internal staff.
This one needed both. Service for internal BI across the 1,200 staff (Premium capacity, F64 in Fabric for the data platform side too). Embedded for the customer portal, where roughly 300 contractors viewed job and safety reports. We built the Embedded capacity at A3 so it could be paused on weekends. Two separate workloads, two separate licensing models, both correct for their use case.
Scenario 4: SaaS founder who picked Embedded too early.
Founder had 30 paying customers, total. Bought A2 capacity at $2,200/month because someone told them Embedded was "the right way." Their actual usage was three concurrent viewers most days. We helped them rebuild on Pro licences for their early customers (their customers happily accepted Power BI logins for the early version) and saved them about $20,000 a year while their product matured. Lesson: don't buy the enterprise model when the bootstrap model would work.
Common Misconceptions That Cost Money
"Embedded is for serious enterprises." It isn't. It's for embedding reports in your application. Lots of small businesses run on A2 capacity. Lots of large enterprises run entirely on Service. The size of your business doesn't determine the model. The use case does.
"Premium is just a bigger version of Pro." It's substantially different. Premium gives free viewing to non-licensed users, but only inside Premium-hosted workspaces. If you bought P1 and left half your content in standard workspaces, those viewers still need Pro licences. We've seen this catch finance teams off guard at renewal time.
"We need Premium for AI features." Many AI features are now available on Pro and PPU. Check the current Microsoft licensing page before assuming you need Premium just for AI. The line keeps moving.
"Embedded means we can't use Microsoft 365 logins." You can, if you want to. Embed token generation is just an API call. You can use Microsoft Entra ID, your own identity provider, or anything else. The embed token is independent from how your users authenticate.
"Fabric replaces Premium." Sort of. F64 and above give you Premium Power BI equivalence and you can use that capacity for non-Power-BI Fabric workloads too. F2 through F32 don't have all the Premium Power BI features. If you bought F8 expecting Premium-level Power BI access, you'll be disappointed.
What to Budget for Your First Year
For Service at typical Australian mid-market scale:
- 10-50 users: $150-$750/month in Pro licences
- 50-500 users: $750-$7,500/month in Pro licences, or move to F64 once you hit ~475
- 500-2,000 users: F64 capacity ($7,200/month) plus Pro licences for authors
- 2,000+ users: F64 or higher capacity, plus Pro for authors
Add roughly 15-25% for storage, refresh compute on Premium, and the occasional spike.
For Embedded:
- Small SaaS, ~50 customers: A2 paused outside business hours, around $800-$1,200/month
- Mid-size SaaS, ~500 customers: A3 capacity, possibly autoscaling, around $4,400-$6,000/month
- Large SaaS, 2,000+ customers: A4 or A5, around $9,000-$18,000/month
- Plus initial embedding development: $40,000-$80,000 AUD for a typical implementation done well
Our Power BI consultants can walk you through what your specific Australian situation should cost. The above ranges work for most clients, but we've seen genuine outliers in both directions.
When the Answer is Actually Both
Quite a few Australian businesses end up running both models. Internal BI on Service, customer-facing analytics on Embedded. This is fine and often optimal. The two models don't conflict and can share underlying datasets if your architecture supports it.
If you're running this dual model, plan for governance. Reports built for internal Service consumption are often not suitable to embed as-is. They reference internal terminology, assume internal data access patterns, and don't have the row-level security pattern that external scenarios need. A separate set of reports, sharing the same semantic model, is usually the cleanest approach.
For larger organisations we work with, we generally recommend running both inside a Fabric capacity environment so the data platform is unified, with Embedded capacity (A-SKU) added separately for customer-facing scenarios. It's not the cheapest possible architecture but it's the easiest to govern.
How to Get This Decision Right
If you've read this far and still aren't sure which model fits, here's a short checklist:
- Write down exactly who the users are. Not "customers" or "staff" but the specific groups.
- For each group, write down whether they have a work email at your company.
- For each group, write down where they'll view reports (app.powerbi.com vs your app).
- Total up the user count by group.
- Use the framework above to map each group to a licensing model.
- Add up the costs across all models.
If you end up with two models, that's normal. If you end up confused, that's also normal, and it's the point where most Australian businesses bring in a partner.
We help businesses across Sydney, Melbourne, Brisbane, and beyond figure out the right Power BI licensing model and design the data platform behind it. If you'd like a straight answer for your situation, our Power BI consultants and our Microsoft Fabric consultants can take you through it in a 30-minute call. No pitch deck, just the numbers.
Get in touch through our contact page and we'll set something up. The licensing decision is one of those things that's much cheaper to get right at the start than to fix later.