Microsoft Power Platform Licensing Guide: The 2026 Buyer’s Map

The five Power Platform products

Power Platform licensing covers five products with materially different commercial models. The honest summary:

• Power Apps. Low-code application platform. Two SKU paths: per-user ($20/month, unlimited apps), per-app ($5/month/user/app). Plus the M365 "seeded" Power Apps entitlement included in E1/E3/E5 (limited to Office data sources only — this is where the Premium Connector trap lives).
• Power Automate. Workflow and RPA platform. Four SKU paths: per-user ($15/month), per-flow ($100/month, unlimited users on one flow), Hosted RPA ($215/month, unattended RPA on Microsoft-hosted infrastructure), Attended RPA ($40/month).
• Power BI. Analytics platform. Three SKU paths: Pro ($14/month, content authoring), Premium per User ($24/month, authoring with Premium features), and Power BI Premium per Capacity now migrating to Microsoft Fabric F-SKUs — see our analysis of the P-to-F migration playbook.
• Power Pages. External-facing website platform. Two SKU paths: $200/month/authenticated-site, $75/month/anonymous-site. Per-site, not per-user.
• Copilot Studio. Conversational AI agent builder. $200/month/tenant for 25,000 messages plus the Copilot Credit / CCCU economy — see our Copilot Studio 2026 pillar.

The Premium Connectors trap

The single most expensive mistake we see in Power Platform deployments is the Premium Connectors trap. The M365 E3 and E5 seeded entitlement to Power Apps and Power Automate only covers Standard connectors — SharePoint, Outlook, Teams, OneDrive, Excel, and a small number of Microsoft surfaces. As soon as a Power App or Power Automate flow uses a Premium connector (Dataverse, SQL Server, Salesforce, ServiceNow, SAP, HTTP, custom connectors, or any third-party SaaS), every licensed user of that app or flow requires a paid Power Apps or Power Automate licence. Microsoft does not enforce this gracefully — the app keeps working until the next compliance review or audit, at which point the licensing-shortfall amount lands in the EA amendment.

The procurement discipline: maintain a Centre of Excellence inventory of every Power App and Power Automate flow with its connector list. Any app touching a Premium connector requires explicit Power Apps per-user or per-app licensing for every user touching it. This is the largest single source of audit exposure in Power Platform estates today.

Power Apps: per-user vs per-app sizing

The Power Apps per-user vs per-app decision turns on app density per user. The breakeven equation is straightforward: at $5/user/app, four apps per user is the breakeven against $20/user unlimited. Below four apps per user, per-app is cheaper. Above four apps per user, per-user is cheaper. The reality across deployments we have audited:

• Power Apps maker population (citizen developers). Almost always per-user. Makers build apps; they hit four apps very fast.
• Departmental utility apps (single-purpose). Almost always per-app. The expense-claim app, the visitor sign-in app, the equipment checkout app — each touches a single Premium connector and one user population.
• Enterprise apps (organisation-wide). Modelled per-app at scale — even when an app reaches 8,000 users it can be cheaper at $5/user-app than $20/user covering a single use case.

Power Automate: per-user vs per-flow vs RPA

Power Automate sizing follows the same logic but adds the RPA dimension. The per-flow SKU at $100/month covers an unlimited number of users on one flow — useful for organisation-wide workflows (approvals, notifications). The Hosted RPA SKU at $215/month is the only path to unattended bot execution; the Attended RPA SKU at $40/month covers desktop RPA running alongside a human user. The Microsoft hosted infrastructure entitlement matters for security and compliance posture; running RPA bots on customer VMs requires a separate Hosted RPA add-on.

Power BI: the Fabric migration

Power BI Premium per-capacity is migrating to Microsoft Fabric F-SKUs through 2026. Customers on Power BI Premium P1 ($5,000/month), P2 ($10,000/month), P3 ($20,000/month), P4 ($40,000/month) are being moved to Fabric F-SKU equivalents at varying ratios. The migration is not commercially neutral — the F-SKU pricing is materially different at certain capacity levels, and the autopause / autoscale model changes the consumption pattern. See our detailed coverage of Power BI Premium retirement options and the Fabric capacity EA negotiation leverage for the structural detail.

Dataverse and the per-environment capacity trap

Every Power Platform deployment consumes Dataverse storage (database, file, log) at per-environment granularity. The base capacity is seeded by Power Apps and Dynamics 365 licences; additional capacity is sold in storage packs (Database $40/GB/month, File $2/GB/month, Log $10/GB/month). Two procurement traps live here: (1) the seeded entitlement is across the tenant not per environment, so a single high-storage environment can consume the tenant’s seeded entitlement and trigger pack purchases; (2) deleted records remain in Dataverse log storage for 30 days by default, inflating log consumption beyond what makers expect. Audit the per-environment Dataverse consumption before any EA renewal — it is the most common surprise line item in Power Platform amendments.

EA negotiation levers specific to Power Platform

1. Connector-list disclosure. Walk into the renewal with a documented connector list per app and flow. The licensing case Microsoft makes turns on Premium connector use; documented inventory neutralises the case.
2. Per-user vs per-app cohort sizing. Segment the population by app density. The LSP default is per-user across the full base; the buyer’s default is per-app for the long tail and per-user for the maker cohort.
3. Capacity headroom vs commit. Microsoft prefers tenant-level capacity commits; the buyer’s preference is environment-level consumption with growth flex. Negotiate the commit shape — not just the unit price.
4. Fabric migration credits. Customers migrating from Power BI Premium P-SKU to Fabric F-SKU should expect transition credits, especially if the F-SKU pricing puts the customer at a structural disadvantage. Capture in the EA amendment.
5. RPA hosted vs customer-hosted. Hosted RPA at $215/month is materially more expensive than customer-hosted RPA at $150/month. For security-sensitive deployments hosted is the right call; for non-sensitive bots the customer-hosted path saves $780/year per bot.
6. Step-up from M365 seeded entitlement. If a Power Apps deployment is growing out of the seeded M365 entitlement, Microsoft typically offers a step-up path that is cheaper than retroactive per-user licensing. Capture before the audit lands.

Anonymised case study: $1.4M Power Platform rebuild

A 12,000-employee insurance carrier carried 8,400 Power Apps per-user licences ($2.02M annualised), 3,200 Power Automate per-user licences ($576K annualised), and a Power BI Premium P3 capacity ($240K annualised). Total annual Power Platform spend: $2.84M. We audited the estate: of 8,400 per-user Power Apps licences, 6,100 users were on 1–2 apps each (cheaper as per-app); of 3,200 per-user Power Automate licences, 2,400 users were running a single high-traffic flow (cheaper as per-flow). The Power BI P3 capacity was at 38% utilisation against the Fabric F64 migration equivalent ($96K annualised). We rebuilt: 2,300 Power Apps per-user (the makers and high-density users), 6,100 per-app at $5/user/app across the long-tail apps, 800 Power Automate per-user (the makers), 6 per-flow ($600/month covering the organisation-wide flows), and a Fabric F64 capacity. Total annual cost after rebuild: $1.44M. Net annualised saving: $1.4M. The Premium Connector audit also surfaced 14 apps using SAP and Salesforce connectors against unlicensed users — resolved inside the EA amendment with no penalty.

Power Platform licensing is the most complex SKU surface in the Microsoft estate after Azure. The procurement task is portfolio rationalisation against actual app density, flow patterns, and capacity consumption — not unit-price negotiation. Pair the licensing rebuild with a structured enterprise spend reduction playbook and a clear view of the 2026 EA tier-collapse landscape, and Power Platform stops being a per-product mystery and starts being a managed line item.