Power BI is Microsoft's business intelligence platform — free Windows authoring in Power BI Desktop, publishing and sharing through the Power BI service at a per-user price, and an on-premises Report Server for organisations that cannot publish to the cloud.
Authored in Desktop, published to the managed service on per-user licences or Fabric capacity, with Report Server on-premises for what cannot leave. Per dataset, data imports into the VertiPaq in-memory engine, stays in DirectQuery pushdown, or reads Direct Lake on Fabric.
How Power BI answers the questions Business Intelligence turns on.
| How it works | |
| Authoring model | Power BI Desktop: drag-and-drop canvas, Power Query for prep, DAX for measures |
| Semantic layer | Semantic model: a tabular model with DAX measures and Power Query M |
| Query execution | Import into the VertiPaq in-memory engine, DirectQuery pushdown, or Direct Lake on Fabric |
| Chart library | A large built-in set plus the AppSource marketplace of certified and uncertified custom visuals, an SDK for your own, and R and Python visuals, those last unsupported in app-owns-data embedding |
| Definitions in | Binary .pbix, or the plain-text PBIP project format, TMDL and JSON, still preview |
| Alerts and delivery | Dashboard tile alerts on gauges, KPIs and cards only, to the notification centre and optionally email, private to whoever set them and capped at 250; report alerts go through Fabric Activator, and Power Automate picks up the rest |
| Connections | |
| Connectors | Power Query reaches hundreds of sources, spreadsheets and files, every major database and warehouse, Azure and third-party SaaS, web, OData, ODBC and OLE DB; an on-premises gateway covers private networks |
| Embedding | Two shapes: embed for your customers, where a service principal authenticates and app users need no licence, and embed for your organization, where each user signs in with their own; production needs an A, P/EM or F capacity, and secure embed is the no-code URL option |
| Access | |
| Access control | Enforced in the semantic model, RLS roles are DAX filters that bind only to Viewers, object-level security hides columns and measures, and sensitivity labels travel with exports |
| Cost | |
| Billing unit | Per user for Pro and Premium Per User, or capacity by the SKU, and the line that decides it is F64 and above, where viewers stop needing a Pro licence at all |
vs Power BI: Managed · Serverless · Subscription
vs Power BI: Open source (permissive) · Operational complexity: High · Python, TypeScript
vs Power BI: Open source (permissive) · JavaScript, Svelte