The Power BI and Fabric ecosystem, explained

Power BI is no longer one product. It sits inside Microsoft Fabric, next to lakes, warehouses, pipelines and AI. This map shows how the pieces fit, from the data coming in to the report on someone's screen, and what each one is for.

How the pieces fit

Data flows left to right, from the systems you run to the people who decide. Pieces with an arrow have a full page.

  1. 1. Get data

    Bring data in from the systems the business already runs.

    • On-premises data gateway
    • Power Query
    • Dataflows Gen2
    • Data pipelines
    • Mirroring
    • Shortcuts
  2. 2. Store

    Keep that data in one place, in open Delta tables.

  3. 3. Engineer

    Clean, join and reshape data with code.

    • Notebooks and Spark
    • Semantic link
  4. 4. Model

    Turn tables into business numbers people can trust.

  5. 5. Show

    Put those numbers in front of the people who decide.

    • Reports
    • Paginated reports
    • Dashboards
    • Apps
    • Embedded analytics
  6. 6. AI

    Let people and agents ask questions of the model.

  7. 7. Run and govern

    Pay for it, ship changes safely and keep it under control.

Source: Microsoft Learn documentation, grouped by me. 10 of 32 pieces have a page so far; the rest are on the way.

Every piece, in one line

A plain definition of each part of the ecosystem, grouped by what it does.

Get data

On-premises data gateway
The bridge that lets the Power BI service refresh data that lives inside your own network.
Power Query
The low-code tool that connects to a source and cleans the data before it reaches the model.
Dataflows Gen2
Power Query in Fabric, writing its output to a Lakehouse or Warehouse instead of one model.
Data pipelines
Scheduled workflows that copy data and run the steps of a load in order.
Mirroring
A near real-time copy of an operational database, kept in sync inside OneLake.
Shortcuts
Links to data in other storage, used as if it were in OneLake, without copying it.

Store

OneLake
The single data lake under every Fabric workspace, storing tables in the open Delta format.
Lakehouse
Files and Delta tables in OneLake, written with Spark and read with SQL.
Warehouse
A T-SQL data warehouse in Fabric, with full read and write, storing its tables in OneLake.
Eventhouse
A store for streaming and time-series events, queried with KQL in seconds.

Engineer

Notebooks and Spark
Python, SQL or Scala code that cleans and joins data at scale, writing Delta tables.
Semantic link
A Python library that lets notebooks read semantic models and run DAX against them.

Model

Semantic model
The tables, relationships and measures that turn raw data into the business numbers every report uses.
Import and DirectQuery
The two classic storage modes: copy data into memory, or query the source every time.
Direct Lake
A storage mode where the semantic model reads Delta tables in OneLake directly, with no import refresh.
DAX
The formula language of semantic models, used to write the measures every report shows.
Calculation groups
Reusable calculation patterns, such as year-to-date, applied to any measure without copying it.
Row-level security
Rules in the semantic model that limit which rows each user can see.

Show

Reports
Interactive pages of visuals, each visual a query against the semantic model.
Paginated reports
Pixel-exact, printable reports for invoices, statements and long tables.
Dashboards
Single pages of pinned tiles from several reports, for a quick view.
Apps
A packaged set of reports, shared to an audience without giving them workspace access.
Embedded analytics
Power BI reports shown inside your own app or portal.

AI

Copilot
The AI assistant that answers questions, writes DAX and builds report pages from your semantic model.
Data agents
AI agents in Fabric that answer questions over your Lakehouse, Warehouse and semantic models.
MCP servers
A standard way for AI tools to read, and sometimes change, a semantic model.

Run and govern

Workspaces
Containers for reports, models and Fabric items, with roles that decide who can do what.
Capacity and F-SKUs
The compute you buy for Fabric and Power BI, sized from F2 to F2048 and shared by every workload.
PBIP and Git
Saving reports and models as text files (PBIP, TMDL, PBIR) so changes can be reviewed in Git.
Deployment pipelines
Stages such as Dev, Test and Prod that move content between workspaces with a click.
Purview and governance
Sensitivity labels, lineage and catalogue tools that show what data exists and who may see it.
Admin and monitoring
Tenant settings, the admin portal and the Capacity Metrics app that keep the platform healthy.

Not sure which pieces you need?

Tell me what you run today and where it hurts. In 20 minutes I'll tell you which parts of this map matter for you, and which you can ignore.

Book a 20-min call