Concepts

Where Power Query Runs

A complete map of every environment where Power Query M can be executed — from desktop tools to cloud dataflows to the new Fabric programmatic API.

Power Query M runs in more places than most users realize. What started as a desktop data import tool in Excel and Power BI has expanded into a cloud execution engine, a pipeline activity, and — as of February 2026 — a callable REST API. The engine is the same across all of these surfaces, but each environment has different capabilities, limits, and licensing requirements.

Microsoft Fabric Dataflows Gen2

Dataflows Gen2 is the most capable Power Query environment as of 2026. It runs in the cloud on the Modern Evaluator engine (rebuilt on .NET 8), supports 80+ connectors at GA, and stores results as Delta tables in OneLake. Output destinations include Fabric Lakehouse, Fabric Warehouse, Azure SQL, ADLS Gen2, and more.

Requires a Fabric capacity (not just Power BI Pro). Supports on-premises sources via gateway.

The Fabric Execute Query REST API (Preview, February 2026)

Microsoft announced in February 2026 that Power Query M can now be executed programmatically through a public REST API:

// API endpoint:
// POST https://api.fabric.microsoft.com/v1/workspaces/{workspaceId}/dataflows/{dataflowId}/executeQuery

The API executes a query within a Dataflow Gen2 artifact, which provides the execution context — connections, credentials, and gateway routing. Results are returned as an Apache Arrow stream, consumable as Pandas or Spark DataFrames in Fabric Notebooks or any HTTP client.

Key constraints in the current preview:

  • Queries must complete within 90 seconds
  • Read-only — no write operations
  • Requires a Dataflow Gen2 artifact as the execution context
  • Requires Fabric capacity
  • Billed as "Dataflows Gen2 Run Query API" in capacity metrics

This is a foundational shift: Power Query is now a callable data transformation service, not just an authoring UI.

Power BI Desktop

The original Power Query environment. Runs fully local on Windows (or Mac with limitations). Supports all connector types including local files, has a full Advanced Editor for M code, and supports custom connectors (.mez files). Data stays local until the model is published to the Power BI Service. No scheduled refresh without publishing.

Mac users have a reduced connector set and some M functions are unavailable.

Power BI Service — Dataflows Gen1

Cloud-only dataflows available with Power BI Pro, PPU, or Premium. Stores data in CDM format (CSV files in ADLS Gen2 or internal Power BI storage). Supports scheduled refresh and on-premises sources via gateway. Limited to 50 tables per dataflow and a 512 KB mashup script size.

Dataflows Gen1 does not support the new executeQuery API — that is exclusive to Fabric Dataflows Gen2.

Excel Desktop

The same Power Query engine as Power BI Desktop — both use the Mashup Engine under the hood. Available in Excel for Windows via Get & Transform Data. Full M language support, custom connectors, local files, databases, web sources, and SharePoint. The Advanced Editor exposes raw M code.

Mac users face the same limitations as Power BI Desktop on Mac: fewer connectors, some functions unavailable.

Excel Online (Rolling Out)

Power Query is rolling out to Excel for the web as of 2024–2025. Viewing and refreshing existing queries is available to all Microsoft 365 subscribers. Creating and editing queries requires a Microsoft 365 Business or Enterprise plan — Personal and Family plans do not have the editor.

Local file sources are not yet available in the web version. The feature is still rolling out and is not at full parity with Excel Desktop.

Power Automate and Power Apps Dataflows

Both Power Automate and Power Apps include Power Platform Dataflows — a cloud execution of Power Query Online that loads data into Dataverse, ADLS Gen2, or Power BI workspaces. On-premises sources are supported via gateway. Subject to the same 50-table and 512 KB limits as Power BI Dataflows Gen1. Dataverse has a 5-minute per-query timeout.

These are separate from Fabric Dataflows Gen2 and do not support the executeQuery API.

Azure Data Factory — Wrangling Data Flows

ADF embeds Power Query Online for transformation authoring in its "Wrangling Data Flows" feature. However, the M code you write is translated into Spark for distributed cloud execution — it does not run on the native M engine. This translation enables cloud scale but means not all M functions are supported. Unsupported functions produce: "Expression.Error: The transformation logic is not supported as it requires dynamic access to rows of data, which cannot be scaled out."

Use ADF wrangling flows when you need Power Query's authoring experience at Spark scale; be aware that behavior may differ from native M execution.

Analysis Services (SSAS / Azure AS)

Tabular models in SQL Server Analysis Services 2017+ and Azure Analysis Services use Power Query M as the data source definition language for partitions. M is authored in Visual Studio (Analysis Services Projects extension) and executed at model processing time. Not all connectors are available in this context, and execution is triggered by model processing — not on-demand.

Environment Summary

  • Full local M engine: Power BI Desktop, Excel Desktop (Windows)
  • Cloud M engine (UI-driven): Fabric Dataflows Gen2, Power BI Dataflows Gen1, Excel Online, Power Automate/Apps Dataflows
  • Cloud M engine + programmatic API: Fabric Dataflows Gen2 only (as of Feb 2026 preview)
  • M-to-Spark translation: Azure Data Factory Wrangling Data Flows
  • Model processing context: SQL Server / Azure Analysis Services

The On-Premises Data Gateway works across all cloud environments to route queries to on-premises or private network data sources — it does not execute M; execution always happens in the cloud, with the gateway acting as a relay.

Related Concepts