If you are still using the old flow of uploading an Excel or CSV file directly through the browser in Power BI Service, this article is a warning. Microsoft is shutting down support for the legacy importer, and the final deadline is August 31, 2026.
This is not a subtle change. After that date, reports that depend on semantic models created through the legacy flow will simply stop opening. No stale data warning, no friendly message. Just an error on the screen.
What is the legacy importer?
Back when Power BI Service was still young, Microsoft offered a straightforward way to work with data: you went to the 'Create' page, uploaded an Excel or CSV file directly through the browser, and Power BI generated a semantic model on the spot. No Power BI Desktop needed, no gateway configuration, nothing complicated.
That convenience came with a cost. Models created this way were limited: you could not edit them in the browser, there was no scheduled refresh option, and you could not download them back. For anyone who needed a quick, static report, it worked. For anyone who needed live data, it was a trap.
Microsoft decided to deprecate this flow, and did so in stages.
The dates you need to know
The deprecation happened in three phases:
- May 31, 2026: Creating new semantic models through the legacy flow was blocked. No one can start down that path anymore.
- July 31, 2026: Existing semantic models stopped refreshing. Reports still opened, but the data froze at that point in time. Anything updated in the source file after that date simply would not show up.
- August 31, 2026: The final deadline. Legacy semantic models are no longer available for querying. Reports linked to them return an error and will not open.
If you are reading this after August 31, the damage may already have happened. But it can still be fixed.
Who is affected?
The good news is that this change does not affect everyone. It specifically impacts users who created semantic models by uploading Excel or CSV files directly through the Power BI Service web interface and have reports connected to those models.
If you built your models in Power BI Desktop and published them to the Service, you are fine. If you use the Excel and CSV connectors inside Desktop and work with files on OneDrive or SharePoint, there is nothing to worry about either.
The quickest way to identify legacy models is to check in Power BI Service whether they have the options to edit in the browser, download the file, or configure scheduled refresh. If none of those options appear, it is a legacy model.
Tenant administrators can use the Power BI REST API to examine the ContentProviderType property of datasets and identify all legacy models across the organization at once.
How to migrate
The migration is not complicated, but it requires attention so nothing gets lost.
Step 1: open Power BI Desktop. Use the 'Get Data' option to connect to the same Excel or CSV file that was feeding the legacy model. Desktop will create a new, modern semantic model with all available options.
Step 2: publish to Power BI Service. After configuring transformations and validating the data, publish the model to the correct workspace.
Step 3: rebind the reports. Old reports need to be pointed at the new semantic model. In Power BI Service, there is an option to reassign the data source of an existing report, which avoids the need to rebuild everything from scratch.
If the Excel or CSV file is stored in OneDrive for Work or SharePoint, Power BI can automatically sync any changes within about an hour, with no gateway configuration required.
For organizations looking to go further, moving data to Microsoft Fabric's OneLake is an option that eliminates duplication and centralizes everything in one place.
Why did Microsoft do this?
The legacy flow was technically inferior to what the platform has evolved to support. Maintaining two parallel systems, one modern and one old, generates maintenance costs, behavioral inconsistencies, and confusion for users.
The deprecation forces a standardization that, in the end, benefits everyone who migrates: modern models support scheduled refresh, are editable, can be downloaded, and integrate better with the full Power BI and Fabric ecosystem.
It is the kind of change that stings in the short term but makes sense over time.
What to do now
If you have not yet mapped your legacy models, that is the first step. Use Power BI Service to check workspace by workspace, or use the REST API for a complete tenant-wide view.
Then prioritize the most-used reports. Start the migration with those, validate the data, and only then move on to lower-traffic ones.
Do not wait until the last minute. The final deadline is August 31, and a broken report on the morning after is exactly the kind of problem nobody wants to explain to their manager on a Monday morning.