Power BI Pro - Full Refresh Fails Because Dataset Size Exceeds 1 GB Limit (Dataset Size: 1053 MB)

Hi manoj_0911 

Yes — in Power BI Pro / shared capacity, the 1 GB semantic model limit is effectively a hard limit for an imported semantic model in the service.

Answers to your questions

1) Is the 1 GB semantic model limit a hard limit for Power BI Pro workspaces?

Yes. In shared capacity, imported semantic models are limited to 1 GB compressed size. If the refreshed model ends up at 1,053 MB, the service will reject it.
Microsoft’s refresh troubleshooting guidance also notes that shared capacity refreshes are subject to these model size constraints.
Docs:

• https://learn.microsoft.com/en-us/power-bi/connect-data/refresh-troubleshooting-refresh-scenarios
• https://learn.microsoft.com/en-us/power-bi/connect-data/service-dataset-modes-understand

2) Can this be increased without moving to Fabric/Premium?

No, not in Pro shared capacity.
To exceed the 1 GB limit, you generally need to move the workspace/semantic model to:

• Premium capacity
• Fabric capacity
• or sometimes Premium Per User (PPU), depending on your org setup and requirements

On shared capacity, Microsoft does not provide a way to raise the semantic model size limit.

3) Does Power BI build the entire model during a full refresh and then check the final compressed size before saving?

Yes, that’s the right mental model.
For an import refresh, Power BI typically:

• processes the incoming data,
• compresses/builds the new model,
• validates it against service limits,
• and only then swaps it in.

So a model can fail even if the old one was just under the limit, because the new compressed result is now over 1 GB.

Also important: refresh uses more memory than the final model size because Power BI may need to hold the existing model and the incoming data during processing. Microsoft notes refresh scenarios can require significantly more memory than the compressed model size.
Doc:

• https://learn.microsoft.com/en-us/power-bi/connect-data/refresh-troubleshooting-refresh-scenarios

4) Best practices to reduce semantic model size

Yes — beyond simply deleting rows/columns, these usually help most:

• Remove unused columns
  • Especially text columns and long descriptions
• Reduce cardinality
  • High-cardinality columns compress poorly
  • Examples: timestamps to the second, long IDs, GUIDs, free-text fields
• Use surrogate keys / integers instead of text
  • Integers compress better than strings
• Avoid loading detail you don’t need
  • Prefer summarized tables where possible
• Use incremental refresh
  • Only if your refresh pattern supports it
• Disable Auto Date/Time
  • It can create hidden date tables and add model bloat
• Use a star schema
  • Fact table + compact dimensions is usually smaller and faster
• Shorten text values if possible
  • Long repeated text can inflate dictionary size
• Move rarely used detail to DirectQuery or a separate model
  • Keep the import model focused on analysis
• Create aggregations
  • Especially if users mainly analyze at monthly/weekly/summary level

Microsoft’s modeling guidance on reducing import model size is a good reference:

• https://learn.microsoft.com/en-us/power-bi/guidance/import-modeling-data-reduction

5) Has this happened to others after gradual growth beyond 1 GB?

Yes, very commonly.
This is a classic issue: the model grows gradually over time, then one refresh pushes it just over the limit and the service rejects it.