What causes Uncompressed data limits for refresh?

Importing high-cardinality columns (e.g., long text, GUIDs, or free-form strings) that compress poorly and expand significantly in memory during refresh. Importing more data than necessary — no query folding, missing filters, or overly wide tables with many unused columns. Semantic model size approaching or exceeding the 1 GB compressed limit for shared capacity, indicating the uncompressed data is far larger. Running large refreshes on Power BI shared capacity (Pro) rather than Premium or Embedded capacity, which has higher uncompressed data allowances

How do I fix Uncompressed data limits for refresh?

Step 1: Audit your data model in Power BI Desktop using the Model view — identify and remove unused columns, tables, and measures to reduce unnecessary data import.. Step 2: Apply aggressive query filtering in Power Query to import only the rows and columns required for reporting — push WHERE clauses and column selection to the source query for maximum query folding.. Step 3: Replace high-cardinality text columns with integer keys where possible — integer columns compress far more efficiently than string columns in the VertiPaq engine.. Step 4: Implement incremental refresh to partition the data and process smaller chunks during each refresh cycle, reducing the peak uncompressed data size in memory.. Step 5: If the dataset legitimately requires more than 10 GB of uncompressed data, upgrade to Power BI Premium Per User (PPU) or Premium capacity, which supports larger semantic models and higher uncompressed data limits.

High severitycapacity

Power BI Refresh Error:
Uncompressed data limits for refresh

What does this error mean?

The semantic model's uncompressed data volume exceeds the Power BI service limit during refresh processing. On shared capacity, the limit is 10 GB of uncompressed data, even though the compressed model on disk may be well under 1 GB.

Common causes

1Importing high-cardinality columns (e.g., long text, GUIDs, or free-form strings) that compress poorly and expand significantly in memory during refresh
2Importing more data than necessary — no query folding, missing filters, or overly wide tables with many unused columns
3Semantic model size approaching or exceeding the 1 GB compressed limit for shared capacity, indicating the uncompressed data is far larger
4Running large refreshes on Power BI shared capacity (Pro) rather than Premium or Embedded capacity, which has higher uncompressed data allowances

How to fix it

1Step 1: Audit your data model in Power BI Desktop using the Model view — identify and remove unused columns, tables, and measures to reduce unnecessary data import.
2Step 2: Apply aggressive query filtering in Power Query to import only the rows and columns required for reporting — push WHERE clauses and column selection to the source query for maximum query folding.
3Step 3: Replace high-cardinality text columns with integer keys where possible — integer columns compress far more efficiently than string columns in the VertiPaq engine.
4Step 4: Implement incremental refresh to partition the data and process smaller chunks during each refresh cycle, reducing the peak uncompressed data size in memory.
5Step 5: If the dataset legitimately requires more than 10 GB of uncompressed data, upgrade to Power BI Premium Per User (PPU) or Premium capacity, which supports larger semantic models and higher uncompressed data limits.

Frequently asked questions

What is the difference between the 1 GB compressed limit and the 10 GB uncompressed limit?

The 1 GB limit refers to the size of the .pbix file or the stored model on disk after VertiPaq compression. The 10 GB limit applies to the in-memory uncompressed size of the data during refresh processing on shared capacity — a 500 MB compressed model can easily expand to 10+ GB in memory depending on the data types used.

Does upgrading to Power BI Premium completely remove the uncompressed data limit?

Premium capacity significantly raises the limits — the maximum semantic model size increases to 25 GB or more depending on the SKU — but limits still apply. Very large models may require Large Dataset Storage Format to be enabled in Premium workspace settings.