Power BI semantic models often use Azure Databricks, posing a critical choice of storage mode. This decision impacts cost, security, ease of development, and most importantly, report performance. Developers frequently rely on intuition rather than empirical evidence for this choice. A new white paper, "Modern Power BI Architecture Choices for Reporting on Azure Databricks," benchmarks four storage modes. These modes include Direct Lake on OneLake, Direct Lake on mirrored Unity Catalog tables, DirectQuery on a Databricks SQL warehouse, and a Composite Model combining DirectQuery with Import-mode aggregations. The study revealed no single best solution, but clear trends emerged. Direct Lake on OneLake generally performed well across various situations, especially for smaller to mid-sized volumes and typical repeated Power BI workloads. For extremely large datasets with billions of rows, a Composite Model with aggregations proved fastest and most consistent. However, the Composite Model's advantage is limited to queries resolved by aggregation tables, requiring careful alignment with user behavior. These findings are preliminary; detailed results vary by data volume, cache state, filter scenario, and query type. The white paper, released in June/July 2026, focuses on end-user query experience within Power BI and should guide individual testing.