Databricks unlocks SAP semantic metadata

Enterprise data is usually not missing rows. It is missing meaning. That is the gap Databricks and SAP are trying to close with a new integration that pushes SAP semantic metadata straight into Unity Catalog alongside shared data. The news is that this capability is now generally available, with Databricks announcing on April 30, 2026 that SAP Business Data Cloud shares can automatically sync their business context and governance tags into Databricks’ catalog layer. (databricks.com) ### What is the actual thing that launched? It is automatic semantic metadata sync between SAP Business Data Cloud and Databricks Unity Catalog. When a team mounts an SAP BDC Delta Share into a catalog and accesses a table, Databricks now pulls in the SAP-side metadata at the table level instead of leaving users (databricks.com) roadmap promise. (databricks.com) ### What counts as “semantic metadata” here? Not just comments. The synced package includes table and column descriptions, primary-key and foreign-key relationships, and governance tags from SAP. That sounds boring, but it is the part that tells a human — or a model — whether a field is a customer ID, how two tab(databricks.com)uth, and the synced metadata is read-only inside Databricks. (databricks.com) ### Why was this a problem before? Because zero-copy access alone does not solve the “what am I looking at?” problem. Delta Sharing already let SAP data products flow live into Databricks, but users still had to reconstruct business context if the metadata did not come with the tables. In practice, that means slo(databricks.com)iness meaning embedded in SAP’s model. (databricks.com) ### Why does Unity Catalog matter so much? Unity Catalog is where Databricks handles discovery, permissions, and governance. So if SAP semantics land there, they become visible where analysts and engineers already search for data. More importantly, SAP governance tags can be used in Databricks ABAC policies, whic(databricks.com)tem. (docs.databricks.com) ### Why is this an AI story too? Because AI systems are unusually sensitive to missing context. A model can summarize a table full of abbreviations, but if it does not know which fields are authoritative, how entities relate, or which columns are restricted, the output gets less reliable fast. Databricks explicitly pitched this launch as making SAP data(docs.databricks.com)get lost when enterprise data moves between platforms. (databricks.com) ### Is this one-way, or part of something bigger? It is part of a broader SAP-Databricks setup that is increasingly bidirectional. Databricks already supports live Delta Sharing access to SAP BDC data products, and SAP documentation describes ways to publish derived Databricks outputs back into SAP Business Data (databricks.com) shared metadata plane between the two systems. (docs.databricks.com) ### What is the catch? The catch is that the metadata is synced for mounted SAP BDC shares and exposed through Unity Catalog rather than as a directly queryable metadata feed for Delta Sharing recipients. That means the feature is powerful inside the Databricks governance layer, but it still depends on SAP BDC being the upstream owner of the semantics. In(docs.databricks.com)both sides. (docs.databricks.com) ### So what changed for enterprise teams? The practical change is simple. SAP data arriving in Databricks no longer has to arrive stripped of its labels, joins, and policy hints. That reduces one of the most expensive parts of enterprise analytics — rebuilding meaning after moving data. Basically, Databricks is trying to make the metadata travel with the table, becau(docs.databricks.com)databricks.com)