ClickHouse now writes Apache Iceberg tables to Microsoft OneLake
ClickHouse now writes Apache Iceberg tables directly to Microsoft OneLake, registering them in the OneLake catalog via REST APIs. This allows analytical results, including those from AI agents, to be distributed to Power BI and Fabric Spark. The feature closes a previous gap where object storage writes lacked catalog discovery, enabling secure, shared data access across organizational tools.
In late June 2026 ClickHouse added the ability to write Apache Iceberg tables to object storage, but those tables could not be discovered by other tools without catalog registration. At OpenHouse 2026 the company announced that ClickHouse can now write Iceberg tables directly to Microsoft OneLake, registering them through the Iceberg REST‑based Table APIs. The write operation uses the same DataLakeCatalog connection that already supports reads, so no extra configuration is required beyond the existing service‑principal credentials. The new capability closes the gap between fast analytical queries in ClickHouse and secure, governed data sharing across the Microsoft Fabric ecosystem. Results produced by AI agents or applications can be published to OneLake where they appear automatically in Power BI, Fabric Spark and any other Iceberg‑compatible engine. Because OneLake Security controls apply from the moment data land, organizations gain a unified governance model without an additional ingestion layer. It remains unclear how performance will compare with the previous object‑storage‑only approach, as ClickHouse has not released benchmark data. The announcement does not specify whether future updates will support multi‑namespace tables beyond the back‑ticked workaround. Any limitations of the Iceberg REST APIs themselves are also left uncertain by the presenters.