ClickHouse previews Arrow Flight connector for Power BI
ClickHouse's public-preview Arrow Flight SQL connector gives the Power BI service a direct path to the database, while DirectQuery support is still maturing.
ClickHouse has opened a public preview of an Arrow Flight SQL connector for Microsoft Power BI. The connector allows the Power BI service to connect directly to ClickHouse without an on-premises data gateway or a separately installed driver. ClickHouse describes it as open source.
For supported imports, the Arrow Flight connector preserves a columnar path between the database and Power BI. It uses the Flight SQL ADBC driver shipped with Power BI to open a TLS-encrypted gRPC connection to ClickHouse, send SQL queries and return results as Arrow record batches. Apache Arrow defines Flight SQL as a protocol for interacting with SQL databases through the Arrow in-memory format and Flight RPC framework.
That differs from ClickHouse’s existing ODBC connector, which remains supported. Microsoft’s documentation for the generally available connector says Power BI cloud-service connections require an on-premises data gateway because the ODBC driver is an extra component. Microsoft’s ADBC guidance says its open-source Flight SQL driver ships with Power BI Desktop and can return columnar Arrow batches without a separate driver installation.
ClickHouse says the preview supports query folding for filters, sorts, group-bys and inner joins, along with native SQL queries. The company also says Decimal and UInt64 values reach Power Query without an intermediate floating-point conversion. According to ClickHouse, every connection uses TLS with certificate validation, and there is no plaintext mode.
The preview still has several availability boundaries. ClickHouse Cloud customers must ask support to enable Arrow Flight for their service. Self-managed deployments require ClickHouse 26.5 or later, a configured Arrow Flight port, and a TLS certificate and key. ClickHouse says Import mode is currently the connector’s strongest mode, while DirectQuery support is still maturing. A separate Unity Catalog metric views integration for Power BI provides context for another preview-stage DirectQuery path.
ClickHouse also published its own performance comparison, which has not been independently reproduced. Across three alternating runs, each connector imported the same 10-million-row, 20-column table into a shared-capacity Power BI workspace. ClickHouse reported a median refresh time of 278 seconds through Arrow Flight, compared with 390 seconds through ODBC. It also reported about five seconds of ClickHouse query time versus about 150 seconds, and 5.2 seconds of ClickHouse CPU per load versus 32.4 seconds.
The test used ClickHouse 26.11 with three 16 GiB replicas in AWS us-west-2 and Power BI in Azure West US. The ODBC route ran through a 4-vCPU, 16-GB gateway virtual machine in Azure West US 2, using the October 2026 gateway build and ClickHouse ODBC driver 1.5.5. ClickHouse said both paths loaded identical data.
When the research was conducted, Microsoft’s public ClickHouse connector page and the repository linked from ClickHouse’s announcement still documented the older ODBC connector rather than a separate Arrow Flight preview.
More news

Databricks opens Genie ZeroOps public beta with pull-request workflow

Polars brings distributed queries to existing Ray clusters

Apache Airflow 3.4.0 adds bounded task loops and AI agent workflows
