ClickHouse in Fabric: workload preview and OneLake Iceberg read GA

ClickHouse announced four product milestones with Microsoft on 2026-09-29 in Barcelona at The European Microsoft Fabric + SQL Community Conference (Alex Francoeur and Aditya Chidurala): a native ClickHouse workload for Microsoft Fabric in public preview, generally available reads of Apache Iceberg tables from OneLake via OneLake Table APIs, public preview Iceberg writes to OneLake (Table APIs with credential vending), and ClickHouse Bring Your Own Cloud (BYOC) for Azure through the Microsoft Marketplace (announcement).

This is cloud / Fabric / OneLake interoperability—not the embedded chDB Durable Layer flush-and-checkpoint story.

Stages at a glance

Surface	Stage
ClickHouse workload for Microsoft Fabric	Public preview (Fabric workload hub)
OneLake Iceberg read (OneLake Table APIs)	Generally available
OneLake Iceberg write (Table APIs + credential vending)	Public preview
ClickHouse BYOC (Azure)	Available on the Microsoft Marketplace — not a new GA of ClickHouse Cloud itself

Fabric workload (public preview)

The workload is available in public preview through the Fabric workload hub: ClickHouse as a query engine inside Fabric aimed at accelerating OneLake queries. ClickHouse frames the benefit as sub-second responses without provisioning a separate cluster or maintaining an ETL pipeline—vendor claims, not independent aitamer measurements (announcement; workload detail).

A related workload path can sync OneLake tables into a dedicated ClickHouse Cloud service as accelerated copies (ClickPipes / point-in-time snapshots in the product write-up). That copy path is not the same as the in-place Iceberg Table-API read below—“no duplication / no ingestion” applies to the Iceberg read path, not the workload lead (workload detail).

OneLake Iceberg read (GA) and write (preview)

Read (GA): ClickHouse can query Iceberg tables managed in OneLake directly via OneLake Table APIs. For this path, ClickHouse states there is no data duplication and no ingestion step (announcement).

Write (public preview): Writing Iceberg tables to OneLake via the same Table APIs with credential vending, so results can land as governed open-format tables other Fabric engines can read (write detail).

BYOC on the Marketplace

ClickHouse BYOC is available to Azure customers through the Microsoft Marketplace: managed ClickHouse Cloud inside the customer’s Azure tenant, with customer-owned IAM, network, and encryption controls as ClickHouse describes them (announcement).

ClickHouse also says Marketplace procurement can count toward Microsoft Azure Consumption Commitments (MACC) under existing agreements—vendor claim; Microsoft notes MACC eligibility is offer- and purchase-path specific.

Who should care

Teams already on Fabric / OneLake who want ClickHouse speed against lakehouse Iceberg data should start at the ClickHouse × Microsoft announcement—lead with workload preview and Iceberg read GA, keep write preview and BYOC Marketplace as supporting milestones, and keep each stage’s wording intact.