Snowflake Expands Partnership with Microsoft to Improve Interoperability Through Apache Iceberg

As of November 2024, this feature is now in public preview. You can watch a demo, try it out, and read the documentation.

Today we’re excited to announce an expansion of our partnership with Microsoft to deliver a seamless and efficient interoperability experience between Snowflake and Microsoft Fabric OneLake, in preview later this year. This will enable our joint customers to experience bidirectional data access between Snowflake and Microsoft Fabric, with a single copy of data with OneLake in Fabric. Organizations using both platforms will be able to do so more cost-effectively, rather than building pipelines or maintaining copies of data in each platform. This interoperability is possible because of Snowflake’s and Microsoft’s commitment to supporting the industry’s leading open standards for analytical storage formats — Apache Iceberg and Apache Parquet.

This new integration means that organizations can make more cost-effective use of their data. For example, data from Snowflake can be more seamlessly integrated with Microsoft 365 apps, like Teams, Excel, PowerPoint and Outlook, via Fabric. And data from Fabric OneLake can be extended to Snowflake’s Data Cloud for cross-cloud AI, applications, sharing and collaboration, and advanced analytics, among many other workloads.

In order to deliver these benefits, Snowflake and Fabric will add these capabilities for customers now and later this year:

• Snowflake will be able to store data in Iceberg format in OneLake (official preview later this year).
• Fabric will be able to store data in Iceberg format in OneLake via Apache XTable translation in OneLake.
• Data written by either platform, Snowflake or Fabric, will be accessible from both the platforms.
• Snowflake will be able to read Fabric data artifacts in OneLake, stored physically or virtually through shortcuts.
• Data from Snowflake and Fabric can be used by a variety of services, such as Microsoft 365, Copilot in Microsoft 365, Snowflake Cortex AI, Azure AI Studio, Snowflake Copilot and more.

Let’s walk through an example of how this new integration works. Suppose you’re working on a customer loyalty project, but a loyalty members table is managed by a team in Snowflake while sales and support data is in Microsoft Azure. Rather than maintaining copies of data in each platform, you can use either to access a single copy of all data stored in Fabric OneLake.

First, create an external volume in Snowflake that points to your Fabric OneLake account. This is where Snowflake will store loyalty member data in Iceberg format. When you create this Iceberg table from Snowflake, you can specify to use Snowflake as the Iceberg catalog and Fabric OneLake as the external volume. As Snowflake operates on the tables and writes data, OneLake will automatically convert the Iceberg metadata to Delta Lake format, without rewriting the Parquet files, so that Fabric engines can query the same tables.

Similarly, Fabric OneLake enables Snowflake to read all OneLake data in Iceberg format, for consumption by Snowflake’s engine. This means you can create an Iceberg table from Snowflake that points to the Iceberg data in OneLake, which can then be queried. For example, you can use Snowflake Cortex AI LLM functions to calculate sentiment of support case notes or summarize call transcripts.

Snowflake and Microsoft will be developing this new integration collaboratively over the coming months. Stay tuned for more announcements at Snowflake’s Data Cloud Summit on June 3–6 in San Francisco. However, Iceberg Tables in Snowflake are currently available in public preview. If you haven’t already, you can try Iceberg Tables with this quickstart guide, or watch a demo to see how Iceberg is integrated with the Snowflake platform, like Cortex AI LLM functions.