Snowflake opens live Iceberg data to external REST Catalog clients
Snowflake's Open Data Sharing lets users outside its platform query supported Snowflake-managed Iceberg tables through REST Catalog-compatible clients, with governed, read-only access to live data.
Snowflake has made Open Data Sharing generally available, allowing providers to expose supported Snowflake-managed Apache Iceberg tables to external users through Iceberg REST Catalog-compatible clients, including Apache Spark.
The read-only route brings live data into consumers’ existing analytics or AI tools without requiring a Snowflake account. Snowflake presents the feature as an alternative to scheduled or manual exports. In the normal access path, providers do not need to maintain a separate consumer-facing ETL pipeline or data copy. Cross-region delivery is a qualified case: Snowflake automatically replicates listing data into the external consumer’s region.
The rollout has two dates. Snowflake’s release notes place the move out of preview on November 2, while the company published its product announcement on November 16. One external listing can serve Snowflake users and open clients, so providers do not need a separate share or listing for the latter.
Under the provider workflow, the provider creates a restricted external-consumer identity, prepares eligible data, grants it through a share and external listing, then sends the consumer the catalog connection values. The provider retains control in Snowflake and can revoke access. Consumers can authenticate through workload identity federation or a provider-managed programmatic access token.
The current table boundary is narrow. External clients can see only Snowflake-managed Iceberg tables identified by CATALOG = 'SNOWFLAKE'; standard Snowflake tables and externally managed Iceberg tables included in the share remain invisible. Snowflake says providers can also share supported dynamic Iceberg tables. They can materialize supported data into a dynamic Iceberg table for sharing, including filtered data derived from Snowflake-native tables.
Snowflake’s connection guide lists AWS, Azure, Google Cloud and OIDC as supported workload identity provider types. Its Spark example requires vectorized Iceberg reads to be disabled. Cross-region external consumption is limited to AWS and Azure commercial regions. Google Cloud consumers must be in the provider account’s region, US government regions are unsupported, and failover and disaster recovery are not yet available for external consumers.
More news

Snowflake makes CoCo’s automated Snowpipe setup generally available

Oracle sets 12-month window for AI Lakehouse semantic layer

Databricks makes Apache Spark 4.3 available in Runtime 19
