Is your Data Intelligence Platform Truly Open?

A Comparison of Openness & Interoperability of Snowflake & Databricks

By:  James Dinkel

As technology cycles accelerate and AI reshapes entire industries, openness and interoperability have become foundational requirements for any enterprise-ready data intelligence platform. Organizations must ensure that their enterprise platform delivers interoperability and consistent governance across engines, avoids vendor lock-in, and enables seamless sharing. While Databricks has historically been associated with open-source technologies (Apache Spark), its recent architectural direction increasingly relies on proprietary components. Snowflake, by contrast, has aligned deeply with open standards—particularly through Apache Iceberg™, Apache Polaris™, and leading the charge on Open Semantic Interchange (OSI is a vendor-neutral standard for defining and exchanging business logic and semantics across data platforms)—resulting in a platform that is more open and interoperable, where it meaningfully impacts enterprise architectures. In this blog, we will compare Snowflake and Databricks across these dimensions of openness and interoperability:

  1. Open Table Formats
  2. Open, functional, and interoperable catalogs
  3. Open data and AI Sharing
1. Open table formats

An open table format is a vendor-neutral metadata and storage specification that enables any compute engine to read and write data in cloud object storage (S3, ADLS, GCS) without being locked into a single vendor

Snowflake (4 Stars): Snowflake’s strategy is centered on an early and strong commitment to Apache Iceberg™ as a truly open table format. This strategy ensures that customers retain full control over their data location and access policies, while using the market’s most powerful and cost-efficient compute engine.

Databricks (3 Stars): Databricks continues to prioritize Delta Lake, which remains predominantly Databricks-led and not fully vendor-neutral. While it has added Iceberg features, its Iceberg support is still maturing, with several features in preview[1].

2. Open, functional, and interoperable catalogs

Open table formats require an open catalog. The catalog defines how open your governance and metadata are crucial for security, governance, and interoperability.

Snowflake (4 stars): The Snowflake Horizon Catalog implements open APIs (Iceberg REST Catalog, Apache Polaris) and adheres to open-source principles. This, in turn brings the following benefits:

  • Migrate to an OSS catalog anytime: Snowflake interoperability is powered by a Polaris instance within Horizon Catalog, enabling you to migrate to the leading open-source catalog for Iceberg and other formats.
  • Ability to use any engine: Integrate data from any system and make it interoperable with any compatible engine, ML platform, and AI agent with Snowflake Horizon Catalog’s embedded open APIs (Apache Polaris™ and Apache Iceberg™)
  • Ability to read/write to any Iceberg Table: Ability to process any Iceberg table, regardless of Catalog, for true bidirectional interoperability.
  • Iceberg catalog federation: Snowflake has championed the open lakehouse strategy with Catalog Linked Database, which provides read and write access to any foreign catalog via the Iceberg REST Catalog (IRC) Interface.
  • Comprehensive business continuity, security, and governance: Snowflake provides robust cross-region or cross-cloud disaster recovery capabilities, managed through an easy-to-use interface. This is layered on top of comprehensive security and governance features, including role-based access controls (RBAC), attribute-based access controls (ABAC), privacy controls, cloud security posture management, and immutable backups.

Databricks (2 stars): Databricks has two catalogs: a proprietary Databricks catalog and an open-source version (Unity Catalog Open Source). However, these catalogs are materially different from each other, which can introduce complexity and make it harder for customers to adopt a unified open-catalog strategy:

  • Databricks Unity Catalog is not open, and there’s no easy way to switch from Databricks Unity Catalog to the open-source one – Unity Catalog Open Source (OSS)
  • Limited read/write capabilities: Customers cannot write from Databricks engines to Iceberg Tables and Delta Tables managed by remote catalogs. Moreover, external engines cannot write to Delta Lake tables managed by Databricks Unity Catalog (Iceberg writes are supported).
  • Iceberg catalog federation limitations: Unity Catalog has a strong federation story; however, it only supports reading from foreign sources, and does not permit full interoperability with reading or writing to Iceberg REST Catalog sources (no outbound IRC federation)
  • Incomplete governance: Unity Catalog OSS currently lacks key governance and security features, including ABAC, RBAC, and row- and column-level access policies[2]. Customers must use the proprietary Databricks Unity Catalog for meaningful security and governance, which, as detailed in my previous blog, still lags the Snowflake Horizon Catalog in managed cross-cloud disaster recovery, advanced security and privacy controls, and cyber defense features. Link to enterprise ready security blog:  https://squadrondata.com/enterprise-ready
3. Open data and AI sharing

True openness allows external engines to use your data easily without duplication.

Snowflake (5 stars): Snowflake Sharing is open to all, allowing sharing of data and AI assets, even with non-Snowflake customers, and supporting open table formats like Iceberg and Delta. Snowflake also provides a managed and secure shared area that simplifies cross-cloud and cross-region sharing of open table formats, without unpredictable egress costs. Combining this with Delta Direct and Catalog Federation extends Snowflake’s open table sharing to Delta Lake tables, including those written by Delta engines (e.g., Databricks) or managed by non-Snowflake catalogs. This enables truly interoperable, open sharing across table formats and catalogs.

Databricks (2 stars):Delta Sharing Open Source (OSS) offers limited value as it only supports raw data sharing, lacking essential metadata exchange for understandable, searchable, and reusable data. Only within Databricks can metadata be shared and customers must use the Databricks version of Delta Sharing for meaningful capabilities. And both versions of Delta Sharing lack automatic replication and incremental sync, making cross-region/cloud sharing tedious and expensive with manual setup and per query egress costs. Consequently, Delta Sharing is less suitable for open, scalable, multi-cloud data collaboration.

Summary

For organizations designing future‑proof data and AI platforms, adopting technologies built on open standards ensures:

  • Greater flexibility to integrate best‑of‑breed tools and reduced lock‑in risk
  • More consistent governance and data portability
  • Interoperability across clouds, engines, and AI systems
  • Open standards reduce long-term architectural risk by ensuring that data, semantics, and governance can move with the business as technology evolves.

Based on our evaluation, Snowflake is significantly more open and interoperable, contributing to their overall enterprise-readiness. Databricks offers several capabilities but relies on platform‑specific components, which limit openness where it matters most. The evaluations and comparisons in this analysis reflect my professional opinion, based on my experiences, observations of both platforms, and an interpretation of the available information.  Do you have any specific questions regarding the differences in openness or interoperability between the two platforms? Please contact us at info@squadrondata.com.

1 https://docs.databricks.com/aws/en/iceberg/

2 https://github.com/unitycatalog/unitycatalog/discussions/411