Snowflake boosts governed data for AI with Iceberg

Snowflake has introduced a new framework for interoperable data and artificial intelligence across enterprise systems, designed to let customers work from a single governed copy of data.

The update centres on broader support for Apache Iceberg and new governance tools in Snowflake Horizon Catalog. The aim is to reduce the need for businesses to copy data between cloud services, data lakes and other platforms. Snowflake now also offers general availability support for Apache Iceberg v3 and Snowflake Storage for Apache Iceberg Tables.

The move reflects growing pressure on companies to manage data spread across multiple systems while deploying AI tools that depend on reliable, consistent information. Many businesses still duplicate and move data between platforms before it can be analysed or used in applications, adding cost and complexity.

Snowflake's latest products are intended to let organisations access, govern, share and use data across Snowflake, external lakes and open systems without creating multiple copies. Horizon Catalog, powered by Apache Polaris, serves as a central layer for discovery, access control and monitoring across data held both inside and outside Snowflake.

Open standards

A key part of the announcement is Snowflake's deeper adoption of Apache Iceberg, the open table format that has become an important standard for managing large analytical datasets across different processing engines. By backing Iceberg v3, Snowflake is positioning itself more firmly within an open data ecosystem at a time when customers are wary of vendor lock-in.

Snowflake Storage for Apache Iceberg Tables gives customers a managed way to handle open-format data while keeping it accessible across different tools and environments. Horizon Catalog also allows bi-directional read and write access to Snowflake-managed Iceberg tables from external engines through standards-based controls.

That matters because many large companies now use a mix of cloud warehouses, external lakes, operational systems and specialist analytics tools. In those environments, a common complaint is that governance policies, metadata and access controls become inconsistent once data moves between systems.

Snowflake is extending governance controls to external engines through features including external engine access management and support for the Iceberg REST Scan Plan API. Customers will also be able to apply data protection policies such as column masking and row-level access controls across Iceberg-compatible engines.

Customer use

Snowflake pointed to deployments by Affirm, NTT DOCOMO and Samsung Ads as examples of how the model is being used in production. Those customers are adopting the software to simplify data estates and create a more consistent base for AI projects and decision-making.

"Most organizations still rely on moving and duplicating data just to make it usable, and that approach simply cannot keep up with the pace of AI. As innovation accelerates, data fragmentation becomes the constraint," said Christian Kleinerman, EVP of Product, Snowflake.

"We are fully committed to interoperability and openness. With Snowflake's capabilities, we are ushering in a new model for enterprise data, where customers can work directly on live, governed data wherever it resides through a single, connected governance plane. By eliminating duplication and defining shared business meaning through semantic views, we're establishing a consistent, trusted foundation for both teams and AI agents," Kleinerman added.

Affirm said it has used Snowflake to reduce duplication across systems while applying governance more uniformly. The financial technology group also said it migrated thousands of tables and key financial workloads to Polaris.

"At Affirm, delivering transparent and responsible financial products starts with having a clear, consistent view of our data," said Vivek Anandpara, VP of Engineering, Affirm.

"Snowflake enables us to work across systems without duplicating data, while applying governance consistently across our environment. This gives our teams a trusted foundation to move faster, improve decisioning, and scale AI with confidence. Our migration of thousands of tables and critical financial workloads to Polaris using Snowflake's interoperable and governed data foundation proved that out - Snowflake partnered closely with us to deliver zero-downtime correctness at scale," Anandpara said.

Snowflake also said customers can connect to data from enterprise software providers including SAP, Salesforce and Workday without replication, alongside newer links with AVEVA and IBM. It added that its CoCo coding agent now includes a dedicated skill for SAP data, while natural-language querying can extend across Snowflake, external data lakes and some external relational database systems.

Governance push

The governance element is central to Snowflake's strategy as businesses seek to let internal teams and software agents query data directly in natural language. If data definitions, permissions and quality checks differ across systems, the results generated by AI models can quickly become unreliable.

Snowflake said Horizon Catalog is intended to become a universal catalogue for enterprise data, unifying governance, policy enforcement and security controls across multi-catalog environments. It also introduced what it described as connected audit access and new observability for externally managed Iceberg tables, aimed at giving customers a broader view of data activity and pipeline health.

NTT DOCOMO said the approach helps it handle growing data and AI workloads without adding operational friction. Samsung Ads described similar needs in advertising, where data often sits across multiple internal and external systems.

"As we expand our data and AI initiatives, it's critical that we can work across systems without adding complexity," said Yoshio Umezawa, Vice President General Manager of Service Innovation Department R&D Innovation Division, NTT DOCOMO, INC.

"Snowflake allows us to access and govern data wherever it resides, while maintaining a consistent, trusted foundation. This helps our teams move faster, reduce operational overhead, and deliver more intelligent services to our customers," Umezawa said.

"At Samsung Ads, delivering relevant and measurable advertising experiences depends on having seamless access to trusted data across a complex ecosystem," said Hervé Marcellini, Vice President of Engineering, Samsung Ads.

"Snowflake enables us to work across systems without duplicating data, while maintaining consistent governance throughout our environment. This allows our teams to move faster, improve targeting and measurement, and scale AI-driven innovation with confidence," Marcellini said.