Teradata links Autonomous Knowledge Platform to OneLake

Teradata has integrated its Autonomous Knowledge Platform with Microsoft OneLake, allowing customers to run Teradata AI workloads on data stored in OneLake.

Customers can now read OneLake tables through Teradata, with the integration designed to avoid copying or moving data between systems before analysis.

The arrangement is built on Apache Iceberg, an open table format used across data platforms. Customers can query OneLake tables in place through standard Iceberg application programming interfaces, while authentication and access controls are handled through Microsoft Entra ID.

That means organisations using both Teradata and Microsoft Fabric can analyse data across the two environments without first consolidating it into one system. It also removes the need for separate extract, transform and load pipelines when teams would otherwise maintain duplicate datasets for analytics and AI work.

Data access

Moving data between platforms has long been an operational issue for companies building AI systems at scale. Data transfers can add storage costs, create delays, and force engineers to recreate security rules and governance processes for each new copy of a dataset.

The new link is intended to let data remain in OneLake while Teradata reads and analyses it directly. This should reduce the time data engineers spend maintaining pipelines and allow AI teams to begin feature engineering without waiting for data replication.

The announcement also reflects a broader push by technology suppliers to support more open data architectures, rather than requiring customers to move large datasets into a single proprietary environment. Apache Iceberg has become a key part of that approach because it offers a standard way to manage large analytic tables across different systems.

For Microsoft, the integration adds another route for enterprise customers to work with data held in OneLake, the storage layer within Microsoft Fabric. For Teradata, it gives existing customers a way to work with Microsoft's data environment without shifting information out of OneLake.

Governance focus

Governance remains attached to the data rather than being rebuilt for replicated copies. In practice, that addresses one of the more difficult aspects of large AI deployments, where data lineage, access permissions, and policy enforcement often become harder to manage as information spreads across multiple tools and clouds.

Teradata described the OneLake integration as part of its broader effort to let customers run AI and analytics where their data already sits. It has been promoting its Autonomous Knowledge Platform as a framework for governed AI use across hybrid and multi-cloud environments, with an emphasis on business context and operational controls.

The latest step is aimed at customers with data estates split between Teradata systems and Microsoft platforms. Those organisations often face a trade-off between centralising data for analysis and leaving it in place to reduce movement, cost, and governance overhead.

The current release covers read access to OneLake tables through Teradata. The announcement did not describe write-back functions.

"Moving data to analyse it is a constraint that adds cost and complexity without adding value. Bringing Teradata enterprise AI directly to Microsoft OneLake means customers can run the workloads that matter on the data where it already lives, inside the governance model they've already built. This is a concrete step in making the Autonomous Knowledge Platform work wherever enterprise data exists," said Sumeet Arora, chief product officer at Teradata.

Microsoft framed the link as an example of interoperability within its Fabric ecosystem.

"Microsoft OneLake is built to be the unified data foundation for the AI era, connecting organisations with the platforms, tools, and data ecosystems they already trust. Teradata's integration through Apache Iceberg exemplifies the openness and interoperability at the heart of Microsoft Fabric, enabling customers to bring their data together and accelerate their AI transformation without the complexity or rearchitecting around a single vendor," said Dipti Borkar, vice president of Microsoft OneLake and ISV Ecosystem at Microsoft.