Teradata (NYSE: TDC) announced the general availability of Elastic Python Compute within AI Studio, a managed distributed Python environment designed to scale AI and machine learning workloads without requiring data to leave the Teradata platform.

The capability is built on Ray, a distributed computing framework, and integrates with the Teradata Autonomous Knowledge Platform. It supports parallel model training, custom Python application execution, and ML workload serving across single-server and cluster deployments. The feature is available on AWS and Microsoft Azure, with Google Cloud availability expected in the fourth quarter of 2026 and on-premises support available now through Teradata Factory.

Elastic Python Compute reads and writes data directly to the Teradata database, S3 object storage, and open table format catalogs including Apache Iceberg and Delta Lake. The platform also supports Arrow Flight SQL for parallel data transfer, which Teradata states is up to 84 times faster than traditional row-based methods based on its own benchmark testing.

According to the company, a retail bank using the platform reduced credit scoring pipeline runtimes by three times while keeping data within its governance perimeter. A telecommunications provider reduced feature engineering runtime from four hours to 20 minutes and reported a 12% improvement in model accuracy after training on full datasets instead of constrained samples. A government maritime agency used the tool to fuse satellite imagery, radar, and AIS data to track vessel movements.

"Elastic Python Compute removes that constraint. Teams can now run distributed AI workloads at scale where their data already lives," said Sumeet Arora, Chief Product Officer at Teradata.

The announcement was made at the company's San Diego location and is based on a press release statement from Teradata Corporation.