Teradata launches on-premises AI system for hybrid use
Tue, 19th May 2026 (Today)
Teradata has introduced Teradata Factory, extending its Autonomous Knowledge Platform to on-premises deployments for hybrid AI and analytics environments.
The system combines Teradata's full software stack with Dell compute and storage under a single management plane. It is designed to run enterprise data warehousing, lakehouse and AI workloads, including AI Studio, on one integrated system.
The launch reflects growing demand from companies that want to keep more AI workloads closer to their data, particularly in sectors with strict governance and data residency requirements. Teradata is targeting organisations that already operate across cloud and on-premises infrastructure and want a more unified way to manage both.
Hybrid set-up
Teradata says the system is pre-engineered, supports CPUs and GPUs, and can scale modularly from pilot projects to larger production deployments. It also supports Apache Iceberg, Delta Lake and S3-compatible object storage, with the aim of reducing lock-in and enabling data access across cloud and on-premises environments.
The product also includes autonomous management tools called Tera agents, intended to handle infrastructure and operational tasks such as monitoring compute resources, query execution, telemetry processing and spend control across cloud and on-premises estates.
Teradata presents the product as an on-premises foundation for what it describes as an AI-native, agentic enterprise. In its view, autonomous knowledge is software that can turn structured and unstructured data, operating models and organisational experience into governed decisions and actions with limited human intervention.
This comes as more companies assess whether generative AI, large language models and other advanced AI workloads should run in public cloud environments, private infrastructure or a mix of both. Cost, governance and reliability have become central questions as AI systems move from experiments into day-to-day operations.
Dell is the strategic collaborator for the on-premises deployment, providing the underlying compute and storage layer through its AI Factory and AI Data Platform. Teradata provides the software stack, management layer and customer support as a complete product.
Executive view
Sumeet Arora, chief product officer at Teradata, outlined the company's case for bringing AI processing closer to enterprise data.
"The data platform and the AI platform are converging - yet most enterprises are still running AI far from their most critical data. The Teradata Factory brings EDW reliability, Lakehouse flexibility, and AI horsepower together in a single on-premises system - so enterprises get the full performance of the Teradata Autonomous Knowledge Platform wherever their data, regulations, and agents require," said Arora.
Analysts and technology suppliers have increasingly highlighted data sovereignty as an architectural issue, not just a compliance one, especially for regulated industries and public sector bodies. Teradata used that argument to support the case for private AI deployments.
"Data sovereignty is evolving beyond just a compliance requirement. It is becoming a core architectural decision as AI moves from pilot to production. Enterprises are realising that where AI runs can be as important as how it runs. This on-premises deployment of the Teradata Autonomous Knowledge Platform can give enterprises a more direct path to run private AI on-premises, keeping it close to the data and under their governance, while maintaining the control, consistency, and performance needed at scale," said Kramer.
Cost pressure
One of the main claims behind the launch is that some AI workloads may become too costly or too difficult to manage if they rely only on public cloud services. Teradata argues that continuous inference, GPU usage and data-intensive analytics are changing the economics for businesses that need AI systems to run at production scale.
It also criticised on-premises approaches that require customers to assemble and manage separate components for storage, databases, GPUs, AI tooling and orchestration. Its alternative, it says, is a single product with one management plane that combines hardware and software in a packaged deployment.
Teradata says active system management in the new set-up is intended to protect core analytics workloads while AI teams run more experimental or resource-intensive tasks. That distinction may appeal to businesses trying to prevent AI projects from disrupting existing reporting and operational systems.
Teradata expects Teradata Factory to become available in the third quarter of 2026.