Teradata expands sovereign AI with air-gapped cloud

Teradata has expanded its sovereign AI portfolio with new on-premises and air-gapped cloud offerings, including the availability of Teradata Sovereign Cloud on Google Distributed Cloud air-gapped.

The update also expands Teradata Factory with AI Studio, designed to let organisations run AI and data workloads inside customer-controlled infrastructure.

The announcement is aimed at governments, defence organisations, regulated industries and operators of critical infrastructure that need tighter control over where sensitive data is stored, where AI systems run and who manages the underlying environment.

Teradata's approach centres on two deployment models. Teradata Factory is an integrated hardware and software rack for on-premises use, while Teradata Sovereign Cloud on Google Distributed Cloud air-gapped is aimed at customers that want a cloud-style operating model in a fully isolated environment with no public internet connection.

In the on-premises setup, the expanded Factory with AI Studio environment brings together data, vectors, models and agents within customer-controlled infrastructure. Additions include Tera, its agentic coworker for enterprise data work; AgentStack for building and operating AI agents; model services for machine learning operations and model management; data pipelines for multimodal unstructured data and vector search; and analytics tools including native notebooks and Elastic Python Compute.

For the air-gapped option, Teradata provides its analytics layer within Google Distributed Cloud's isolated environment. This is intended to support workloads such as operational reporting, mission analysis, intelligence fusion, training and simulation analytics, and defence logistics within sovereign boundaries.

Growing demand

The expansion comes as businesses and public sector organisations face growing pressure to demonstrate control over data, infrastructure and AI systems. Teradata pointed to research showing that sovereignty requirements are being driven not only by data residency rules, but also by concerns about vendor dependence, operational continuity and regulatory compliance.

A 2026 BARC study of 320 business and IT decision-makers found that 62 per cent of organisations cited data and AI in core business processes as a key driver of sovereignty requirements. Teradata also cited IBM research showing that 91 per cent of executives said they did not fully understand their organisation's dependencies across AI vendors, models and infrastructure.

The company argued that sovereignty now extends beyond the location of stored data. It defines the issue across three areas: data sovereignty, operational sovereignty and digital sovereignty, covering data access, control of the operating environment and dependence on external technologies and AI models.

That distinction is becoming more important as AI is deployed in mission-critical operations. An organisation may meet data residency requirements but still fall short of full sovereignty if AI processing happens outside its boundaries, external administrators can access the environment, or services fail when public connectivity is lost.

Sumeet Arora, Chief Product Officer at Teradata, said the company sees sovereign deployment as an architectural issue spanning multiple environments rather than a single location.

"As AI becomes embedded in mission-critical decisions, organisations can no longer treat sovereignty as a data-location requirement. They need control over where AI runs, which models are used, who can operate the environment, and whether critical workloads can continue if external connectivity disappears. And increasingly, sovereignty is not a single location. It is an architecture that spans public cloud, sovereign cloud, and on-premises infrastructure. Teradata is uniquely built for that reality. We can run in all of those environments, connect them through Teradata Fabric without requiring data to move, and give organisations the freedom to choose the AI models and infrastructure that work for them," Arora said.

Google role

Google Distributed Cloud provides the air-gapped infrastructure element of the new offering. The arrangement is intended to give customers a way to keep classified or regulated data in a fully isolated environment while still running large-scale analytics and AI workloads.

The air-gapped deployment is operated by the customer and designed to meet strict regulatory and audit requirements. It is also intended to support complex queries across very large datasets with predictable performance.

Muninder Singh Sambi, Vice President of Product Management for Google Distributed Cloud, outlined Google's role in the arrangement.

"Google Distributed Cloud is purpose-built, air-gapped infrastructure that delivers complete technical, operational and software sovereignty. By combining Google Distributed Cloud's sovereign infrastructure with Teradata's AI and knowledge capabilities, we are giving enterprise customers in regulated industries the ability to innovate without sacrificing control over their data," Sambi said.

Teradata said the expanded sovereign features for Teradata Factory with AI Studio are expected to be available in the first quarter of 2027, while Teradata Sovereign Cloud on Google Distributed Cloud air-gapped is in public preview.