Teradata has announced the general availability of Autonomous Knowledge Platform, an enterprise platform that brings together data, analytics, models, and AI agents into a single architecture. The solution can be deployed on Amazon Web Services (AWS), on the client’s data center, or via a hybrid setup, with a focus on data governance, technological sovereignty, and control over agent-generated consumption.
The key points of Teradata’s AI platform in 30 seconds
- Teradata Cloud is now available on AWS with persistent capacity and on-demand elastic resources.
- Teradata Factory brings software and hardware solutions to private facilities.
- AI Studio consolidates models, agents, vector services, analytics, and governance within a single environment.
- Organizations can use their own models, commercial models, or open weights models.
- Teradata offers a flat rate with flexible capacity, though prices and independent comparisons have not been published.
The launch completes the schedule introduced by the company in May 2026. At that time, Teradata indicated its cloud platform and Factory on-premises system would arrive in Q3. The general availability announced on 07/15/2026 confirms that both deployment models are now available for purchase, with the cloud version of this architecture specifically starting on AWS.
The term “autonomous knowledge” is part of Teradata’s commercial terminology. It does not mean that data are managed or interpreted correctly without human intervention. Instead, it describes a layer that combines structured and unstructured information with semantics, lineage, permissions, and operational knowledge so that agents can work with the enterprise context, rather than just acting on isolated documents or model-generated responses.
One platform with two deployment options
Autonomous Knowledge Platform revolves around four main components: Teradata Cloud, Teradata Factory, AI Studio, and AI Services. Each covers a different part of the journey from enterprise data to autonomous applications.
| Component | Deployment Announced | Main Function |
|---|---|---|
| Teradata Cloud | AWS | Managed data, analytics, AI, and elastic capacity |
| Teradata Factory | Private data center | On-premises AI and analytics with CPU, GPU, storage, and networking |
| Teradata AI Studio | AWS and Factory | Building, deploying, and governing models and agents |
| Teradata AI Services | All deployments | Consulting, integration, and production rollout |
Teradata Cloud combines two types of capacity. Active Compute is meant for persistent and sensitive workloads like operational queries, reporting, or processes requiring stable service levels. Elastic Compute offers on-demand resources for experimentation, training, ad-hoc analysis, or temporary activity spikes.
This separation aims to prevent exploratory AI loads from directly competing with critical business processes. It also aligns with the nature of agents: a human may run a few queries during a day, while an agent can continuously chain searches, vector operations, model calls, and database queries.
Teradata has termed its new commercial structure as Fixed plus Flex. Customers purchase a predictable base and allow capacity to automatically increase during demand peaks. Both consumption modes are expressed through a unified unit to simplify procurement and budgeting.
The company has not published pricing, minimum commitments, or costs for extensions. Nor has it provided a comparison to determine if this model is more economical than paying per query, reserving capacity, or other cloud consumption options. Actual utility will depend on workload stability and how much work can be shifted to the flexible level.
Teradata Cloud offers workload isolation, integration with corporate identity systems, and open table format compatibility. Its Connected Data Foundation supports Apache Iceberg and Delta Lake, enabling different engines to work on shared data without redundant copies.
Although Teradata’s main website mentions AWS, Microsoft Azure, and Google Cloud as available platforms for the product, the general availability announcement for Autonomous Knowledge Platform initially places the new cloud capabilities on AWS. Therefore, it should not be assumed that all newly announced functions are immediately enabled across all three hyperscalers with identical scope.
Teradata Factory brings models into a private environment
Teradata Factory offers an alternative for organizations that prefer or need to keep data and models on-premises. It is delivered as an integrated hardware and software system for running data warehouses, lakehouse architectures, advanced analytics, and AI workloads within the customer’s facilities.
The announced configuration uses Dell PowerEdge servers, NVIDIA AI infrastructure, GPU acceleration, enterprise storage, and high-speed networking. Teradata presents Factory as a ready-to-run system rather than just a set of licenses to be assembled by the customer.
The company has not disclosed specific configurations, GPU models, minimum capabilities, power requirements, or prices. Nor can it be assumed that any language model can run “without compromising performance” in any setup, as the communication suggests. Model size, quantization, available memory, user count, and latency requirements will influence the necessary infrastructure.
Factory supports the Bring Your Own Model approach, allowing organizations to supply their own models—whether developed in-house, commercially licensed for private deployment, or open-source weights.
This flexibility prevents the platform from being tied to a single model provider by design but entails considerable integration work. Different models have various formats, inference requirements, context limits, and licensing conditions. Organizations must also validate quality, security, maintenance, and inference behavior over their internal data.
The main benefit of on-prem deployment is especially relevant in regulated sectors or environments with sensitive data. Banks, government agencies, infrastructure operators, insurers, and healthcare organizations can keep data, embeddings, logs, and inference processes within their own perimeter.
While this setup helps control data residency, it does not guarantee sovereignty by itself. License terms, telemetry, manufacturer access, updates, hardware supply chain, and reliance on proprietary technologies also need consideration.
The hybrid approach allows combining Factory with Teradata Cloud—organizations might store certain data and models locally, using cloud resources for other workloads. Teradata aims to provide consistent identity, policy, lineage, and governance across environments. Effectiveness will depend on real-world implementations, especially when networks, catalogs, or regional regulations differ.
AI Studio consolidates AI development tools often deployed separately
AI Studio is the platform where teams build, deploy, and monitor AI applications. Teradata states it is already available on AWS and within Factory, and can also be individually contracted for existing environments.
It integrates models, agents, vector services, analytics, notebooks, model operations, and access controls. The documentation highlights three key components: Enterprise MCP Server, Agent Builder, and the management systems needed to productionize agents and data flows.
The Model Context Protocol (MCP) enterprise server provides tools through a common protocol that enables agents to query data or perform operations. In a corporate setting, the value lies not just in connectivity but in controlling which user made the request, permissions inherited, data accessed, and actions executed.
AI Studio supports both visual development and coding approaches. Official documentation mentions compatibility with frameworks like LangChain and LangGraph, deployment via containers on Kubernetes, role-based access control, audit logs, and identity propagation.
The platform also includes vector services for combining semantic searches with structured data. Teradata calls this fusion search: the joint retrieval of tables, documents, and other content, while hybrid search blends vector similarity with keyword matching. The aim is for an agent to relate, for example, a policy in a PDF to customer data stored in a table.
This consolidation can reduce the number of tools an enterprise needs to integrate. It also expands the scope of Teradata’s platform. Centralizing data, models, agents, governance, and execution simplifies operations but may increase costs for future migrations, even with open formats.
Part of the announcement also involves rebranding existing products. Teradata Vantage is now called Autonomous Knowledge Platform; VantageCloud becomes Teradata Cloud; ClearScape Analytics and AI Workbench are grouped under AI Studio; QueryGrid is renamed Teradata Fabric, and IntelliFlex and AI Factory evolve into Teradata Factory.
Thus, the platform combines new capabilities with an overhaul of previous offerings. The general availability indicates that not everything was rebuilt from scratch—Teradata has reorganized its data technology around AI agents and added components for models, agents, elasticity, and private deployment.
The announcement does not include reference customers, financial results, performance comparisons, or savings figures. The value proposition must demonstrate that an integrated platform can better control agent costs associated with continuous queries without becoming a dependency that’s hard to reverse.
Frequently Asked Questions
What is Teradata Autonomous Knowledge Platform?
It is Teradata’s new flagship platform that consolidates data, analytics, models, and AI agents. It can be deployed in the cloud, on private infrastructure, or as a hybrid setup.
Which cloud providers is the new platform available on?
The general availability announcement initially places Teradata Cloud on AWS. Although the company offers other cloud products on Azure and Google Cloud, it has not yet confirmed the same immediate scope of new functions across all these platforms.
Can models be run locally in data centers?
Yes. Teradata Factory supports custom, commercial, and open weights models within customer hardware, provided the resources meet model requirements.
Has Teradata published pricing for the platform?
No. The company explained its Fixed plus Flex structure with a base and elastic capacity but has not disclosed rates, minimum commitments, or example costs.
