Dell Technologies is expanding its Dell AI Data Platform with a unified semantic layer, enterprise knowledge graph and topic-specific knowledge agents designed to give enterprise AI systems more reliable context. The company is also adding GPU-accelerated data processing, expanded multitenancy and storage benchmarking tools as it targets a persistent obstacle to production AI: making fragmented enterprise data searchable, contextualized, governed and ready for AI workloads.
Enterprise AI has a data problem
The biggest constraint on enterprise AI is increasingly not the model itself. It is the data surrounding the model.
Corporate information remains distributed across file systems, databases, cloud services and business applications. Much of that information was created before generative AI and AI agents became part of the enterprise software stack, meaning it was not necessarily structured, labeled or connected for machine reasoning.
Dell Technologies is betting that the infrastructure connecting that information to AI systems needs to become a platform of its own.
The company is expanding its Dell AI Data Platform with three major capabilities: a Unified Semantic Layer, an Enterprise Knowledge Graph and Knowledge Agents. Dell is also adding accelerated data processing and infrastructure-management capabilities designed to move enterprise information from storage into AI workflows more efficiently.
The strategy addresses a problem increasingly familiar to enterprise AI teams: retrieval-augmented generation can find relevant documents, but finding information is not the same as understanding how different pieces of business data relate to one another.
Giving enterprise data consistent meaning
The Unified Semantic Layer is designed to give structured and unstructured data common business definitions.
That matters because different enterprise systems often use different terminology for the same underlying concept. One application might refer to a customer as a “client,” while another calls the same entity an “account.”
Dell’s approach is to create a consistent layer of definitions, rules and searchable terminology that AI applications can use when interpreting enterprise information. Organizations can also import existing ontologies and classification taxonomies.
Dell says the platform will additionally enable NVIDIA Auto-Ontology, an open-source library designed to generate knowledge graphs from enterprise data.
The objective is straightforward: reduce the ambiguity an AI system encounters when it moves between different data sources.
That becomes increasingly important as enterprises move from single-turn chatbots toward AI agents capable of executing multi-step tasks. An agent may need to combine information from databases, documents, operational systems and historical records before deciding what action to take.
Without a common understanding of those entities and relationships, the agent has to reconstruct that context repeatedly.
Knowledge graphs add relationships to retrieval
Dell’s Enterprise Knowledge Graph is designed to address the second part of the problem: relationships.
Rather than treating enterprise information as isolated documents or database records, the knowledge graph maps relationships between structured and unstructured information. Dell says it uses metadata, lineage and query history to continually refine those relationships.
That could change how agents retrieve information.
Consider a manufacturing organization investigating an unexpected production-line problem. The relevant answer may not exist in a single document. It could require combining a sensor reading with information about a specific machine, its maintenance history, a supplier shipment and customer orders associated with the affected production run.
Traditional search can retrieve individual pieces of that information. A knowledge graph is intended to make the relationships between those pieces explicit.
This approach aligns with a broader evolution in enterprise AI architecture. Retrieval systems are increasingly moving beyond simple vector similarity toward combinations of structured metadata, knowledge graphs, hybrid search and other mechanisms for improving context.
Knowledge Agents operate on governed context
Dell is putting another layer on top of that data foundation with Knowledge Agents.
Each agent is designed around a defined business topic and operates against a specific portion of the Enterprise Knowledge Graph. Organizations can determine the data an agent can access, the rules it follows, the quality standards it must meet and the amount it is permitted to spend.
That makes the agents closer to governed enterprise assistants than unrestricted general-purpose chatbots.
Dell says NVIDIA Nemotron Retriever models provide retrieval, reasoning and visual-understanding capabilities for Knowledge Agents, while the platform is intended to remain independent of any single AI model, storage provider or data provider.
The architectural decision is important.
Enterprises increasingly operate multiple foundation models and AI services. Locking the entire data layer to one model can make it harder to change providers as model capabilities, costs and enterprise requirements evolve.
Dell is instead positioning the data platform as the persistent layer beneath those models and applications.
GPU acceleration moves closer to enterprise data
The platform expansion also addresses the performance side of the data problem.
Dell says its Data Processing Engine, using NVIDIA cuDF on NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, processed data nearly four times faster on average than CPU-only processing across a range of tests, with batch workloads reaching up to 20 times faster.
Those figures come from Dell’s own testing and should not be interpreted as universal performance benchmarks.
The platform uses Apache Arrow to move data between Dell storage and processing systems, with the aim of reducing data-preparation overhead and getting AI-ready information to accelerated compute faster.
This matters because AI infrastructure performance is not determined by GPU throughput alone. If GPUs spend significant time waiting for data to be transformed, retrieved or transferred, expensive accelerators can remain underutilized.
The data path therefore becomes part of the AI infrastructure equation.
Security becomes more important as context becomes richer
The more enterprise context an AI platform can access, the greater the security requirements.
Dell says its Unified Semantic Layer, Enterprise Knowledge Graph and Knowledge Agents remain inside the customer’s data center. The company is also expanding PowerScale to support up to 500 tenants in a single cluster, alongside mutual TLS over NFS and more granular role-based access controls.
For shared AI environments, multitenancy can allow multiple business groups or customers to use common infrastructure while maintaining data isolation.
That is particularly relevant for organizations operating internal AI platforms where different teams have different permissions and data-access requirements.
The objective is to make enterprise context available to AI without turning the data layer into a new security boundary that is difficult to manage.
Dell is selling the path to production, not just infrastructure
Dell is also expanding implementation services around the AI Data Platform.
That reflects a larger reality in enterprise AI: deploying infrastructure is often easier than connecting it to production data, governance processes and existing applications.
A production AI system needs reliable data pipelines, search, orchestration, monitoring, security and model-serving infrastructure. It also needs to fit the organization’s existing operational environment.
Dell is therefore positioning the platform as an end-to-end path from stored enterprise information to AI-ready context.
The broader market is moving in the same direction. AI infrastructure is increasingly becoming a combination of compute, storage, data processing, retrieval and governance rather than a standalone GPU cluster.
Context becomes the next AI infrastructure battleground
The next phase of enterprise AI competition may therefore be less about which model can generate the most impressive response and more about which infrastructure can provide the right context reliably and economically.
Dell’s expanded platform is an attempt to make that context persistent.
The semantic layer gives enterprise information common meaning. The knowledge graph connects related information. Knowledge Agents apply that context to defined business tasks. GPU-accelerated processing and storage infrastructure are intended to keep the entire pipeline moving.
None of those components guarantees that an AI agent will produce a correct answer. But together they address one of the fundamental weaknesses of enterprise AI: the gap between having enormous amounts of corporate data and giving AI systems a governed understanding of what that data actually means.
For enterprises moving AI agents from pilots into production, that gap could become one of the most important infrastructure problems to solve.
Market Landscape
Enterprise AI infrastructure is evolving beyond compute and model serving toward a full data-to-inference stack.
Dell’s expansion competes across several overlapping categories: enterprise storage, data processing, vector search, retrieval-augmented generation, knowledge graphs, AI agents and AI infrastructure services.
The emerging architecture increasingly combines structured enterprise data with unstructured documents, metadata, lineage and semantic relationships. This can help agents retrieve more relevant context while maintaining governance controls.
Dell’s partnership with NVIDIA is also significant because it combines storage and data infrastructure with GPU-accelerated processing and AI software. Competitors including AWS, Microsoft, Google Cloud, IBM and other infrastructure providers are pursuing similar convergence between enterprise data platforms and generative AI.
The competitive question is shifting from “Where do we run the model?” to “How do we reliably deliver the right enterprise context to the model?”
Top Insights
- Dell is expanding its AI Data Platform beyond storage with semantic definitions, knowledge graphs and governed agents for enterprise AI applications.
- The Unified Semantic Layer is designed to give different enterprise systems consistent meaning for shared business entities and terminology.
- The Enterprise Knowledge Graph maps relationships across structured and unstructured data, helping agents retrieve broader business context.
- GPU-accelerated data processing aims to reduce the time between enterprise data preparation and AI inference or model training.
- Dell’s architecture keeps core enterprise context within customer-controlled infrastructure while supporting multiple AI models and data sources.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
