Dell Technologies is adding a semantic layer, an enterprise knowledge graph and a set of task-specific agents to its AI Data Platform, aiming to give AI agents reliable business context instead of raw access to data, SiliconANGLE reported Tuesday. The update also brings Nvidia GPU acceleration to Dell's data processing engine and new multitenancy features for its PowerScale storage, The Next Platform reported.

What Happened

The Unified Semantic Layer applies one set of business definitions and rules across applications. SiliconANGLE's example: it can recognize that a "client" in one system and an "account" in another refer to the same customer, so an agent does not treat them as different things.

The Enterprise Knowledge Graph maps how data in different systems is connected, using metadata and lineage, so an agent can pull the related tables and vector indexes it needs. Nvidia's Auto-Ontology library is among the components used to build it.

RelatedOpenAI Says Its Models Broke Into Two Outside Data Sites and Left It Out of Their Answers

Knowledge Agents are topic-specific agents that work on assigned slices of data under configurable rules, access controls and spending limits. They use Nvidia's Nemotron Retriever models.

On the processing side, Dell integrated Nvidia's cuDF library into its Data Processing Engine. In Dell's tests on PowerEdge R770 servers with Nvidia RTX PRO 4500 Blackwell Server Edition GPUs, using default settings, Apache Spark jobs ran 3.9 times faster on average than on CPUs alone and up to 20.4 times faster on batch data mining, according to SiliconANGLE.

PowerScale now supports up to 500 tenants per cluster, with tighter role-based access control and mutual TLS for authentication and encryption over NFS. Dell is also releasing an open-source Storage Performance Tool for benchmarking ObjectScale and PowerScale with the methodology its own engineers use.

Why It Matters

Dell is arguing that the limiting factor for enterprise AI has moved from compute to data. Varun Chhabra, a senior vice president in Dell's Infrastructure Solutions Group, told The Next Platform that the real bottleneck is access to data, not compute. He described a "pilot reproduction gap," in which projects that work as small pilots fail at enterprise scale because of data quality and governance problems.

Dell cited its own survey of 3,800 IT decision-makers and AI experts, which ranked data quality, availability, management and security as the top challenge in adopting AI.

The semantic layer and knowledge graph also move Dell, best known for servers and storage, further up the stack into territory where data platforms such as Snowflake and Databricks compete to supply the business context agents run on.

The Rest of the Stack

The platform's storage layer combines PowerScale network-attached storage, ObjectScale with S3-over-RDMA, and the Lightning File System, a parallel file system Dell announced at Nvidia GTC in March. Nvidia components include NIM microservices, CUDA libraries, the cuVS vector search library and the Nemotron Retriever models, with Apache Arrow used to move data, The Next Platform reported.

What to Watch

Availability is staggered. The Storage Performance Tool and related services are available now, PowerScale multitenancy and security features follow in November, and the accelerated Data Processing Engine in December, according to SiliconANGLE. The semantic layer, knowledge graph and Knowledge Agents are not due until the first half of 2027. Dell did not disclose pricing.