Dell expands AI Data Platform with agent-ready gateway and liquid-cooled storage
Dell announced the expansion of the Dell AI Data Platform at SC26 in Chicago, November 15-20, 2026, transforming the platform from a data pipeline engine into an open, agent-ready platform. The update introduces an Agentic Data Gateway and hybrid orchestration to turn siloed data into governed context, with new high-density and liquid-cooled storage hardware scheduled for release in 1H 2027. (Agentic Data Gateway: An AI gatekeeper that provides a customer-controlled entry point for models and data.) This transition is designed to move AI production beyond experimental pilots by allowing agents to access enterprise data directly on customer-controlled infrastructure.
Is this for you?
Enterprise IT managers and data engineers can deploy autonomous agents on local infrastructure without moving heavy data sets.
- If you manage enterprise IT infrastructure, the Agentic Data Gateway allows you to deploy AI agents while maintaining strict data sovereignty.
- If you are an AI data engineer, hybrid orchestration allows you to run pipelines where data already lives to avoid movement costs.
- If you are a storage architect, the 2OU liquid-cooled storage allows you to align storage with high-density AI rack cooling requirements.
- Skip this if you are a home PC gamer as it focuses on enterprise data platforms and rack-scale storage.
What you can do now
- Check the 1H 2027 release schedule for the 2OU liquid-cooled storage and the high-density 1U platform.
- Compare your current LLM costs against the 76% reduction in token consumption achieved in internal testing.
Key figures
- Agentic Data Gateway
- Customer-controlled Model Context Protocol (MCP) server (AI data standard)
- Token reduction
- Up to 76% via internal testing
- Liquid-cooled storage
- 32 drives in 2OU, expected 1H 2027
- Answer accuracy
- Improved from 60% to 94% (Dell internal testing)
- Time to first token
- 19x faster via CMX offload
If you already use the Dell AI Data Platform to manage unstructured data pipelines, the new Agentic Data Gateway and hybrid orchestration capabilities provide the mechanism to scale those pipelines into autonomous agents without moving your primary data sets.
The Agentic Data Gateway provides a single front door for secure AI access
The Agentic Data Gateway serves as a customer-controlled Model Context Protocol (MCP) server—an AI data standard—and an AI agent gateway that gives external agents secure access to approved models, data, and MCP tools. (Model Context Protocol (MCP): An AI data standard used for agents to discover and interact with data services.) Because it is built into the Data Orchestration Engine, the gateway provides a single entry point for agents built inside the platform as well as external agents built in tools like Claude Code, Cursor, and Codex. This architecture allows the gateway to route requests to internal and external LLMs while managing credentials, quotas, and token accounting in one place.
Credentials are managed centrally by the Agentic Data Gateway
Credentials are managed centrally by the Agentic Data Gateway, meaning agents no longer need to carry their own api keys. The gateway operates entirely inside the customer’s environment and does not store prompts or long-lived secrets. By exposing OpenAI-compatible and MCP endpoints, the gateway allows existing agents to connect without requiring custom integration for every new data source.
Hybrid orchestration solves the problem of data gravity
Hybrid data orchestration enables the Data Orchestration Engine to deploy and run pipeline containers on various types of customer-provided compute, including Kubernetes clusters, Dell AI Factory infrastructure, non-Dell infrastructure, and public cloud endpoints. This mechanism addresses “data gravity” by allowing the orchestration to occur where the data and compute already live, rather than forcing the movement of heavy unstructured data to a separate environment. Dell manages the deployment, scaling, and teardown of these containers automatically to ensure the infrastructure remains efficient.
Storage Intelligence and Data Intelligence create a trusted semantic layer
Storage Intelligence scans existing storage arrays to build a metadata repository, allowing teams to identify and curate relevant files for AI pipelines. This capability provides storage administrators and IT managers with a unified view of data ownership and usage across both Dell and non-Dell storage systems. By identifying what data exists and who owns it, teams can prepare data for AI with higher confidence.
Data Intelligence provides the end-to-end lineage
Data Intelligence provides the end-to-end lineage—or data history—needed for agents to understand the provenance of the information they retrieve. This capability includes metadata management, search across tables and vector indexes, and a business glossary. Together, these tools create a semantic layer where context is built once and reused, which Dell claims in internal testing can cut total token consumption by up to 76% and raise answer accuracy from 60% to 94%.
New Exascale hardware options arrive in 1H 2027
Dell will expand its Exascale portfolio to include the world’s first direct liquid-cooled storage (water-cooled storage) in 1H 2027. This new hardware features 32 drives in a 2OU form factor designed for Open Rack standards, allowing the storage to share the same liquid-cooled rack design as high-density AI compute. For users requiring a smaller footprint, Dell also expects to release a high-density 1U platform in 1H 2027, which provides 67% higher drive density per rack unit compared to previous 2U configurations.
Software-defined PowerScale on Dell Exascale
Software-defined PowerScale on Dell Exascale has already achieved NVIDIA-Certified Storage validation at the Foundation level. This certification confirms that the storage is ready for AI-accelerated data access in large GPU environments. For inference-heavy workloads, Dell notes that NVIDIA CMX enables KV Cache offload—or fast response memory—to shared Dell AI Storage, resulting in a 19x faster Time to first token (TTFT) compared to standard vLLM. Additionally, S3 over RDMA provides 230% higher throughput and 80% lower latency for object-heavy inference pipelines.
Ecosystem integrations with LangChain, OpenAI, and Unstructured.io
Technical partnerships with LangChain, OpenAI, and Unstructured.io integrate familiar development tools into the Dell AI Data Platform. The LangChain collaboration brings LangSmith tracing and evaluation tools into the Data Orchestration engine, allowing for full traces of pipelines from model calls to tool calls. The OpenAI Codex partnership provides secure development environments using Red Hat OpenShift Dev Spaces, while the Unstructured.io partnership introduces a File Transform application for parsing, chunking, and embedding documents into RAG preprocessing pipelines.
Who this is for: Enterprise IT Infrastructure Managers
Enterprise IT Infrastructure Managers should use the Agentic Data Gateway to deploy AI agents while maintaining strict data sovereignty. The gateway provides an auditable path to thousands of concurrent agents by managing credentials centrally. AI Data Engineers can utilize hybrid orchestration to run pipelines where the data already lives, avoiding the costs associated with moving massive amounts of unstructured data. Storage Architects should plan for the 2OU liquid-cooled storage, which offers a way to align storage with high-density AI rack cooling requirements. GPU Cluster Administrators can leverage KV Cache offload to reduce inference latency and keep GPUs productive during concurrent user sessions.
Who this is not for: Home PC Gamers
Home PC Gamers can ignore this content as it focuses on enterprise data platforms and rack-scale storage rather than consumer hardware; gamers should instead look for consumer GPU and SSD reviews. Software Developers focused on front-end libraries should skip this as the technical focus is on backend data orchestration and storage protocols; these users should look for web development frameworks instead.
A retail enterprise scenario for data context
Sven runs a large retail enterprise with siloed customer data and thousands of unstructured PDFs. By deploying the Dell AI Data Platform, Sven uses Storage Intelligence to scan his archives and Data Intelligence to build a semantic layer. He then uses the Unstructured.io integration to process his PDFs and deploys a customer service agent via the Agentic Data Gateway. The agent can now provide accurate answers with cited sources in 0.03 seconds while the data remains in the original silo. If Sven had instead moved all his data to a single cloud bucket first, he would have faced significant data movement costs and increased security risks.
What this can be used for
| Application | Target User |
|---|---|
| Building RAG preprocessing pipelines for raw documents | Data Engineering Team |
| Designing liquid-cooled AI racks with integrated storage | Data Center Architect |
| Reducing LLM costs by reusing context | AI Platform Manager |
| Tracing the lineage of an agent’s answer to specific files | Compliance Officer |
| Speeding up inference for concurrent user sessions | Inference Engineering Team |
What this replaces: The Agentic Data Gateway reduces
The Agentic Data Gateway reduces the need for hard-coded API keys for agents by providing a centralized management system, which reduces credential sprawl. Hybrid orchestration reduces the need for costly data movement caused by data gravity, as it allows pipelines to run on the customer’s existing infrastructure. Finally, the decoupled architecture reduces the need for “rip-and-replace” vector database migrations by allowing storage and data engines to scale independently.
What is not yet known: Dell did not specify
Dell did not specify which individuals or specific teams made the decision to expand the platform in this manner. The source also does not provide specific instructions on how to deploy pipeline containers on kubernetes clusters for data orchestration. The availability of the 1H 2027 hardware depends on manufacturing timelines which Dell did not disclose.
Dell AI Data Platform: what to check, compare and watch
- Liquid cooling
- The Dell AI Data Platform allows storage to share the rack with liquid-cooled compute and networking. Storage Architect: plan the liquid cooling infrastructure to support the Dell AI Data Platform's new storage hardware.
Dell AI Data Platform: what is known and what is not yet disclosed
- Who made the decision or is responsible?
- The original report did not say who made the decision or is responsible.
- Who is affected, and how?
- Enterprises can turn siloed data into governed context for production AI agents on customer-controlled infrastructure without forcing data movement.
- How to deploy pipeline containers on kubernetes clusters for data orchestration?
- The original report did not say how to deploy pipeline containers on kubernetes clusters.
Key terms
- Agentic Data Gateway
- An AI gatekeeper that provides a customer-controlled entry point for models and data.
- Model Context Protocol (MCP)
- An AI data standard used for agents to discover and interact with data services.
- Direct liquid-cooled storage
- Water-cooled storage using fanless cold-plate cooling on all heat-producing parts.
- KV Cache offload
- Fast response memory that moves the key-value cache to shared storage to speed up inference.
- Data Lineage
- Data history providing end-to-end tracking of data quality and observability.
Reported by Dell Technologies on 16 November 2026: Dell AI Data Platform Makes Data Agent-Ready at SC26.
ExtremeSpec is reader-supported: some links on this page are affiliate links, and as an Amazon Associate ExtremeSpec earns from qualifying purchases at no extra cost to you. Learn more.
