HPE does not sell a single product called an AI data platform. It calls its answer a unified data layer, introduced at NVIDIA's GTC in March 2025 alongside the NVIDIA AI Data Platform reference design and expanded in May 2026. Where Dell sells one branded platform with named engines, HPE spreads the same jobs across three products. This article explains each, what shipped when, and how the two vendors line up.
1. HPE Alletra Storage MP X10000: storage that prepares data itself
The X10000 is HPE's object storage system, and the centre of the unified data layer.
- Object and file on one system. Native file storage arrived in the second quarter of 2026 alongside the S3 object store. The system now scales to 16 nodes and 23 PB of raw capacity, with a 100% data availability guarantee; RDMA-accelerated file access and the 16-node scale arrived in the third quarter.
- Data Intelligence Nodes. Optional nodes with NVIDIA L40S GPUs running NVIDIA AI Enterprise sit in the data path: they tag incoming data with metadata, create vector embeddings inside the storage, and expose the results to AI agents through built-in Model Context Protocol servers. They became orderable in January 2026. The processed data can be queried with Spark, Trino or Starburst.
- An SDK for building unstructured-data pipelines on the NVIDIA AI Data Platform design.
- Data protection through an accelerator node that ingests up to 2.5 PB an hour.
The idea is to do the "prepare and index" step where the files already are, rather than copying them into a separate pipeline.
2. HPE Data Fabric Software: one view over all the data
Formerly Ezmeral Data Fabric, this presents files, objects and databases across sites under one namespace, and supports Apache Iceberg tables. The May 2026 release, available immediately, added:
- policy-based data placement and movement across hybrid environments;
- metadata visibility, classification and lineage;
- governance through open standards, including Apache Polaris;
- a conversational, agentic assistant for finding data in the namespace.
This is the governance layer. It is notable that it shipped in 2026, whereas Dell's equivalent, the semantic layer and knowledge graph, is announced for 2027.
3. HPE AI Essentials in Private Cloud AI: the pipeline tools
HPE Private Cloud AI is HPE's turnkey private AI cloud built with NVIDIA, and AI Essentials is its software layer. For data engineers it provides:
- Apache Airflow for pipeline orchestration, Apache Spark and Ray for processing;
- EzPresto, a Presto-based SQL engine, for queries across sources;
- Apache Superset for dashboards;
- a data lakehouse gateway and a global namespace linking S3 and NFS endpoints.
An air-gapped edition exists for isolated environments, which matters for defence and some government deployments.
HPE's block storage, the Alletra Storage MP B10000, completes the picture with agentic support that detects and resolves storage issues, a 5:1 data-reduction guarantee and scale to six controller nodes in the third quarter of 2026.
HPE and Dell side by side
| Job | Dell AI Data Platform | HPE |
|---|---|---|
| File and object storage | PowerScale, ObjectScale, Exascale Storage | Alletra X10000 (object and file), B10000 (block) |
| Prepare, embed and index documents | Data Search Engine (Elastic), Data Orchestration Engine (Dataloop) | X10000 Data Intelligence Nodes |
| SQL across sources | Data Analytics Engine (Starburst) | EzPresto, Data Fabric namespace |
| Processing | Data Processing Engine (Spark) | Spark, Airflow and Ray in AI Essentials |
| Governance and business meaning | Semantic layer and knowledge graph, due 2027 | Data Fabric metadata, lineage and Polaris, available now |
| Packaged AI stack | Dell AI Factory with NVIDIA | HPE Private Cloud AI |
The main differences: Dell builds on partner software (Starburst, Elastic, Dataloop) and names each engine; HPE puts the AI processing into the storage itself and relies on its own Data Fabric plus open-source tools. Both follow NVIDIA's AI Data Platform design. HPE's governance features ship today.
The caution that applies to both
Analyst firm Futurum asked whether HPE's unified stack reduces integration complexity or repackages it under a new management layer. The same question applies to Dell. In our experience the answer depends less on the vendor than on the data work underneath: the pipelines, the quality checks and the access model that any of these platforms has to be fed with. That work is the data platform service; the vendor platform can replace any layer of it where a customer wants one product and one support contract.
Where it fits for us
HPE is one of our server partners; the sovereign AI cloud design pairs Dell GPU servers with HPE ProLiant CPU servers. For defence, the air-gapped edition of Private Cloud AI is worth a closer look, and we would test the X10000's embedding nodes on Bangla documents before relying on them, as we do with every retrieval component.
Sources
What this article was checked against on 19 October 2026.
- HPE delivers unified private clouds and data platforms (HPE, May 2026)
- The same announcement with availability dates (StorageNewsletter)
- HPE drives the agentic AI era with an intelligent unified data layer (March 2025)
- HPE storage in 2025: Alletra rises (Blocks & Files)
- HPE Alletra Storage MP X10000 (HPE)
- HPE AI Essentials for data engineering (HPE)
- Can HPE's unified private cloud and data platforms deliver AI simplicity? (Futurum)