HPE doubled down on its Alletra storage products in 2025, and focussed in selling its own storage software rather than adding partner storage IP to its server systems.
The company’s storage product portfolio consists of its enterprise Alletra arrays, GreenLake File software, entry-level MSA arrays, Store Once deduping backup target appliances, MSL tape libraries, Zerto DR software, Simplivity HCI products, Data Fabric SW, and, up in the HPC/supercomputer world, the ClusterStor arrays.
Its quarterly storage revenue history for the past four years has been one of slight growth in fiscal 2022 over 2021, followed by a dip in 2023 and 2024, but a growth recovery in the final fy 2024 quarter followed by more growth through 2025. This corresponds with its focus on the Alletra storage product portfolio and its own storage intellectual property.
If we look at how HPE storage has progressed, compared with Dell, NetApp and Pure, we observe that Dell consistently sells more than the others. It leads HPE by a general 4x multiple. HPE is positioned midway between NetApp and Pure Storage. The gap with NetApp has been generally increasing since fy 2021, while Pure Storage has been narrowing its gap below HPE as its growth rate is faster.
ClusterStor products have been selling well as HPE wins supercomputer bids, such as the recent Oak Ridge National Labs Discovery supercomputer, and the Mission and Vision supercomputers at Los Alamos National Lab. ClusterStor has not featured in Nvidia GPU-based supercomputers though, despite its IO500-leading data performance.
In general, our view is that HPE storage, while selling well to enterprise buyers, has not benefitted much from neocloud AI training needs, losing out to VAST Data, DDN, and WEKA for example. Like the other main storage suppliers, it’s waiting for a storage sales uplift from AI inferencing workload needs.
A main storage event for HPE has been the development of the Alletra product line. Originally this was the coming together of the 3PAR-based Primera and Nimble and Apollo server brands in 2023. There were file (GreenLake File Storage based on Alletra MP hardware with OEM’d VAST Data file software), block (MP B10000) and object (MP X10000) Alletra systems at the start of 2025. In August this year, HPE announced an upgraded all-flash MP X10000 system with integrated Zerto functionality, saying it was the world’s fastest backup storage. It also had an integrated S3 over RDMA feature for fast object delivery to GPU servers.
CEO and president Antonio Neri told financial analysts at a Securities Day meeting in October that capitalizing on unstructured data market growth with its own IP through Alletra MP was one of HPE’s five strategic priorities. HPE also announced it was developing an Alletra MP X10000-based data fabric that month, running across distributed edge sites, co-lo centers, core data centers and the three main public clouds. This was part of its Nvidia-based AI Factory and Smart Cities initiative and features a global namespace, universal access, multi-protocol support, automated tiering, caching, and mirroring.
Previously, it had its Ezmeral data fabric, plus Alletra Block storage for AWS. It announced Alletra Block Storage for Azure at Nvidia GTC 2025 in March.
The Entry-level Alletra MP B10000 initially supported two nodes working together with point-to-point connections, obviating the need for a nework switch. This was upped to 4 nodes in November.
HPE announced Alletra MP X10000 Data Intelligence Nodes today, and these run an AI pipeline directly. These all-flash nodes are powered by Nvidia L40S GPUs, and use Nvidia’s AI Data Platform reference design and run its AI Enterprise software in the data path, checking object data as it’s ingested, extracting metadata, and creating vector embeddings for use by AI large language models and agents. MCP is supported, and the AI-ready data can be queried through native SQL query engines such as Spark, Trino, and Starburst Data.
HPE will soon, ESG says, offer the ability to deploy the X10000 as part of an HPE private Cloud AI (PCAI), designed by HPE to be a turnkey AI factory for enterprises. The Alletra Storage MP X10000 Data Intelligence Nodes will be available to order in January 2026. Read a Solution Brief document here and an ESG report on the new nodes here.
HPE IP
Apart from file-based Alletra the line is sold with HPE storage software, its own intellectual property (IP). HPE had a storage product sales line composed of its ProLiant servers running third-party storage software, such as Scality object storage, Qumulo scale-out filesystem and WEKA parallel filesystem software. It announced the end of these partnership deals in November, saying it was concentrating on selling its own IP, which would be more profitable. The three affected vendors each said they would continue supporting HPE servers via meet-in-the-channel deals.
At the end of 2025, HPE is a stronger storage market player than it was at the beginning, with Alletra MP sales riding the disaggregated storage market wave, leaving less money on the table by not selling third-party software, winning large-scale ClusterStore supercomputer deals, and its developing Data Fabric. The company says that, with cloud-native data infrastructure Alletra Storage, customers can: “Seamlessly run any app without compromise from edge to cloud with a cloud experience for every workload.” [Watch a video about this here.]
The Data Fabric and Alletra systems are both ready for the hoped-for arrival of the AI Inferencing storage market and HPE could, we think, grow its storage market share.
