HPE extends supercomputing software to ProLiant servers

HPE has expanded its supercomputing software to ProLiant servers and added new multi-tenant networking, storage and retirement services for HPC and AI systems. The changes are aimed at organisations running sovereign AI research and mixed-vendor computing estates.

The update brings HPE Supercomputing Programming Software to HPE ProLiant Compute servers for the first time, extending a software stack previously associated with the company's Cray supercomputing line. The move is intended to give customers a more consistent software environment across different classes of systems as HPC and AI workloads increasingly run on the same infrastructure.

The software is designed to address a common problem for developers and system administrators: maintaining tightly integrated toolchains assembled from multiple vendors and open-source projects. HPE is instead offering pre-validated programming environments that combine vendor, open-source and HPE tools in a single stack.

Delivered as containers, the environments allow customers to roll out updates as isolated packages rather than rebuild software stacks from scratch on each system. HPE will also act as the first point of support for customers using the multi-vendor software environment, taking responsibility for technical triage across suppliers.

The extension to ProLiant covers HPE ProLiant DL and XD servers, which are commonly used for AI training, tuning and inference workloads. By applying the same software approach across supercomputers and more conventional server estates, HPE is seeking to simplify operations for large organisations with mixed environments.

Multi-tenant systems

Alongside the software announcement, HPE introduced multi-tenant functions in networking and storage for HPC users. The additions are aimed at national laboratories and other research organisations that need to isolate users and workloads while running shared infrastructure.

On the networking side, a new version of HPE Slingshot 400 software adds multi-tenancy functions based on media access control learning. The software is intended to enforce separation between user groups and restrict unauthorised routing between them, including from open-source and third-party environments.

These networking features can be applied to Slingshot 400 switches that are already deployed, potentially allowing existing customers to add workload separation without replacing hardware.

Storage changes centre on the HPE Cray Supercomputing Storage Systems E2000. HPE has added a graphical user interface and an application programming interface for setting up and managing fine-grained multi-tenancy in the file system.

The graphical interface is intended to make configuration and administration more straightforward, while the API is aimed at customers automating management across large-scale environments. The broader goal is to support secure separation of sensitive research workloads, particularly in sovereign AI settings where data control and access boundaries are central requirements.

Retirement services

HPE also used the announcement to expand the scope of its end-of-life services for air-cooled HPC and AI infrastructure through HPE Financial Services. The new offer covers security and configuration resets, testing, workload validation, diagnostics, scale-out validation and trade compliance assurance.

The retirement process is intended to return systems to factory settings before they leave service, helping protect data and meet regulatory or sovereign requirements. The focus on retirement reflects the sensitivity of many HPC and AI deployments, where systems may have handled research data subject to strict internal and national controls.

HPE said its Technology Renewal Centres processed large volumes of equipment and data in the past year. According to the company, 85% of servers handled through the centres in 2025 were upcycled and returned to active use, while 1.7 exabytes of data were securely sanitised.

The latest measures come as suppliers across the HPC market adapt products for a world in which traditional simulation workloads increasingly overlap with AI training and inference. That convergence has created demand for systems that can support mixed workloads, tighter security boundaries and more standardised management across different hardware platforms.

HPE linked the latest products to that trend in comments from its Asia Pacific leadership. "Across the Asia Pacific, sovereign AI research is accelerating as governments and enterprises seek to build and control their own AI capabilities, pushing HPC and AI workloads to converge on the same infrastructure," said Fumiki Negishi, Vice President and General Manager, HPC & AI APJ GTM Division, HPE.

Negishi said the company sees software consistency, secure workload separation and lifecycle management as connected parts of the same customer requirement. "In this era of convergence, the advantage will belong to organisations that can transform complexity into AI ecosystems that are simple, secure and sovereign. With HPE Supercomputing Programming Software now extending to ProLiant Compute servers, and new multi-tenant capabilities across HPE Slingshot 400 software and HPE Cray Supercomputing Storage Systems E2000, we are giving customers in the region a consistent, secure foundation from deployment through retirement," he said.