• by Admin

HPE Simplifies Supercomputing for Sovereign AI

The world of high performance computing, or HPC, is currently at a rather unique point. It is becoming more advanced and more essential, yet also more complicated to manage. AI, HPC, and quantum computing now overlap within the same ecosystem, and supercomputing customers are required to have systems that are not only fast, but also flexible and reliable enough to support research that is expected to change the world, ranging from medicine to climate science.

HPE aims to meet these demands through a series of updates it recently announced. These updates are not only about faster hardware, but also about making life easier for developers and system administrators.

Programming Software That No Longer Gives Developers Headaches

Anyone who has worked with a multi-vendor stack understands how frustrating it can be to update a single component without having to consider its ripple effects on other components. This is a real problem that HPC and AI developers have long faced, namely stacks that are too tightly integrated, difficult to update at the pace of technological change, and prone to instability whenever changes occur.

HPE addresses this challenge through HPE Supercomputing Programming Software. Notably, this software is now also available for HPE ProLiant Compute servers, meaning it is no longer limited to specialized supercomputing systems but can also be used by large enterprises running ProLiant DL and XD servers for AI training, tuning, and inference workloads. In short, it offers a more consistent experience across various HPE platforms.

The software offers three key elements.

• First, pre-validated programming environments. HPE integrates tools from vendors, open-source communities, and HPE itself, then validates them into a single, ready to use programming environment, so that setup no longer needs to start from scratch.

• Second, updates delivered as containerized environments. All tool updates are delivered as pre-validated, isolated environments. Developers can adopt new tools at their own pace without having to rebuild everything every time the infrastructure changes. This approach makes the stack more stable and updates safer.

• Third, HPE takes on the role of first call support for multi vendor assistance. Customers no longer need to act as their own vendor coordinator when issues arise. HPE performs technical triage and drives issue resolution across vendors, drawing on its experience from the Cray supercomputing platform.

End to End Multi Tenancy for Sovereign AI Research

For research laboratories, multi-tenant capability is crucial, since they require independent management and secure separation between users or critical workloads. HPE has added multi tenant capabilities at the networking and storage levels, forming a full stack multi tenant solution for sovereign AI research, with stronger data security for sensitive workloads on large scale AI systems.

On the networking side, HPE has released a new version of HPE Slingshot 400 software with multi tenancy capabilities based on media access control, or MAC, learning that ensures secure separation between users while preventing unauthorized routing from other groups, including open source or third party sources. The good news is that this capability can be applied to Slingshot 400 switches that customers have already deployed, so no new hardware is required.

On the storage side, HPE has made multi tenancy setup and management more granular through a new graphical user interface and API for the HPE Cray Supercomputing Storage Systems E2000. The new interface makes multi tenancy setup and management more intuitive, while the API opens the door to automation in large scale multi tenant environments.

All of this is supported by the HPE Cray Supercomputing platform, which unifies AI and HPC architecture, along with a direct liquid cooling system that improves energy efficiency.

Retiring Systems Comes With Its Own Set of Rules

One aspect often overlooked in discussions about supercomputing is how to properly retire systems that are no longer in use. This is far from a simple matter, especially for machines used in advanced scientific research that must comply with environmental, legal, and regional data sovereignty standards.

HPE operates Technology Renewal Centers that handle end of use asset management securely and efficiently. The numbers are quite impressive. Throughout 2025, 85 percent of servers that went through the renewal process were successfully upcycled and returned to active use, while 1.7 exabytes of data were securely sanitized.

Now, through HPE Financial Services, this support is being expanded specifically for air cooled HPC and AI systems. The new retirement capabilities give customers greater control, security, and oversight throughout the process, including security and configuration resets, proprietary testing, workload validation, advanced diagnostics, scale out validation, and trade compliance assurance. Ultimately, systems can be restored to factory settings with data securely handled and in accordance with regulations.

Conclusion

From simplified programming software and end to end multi-tenant capabilities for sovereign AI research to a more secure system retirement process, HPE is working to cover the entire lifecycle of supercomputing infrastructure, from the day a system is deployed to the day it is retired. For research institutions and large enterprises that are increasingly dependent on large scale computing, this translates into a lighter administrative burden, allowing them to focus on what truly matters: the research itself. 

Author: Ghea Devita

Marketing Communication PT Perkom Indah Murni