HPE and AMD converge supercomputing and AI via liquid-cooled GX5000
Hewlett Packard Enterprise is deepening its integration with AMD to converge supercomputing and AI infrastructure, utilizing EPYC 6 processors and Instinct MI430X accelerators. The companies are promoting liquid-cooled platforms like the GX5000 to solve thermal management challenges in high-density data centers and AI factories.
Key Takeaways
- HPE's GX5000 platform supports 40% more AMD EPYC 9996 cores per rack than the nearest dense competitor.
- The new Instinct MI430X accelerator delivers up to 288 TFLOPS of hardware-based FP64 performance for high-precision scientific computing.
- HPE is under contract to deliver the Discovery supercomputer to Oak Ridge National Laboratory, succeeding the current Frontier system.
- Liquid cooling is now a requirement for enterprise data centers to maintain density as AI silicon power requirements soar.
Why It Matters
The convergence of HPC and AI marks a shift where traditional supercomputing hardware becomes the blueprint for enterprise AI infrastructure. By moving high-density cooling and scientific-grade precision (FP64) into mainstream platforms like the GX5000, HPE and AMD are positioning themselves as primary alternatives to Nvidia's integrated systems. For streaming and media companies, this signals the arrival of infrastructure capable of handling autonomous AI agents and complex metadata simulations at scale. Watch for the 2027 deployment of the MI430X as a benchmark for sovereign AI performance against Nvidia's Rubin architecture.
Additional Context
The collaboration between HPE and AMD is central to the U.S. Department of Energy’s modern infrastructure strategy. In October 2025, the DOE announced a $1 billion public-private investment to build the Discovery exascale supercomputer and the Lux AI cluster at Oak Ridge National Laboratory. Per ORNL, December 2025, the Lux cluster is scheduled for deployment in early 2026, utilizing MI355X GPUs and Pensando networking to serve as a sovereign AI cloud for national research. Discovery will follow in 2028, built on the liquid-cooled GX5000 architecture and upcoming 'Venice' EPYC CPUs. HPE is simultaneously diversifying its hardware options to capture more of the burgeoning AI factory market. In March 2026, HPE expanded its portfolio to include the Compute XD700 and GX240 blades optimized for Nvidia’s Rubin GPUs and Vera CPUs, per StorageReview. This dual-track strategy allows HPE to offer multi-tenant, liquid-cooled environments regardless of the customer's silicon preference. Meanwhile, industry data suggests a rapid move toward liquid cooling; per CoolIT Systems, July 2026, AI compute has officially exceeded the practical limits of air cooling in high-density facilities, making direct-liquid platforms a standard requirement for next-generation builds. AMD is positioning its Helios rack-scale systems as a direct counterweight to Nvidia's dominance in the agentic AI era. During the July 2026 Advancing AI event, AMD CEO Lisa Su stated that inference now accounts for roughly 60% of AI compute capacity, surpassing training for the first time. To meet this demand, AMD launched the EPYC 9006 series, featuring up to 256 cores per chip, alongside a unified ROCm software stack designed to streamline AI software development across heterogeneous hardware, per TechPowerUp, July 2026.
Read full article at siliconangle.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source