Hewlett Packard Enterprise has announced major expansions to its NVIDIA AI Computing portfolio, integrating NVIDIA’s Vera Rubin platform and Blackwell architecture into AI factory offerings and supercomputing systems.
The centerpiece NVIDIA Vera Rubin NVL72 rack-scale system handles AI models exceeding one trillion parameters, combining 36 NVIDIA Vera CPUs with 72 NVIDIA Rubin GPUs and advanced networking infrastructure. The HPE Compute XD700 delivers 128 GPUs per rack, boosting AI training and inference density while reducing deployment costs.
HPE’s Cray Supercomputing GX5000 platform now features NVIDIA Vera CPU compute blades, with each GX240 blade supporting up to 16 NVIDIA Vera CPUs. The configuration scales to 40 blades per rack, delivering 640 NVIDIA Vera CPUs and 56,320 NVIDIA Olympus Arm cores.
Networking options include NVIDIA Quantum-X800 InfiniBand switches providing 144 ports at 800 Gb/s connectivity with power efficiency features. NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs are now available across HPE AI factory deployments.
The portfolio includes full-stack integration of compute, GPUs, networking, liquid cooling, software, and services targeting service providers, sovereign entities, and large enterprises. Leading research institutions including Argonne National Laboratory and HLRS have adopted the infrastructure.
Software enhancements feature NVIDIA Mission Control for workload orchestration, multi-tenancy support, and Red Hat Enterprise Linux integration. Availability ranges from immediate for certain GPUs to 2027 for Vera CPU blades.
