HPE, NVIDIA Expand AI Infrastructure with New Blades, Networking, and AI Factories

Event summary

  • HPE and NVIDIA jointly announced updates to the NVIDIA AI Computing by HPE portfolio, focused on large-scale AI factories and supercomputers.
  • HPE is introducing the HPE Cray Supercomputing GX240 compute blade featuring up to 16 NVIDIA Vera CPUs, with availability in 2027.
  • The NVIDIA Vera Rubin NVL72 rack-scale system, designed for AI models exceeding 1 trillion parameters, will be available in December 2026.
  • HPE’s Compute XD700 AI server, inspired by the Open Compute Project, will be available in early 2027, supporting up to 128 Rubin GPUs per rack.
  • HPE AI Factory portfolio with multi-tenancy and GPU passthrough will be available in Spring 2026.

The big picture

HPE's collaboration with NVIDIA represents a strategic push into the high-end AI infrastructure market, catering to organizations and nations requiring massive-scale compute capabilities. This move positions HPE to capitalize on the growing demand for sovereign AI solutions and compete with hyperscalers offering more standardized cloud services. The focus on liquid cooling and high-density architectures underscores the increasing power and complexity of AI workloads.

What we're watching

Adoption Rate
The speed at which organizations adopt the new HPE Cray GX240 blades and NVIDIA Vera Rubin systems will indicate the true demand for this level of AI compute density, particularly within sovereign environments.
Neo-Cloud Competition
The success of HPE's neo-cloud offerings will depend on its ability to differentiate from established hyperscalers and attract customers seeking more control over their AI infrastructure.
Software Integration
The effectiveness of the Red Hat and SUSE integrations will be critical for enterprise adoption, as seamless software management is essential for complex AI deployments.