Hewlett Packard Enterprise (NYSE: HPE) announced new artificial intelligence and supercomputing systems developed with NVIDIA at the NVIDIA GTC 2026 conference. The company introduced compute and networking offerings paired with software and services for service providers, sovereign entities, research laboratories, and large enterprises.

HPE becomes among the first companies to offer an NVIDIA Vera CPU compute blade and adds NVIDIA Quantum-X800 InfiniBand networking to its HPE Cray supercomputers. The HPE Cray Supercomputing GX240 Compute blade features up to 16 NVIDIA Vera CPUs and can scale up to 40 blades with 640 NVIDIA Vera CPUs per rack.

The company enhanced its AI Factory portfolio with NVIDIA Vera Rubin and NVIDIA Blackwell platforms. The NVIDIA Vera Rubin NVL72 by HPE rack-scale system includes 36 NVIDIA Vera CPUs and 72 NVIDIA Rubin GPUs, designed for models exceeding 1 trillion parameters. The HPE Compute XD700 server built on NVIDIA HGX Rubin NVL8 supports up to 128 Rubin GPUs per rack.

HPE's AI factories received NVIDIA Cloud Partner program endorsement and now support multi-tenancy through NVIDIA Multi-Instance GPU technology. The systems integrate with Red Hat Enterprise Linux and OpenShift as part of Red Hat AI Enterprise, along with NVIDIA AI Enterprise solutions.

The HPE Cray Supercomputing GX240 Compute blade will be available in 2027, while NVIDIA Quantum-X800 InfiniBand networking becomes available on HPE Cray Supercomputing GX5000 systems in 2027. The NVIDIA Vera Rubin NVL72 by HPE system will be available in December 2026, and the HPE Compute XD700 will be available in early 2027.

Organizations including Argonne National Laboratory, HLRS, Hudson River Trading, and the Korea Institute of Science and Technology Information have selected HPE AI infrastructure systems, according to the company's statement.