Specifications
Platform
Hyve Orion, featuring the NVIDIA HGX platform
Power density
Up to 50 kW per rack
Configurations
Two 10U options: JBOG (Just a Bunch of GPUs) or complete monolithic server
Graphics module
6U module design, forward-compatible across GPU generations
GPU support
NVIDIA H100, H200, B100 and B200
GPU architectures
Hopper and Blackwell
Cooling options
Air-cooled and direct liquid-cooled configurations
Deployment model
Rack-ready, plug-and-play turnkey solution
Validation
Rack-level and infrastructure-level testing before shipment
Partner status
NVIDIA Elite OEM Partner in the NVIDIA Partner Network
Global services
Hyve Unified Global Services (HUGS) spanning 50 countries
Ownership
Wholly owned subsidiary of TD SYNNEX Corporation
Target customer
Hyperscale and large enterprise AI infrastructure deployments
Design goal
Future-proof GPU transitions while optimising total cost of ownership
Overview
Hyve Orion is a rack-ready implementation of the NVIDIA HGX platform, built by Hyve Solutions — the TD SYNNEX subsidiary that operates as an NVIDIA Elite OEM Partner. Rather than selling a bare accelerator node, Hyve ships a validated rack: two 10U configurations are offered, either a JBOG (Just a Bunch of GPUs) expansion rack or a complete monolithic server.
The differentiator is the 6U graphics module. It is designed to span GPU generations, accepting NVIDIA H100, H200, B100 and B200 modules, and to work in both air-cooled and direct liquid-cooled racks — so an operator can move from Hopper to Blackwell compute, or from air to liquid cooling, without replacing the rack architecture underneath. Orion racks are rated to 50 kW of power density.
Validation happens at rack and infrastructure level, not just at the node, and support runs through the Hyve Unified Global Services programme across 50 countries. Hyve positions the platform explicitly at hyperscalers that need rack-level repeatability and a controlled path from Hopper to Blackwell while keeping total cost of ownership in view.
Key Benefits
Cross-generational module — one 6U graphics module covers H100, H200, B100 and B200, so GPU refreshes do not force a rack redesign. Air or liquid — the same platform is offered in air-cooled and direct liquid-cooled configurations. Rack-level validation — testing extends to rack and infrastructure, not just the server node. Global service envelope — Hyve Unified Global Services spans 50 countries as part of TD SYNNEX.
Applications
Hyperscale AI training and inference deployments, AI infrastructure refresh programmes migrating from Hopper to Blackwell GPUs, data centres transitioning from air to direct liquid cooling, GPU expansion racks (JBOG) attached to existing compute, and enterprise clusters needing validated rack-level delivery rather than component sourcing.
Request a Quote — HYVE ORION — RACK-READY NVIDIA HGX AI PLATFORM
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →