Nvidia GTC 2026: Hyve Solutions Shares What it Actually Takes to Build AI Infrastructure at Scale

Hyve Solutions Corp., a wholly owned subsidiary of TD SYNNEX Corporation and a player in the design to worldwide deployment of hyperscale digital infrastructures, participated to the GTC 2026, Nvidia’s premier AI conference, that took place March 16-19 in San Jose, California.This year, Hyve returned to GTC not only as an exhibitor but as a featured voice on stage, with SVP, global engineering, Rami Khouri, presenting insights drawn from some of the world’s most demanding AI infrastructure deployments.

“There’s no better stage than GTC to highlight what we’ve built at Hyve, exceptional engineering talent, partnerships rooted in decades of trust, and a sharp view of where AI infrastructure is headed,” said Jerry Kagele, president, Hyve Solutions. “That trust isn’t something we take for granted. It’s what drives every decision we make and every product we ship.”

Engineering Insights from the Field
Rami Khouri, SVP, global engineering, Hyve Solutions, presented a session titled “Architecting AI-Ready Data Centers: Lessons from Leading-Edge Deployments,” that draws on firsthand experience designing and delivering AI infrastructure at hyperscale.

Khouri examined how rapidly escalating power densities, advanced cooling requirements, and rigorous validation protocols are redefining what it takes to build and operate AI factories. The session is designed for engineers, IT architects, and infrastructure decision-makers grappling with the real-world complexity of transitioning to AI-optimized data center environments.

“The AI infrastructure challenge is no longer theoretical,” said Rami Khouri. “It’s playing out in data centers right now at power densities and thermal loads the industry has never seen before. My goal at GTC is to share what we have learned from real-world deployments so that engineers and infrastructure leaders can build with confidence and avoid costly missteps.”

Hyve’s End-to-End AI Infrastructure Capabilities
As a System Partner in the Nvidia Partner Network (NPN), Hyve brings proven strength across headnode, GPU, and CPU platforms, spanning the full design-to-deployment spectrum.

Hyve showcased the depth of its AI infrastructure capabilities, including CPU-led AI platforms which play a strong role in new agentic AI workflows.

AI Design & Manufacturing: As a U.S.-based, fully vertically integrated company, Hyve designs and manufactures solutions purpose-built for AI, ML, and DL workloads. In-region production, including U.S. surface-mount technology (SMT) lines, ensures the precision and quality that next-generation AI infrastructure demands
Scalable Architecture: Hyve’s standardized architecture enables seamless GPU communication across servers and clusters, allowing customers to scale AI infrastructure without disruptive architectural changes. The result is a platform built for growth as much as performance
Advanced Testing & Liquid Cooling: Hyve operates in-region liquid cooling capabilities, providing comprehensive testing and validation for Nvidia GPUs at the densities AI workloads require. Custom testing configurations are available to meet specific performance and reliability requirements
Global Deployment Capabilities: Hyve’s global footprint and deep regional compliance expertise enable rapid AI data center deployment across all major geographies, helping enterprises accelerate AI initiatives wherever they operate

Showcasing Next-Generation AI Solutions at Hyve’s Booth
Hyve presented its latest product lineup, headlined by the Hyve Orion featuring Nvidia HGX Rubin NVL8. Built for the emerging demands of large-scale AI training, inference, and agentic AI workloads, Orion delivers a balanced compute architecture where CPUs orchestrate complex workflows while GPUs accelerate model execution.

The Orion featuring Nvidia HGX Rubin NVL8 platform is a 2RU, direct liquid-cooled AI server built within the Nvidia MGX 19″ rack standard. Powered by eight SXM8 Rubin GPUs with up to 2.3 TB of HBM4 memory, the system delivers up to 400 PFLOPS of FP4 inference performance. Direct liquid cooling captures up to 98% of system heat, enabling high-density rack configurations and supporting up to 128 GPUs per rack.

Hyve showcased its Build-to-Order Rack & Roll Networking for rapid, flexible deployment and its Custom Network Switches, engineered for optimized rack-level AI connectivity at scale.