October 12, 2026 - Giga Computing, a subsidiary of GIGABYTE and a leader in open compute and infrastructure solutions, today announced its participation in the OCP Global Summit, where Giga Computing is bringing its rack-scale AI portfolio to the show floor. Highlights at GIGABYTE booth #A39 include a fully configured NVIDIA Vera Rubin NVL72 platform, representative nodes from a NVIDIA Vera CPU rack and NVIDIA Groq 3 LPX rack, and an OCP-compliant rack pairing compute, storage, management, and GPU nodes. Together, the display shows Giga Computing's engineering depth across the full spectrum of AI infrastructure, from flagship rack-scale accelerated systems to open, standards-based building blocks for scalable deployment.
NVIDIA Vera Rubin NVL72 made its debut at NVIDIA GTC 2026, and since then it has gone into production and into data centers of hyperscalers, neoclouds, and frontier AI labs. A single rack shares a 20.7TB of unified HBM4 memory pool to unlock trillion-parameter and million-token-context models. To further accelerate economy at scale and tokens/sec per user, the Groq 3 LPX rack handles the latency-sensitive decode phase to achieve near-instant response speeds needed for AI agents. Visitors can see firsthand a node from the Groq 3 LPX rack.
Another powerful inference solution at the GIGABYTE booth is one from the NVIDIA Vera CPU rack. The Vera CPU Rack delivers the dense, standalone CPU infrastructure that agentic AI and reinforcement learning demand. That is over 22,500 concurrent execution environments per rack for agent orchestration and sandbox testing, completing workloads faster and more efficiently than traditional CPU infrastructure. Built on NVIDIA MGX architecture and integrated with NVIDIA Spectrum-X Ethernet networking, it gives AI factories the scale-out sandboxing muscle that needs CPUs to perform.
The Giga Computing portfolio for OCP-based products has matured significantly since its early work on OCP ORV1, expanding into a comprehensive OCP ORV3 product lineup. A GIGABYTE OCP ORV3 rack on display at the summit introduces solutions that complement rack-scale solutions or stand on their own. Supporting both AMD EPYC™ and Intel® Xeon® CPU platforms, the lineup spans compute-centric nodes, high-speed flash storage, and management solutions, giving customers flexibility across the full spectrum of OCP-based infrastructure. Every OCP server in the lineup incorporates direct liquid cooling (DLC) for stable, efficient high-performance operation. The GIGABYTE TO46 series delivers a dense yet powerful platform for 8-GPU baseboards, including NVIDIA HGX B300 system.
As compute infrastructure is moving into an era where power delivery and cooling systems are becoming more revolutionary, at the GIGABYTE booth will feature an 800V CDU from partner TaiSol Electronics, introducing a solution to support next-gen platforms at scale.
Together, these offerings reflect Giga Computing’s commitment to advancing every layer of AI and cloud infrastructure, from rack-scale accelerated computing to open, standards-based building blocks. Visitors to OCP Global Summit 2026 can experience these solutions firsthand at GIGABYTE booth #A39.
For queries or more information, please contact sales.
