Giga Computing Showcases AI Infrastructure at OCP Global Summit 2026
Giga Computing, a subsidiary of GIGABYTE, has unveiled its latest data center infrastructure solutions at the OCP Global Summit 2026. The company is highlighting its rack-scale AI portfolio, designed to support the growing demands of large-scale AI and cloud computing workloads.
Rack-Scale AI and NVIDIA Ecosystem
A major highlight of the exhibition is the fully configured NVIDIA Vera Rubin NVL72 platform. Already in production for hyperscalers and frontier AI labs, this rack features a 20.7TB unified HBM4 memory pool, enabling the execution of trillion-parameter models and million-token context windows. To address latency-sensitive tasks, the company also showcased the NVIDIA Groq 3 LPX rack, which optimizes the decoding phase to achieve the rapid response speeds required by AI agents.
Additionally, the NVIDIA Vera CPU rack, built on the NVIDIA MGX architecture and integrated with "Spectrum-X" Ethernet networking, was presented. This infrastructure is designed for agent orchestration and sandbox testing, supporting over 22,500 concurrent execution environments per rack to handle complex reinforcement learning workloads more efficiently than traditional CPU setups.
OCP Standards and Advanced Cooling
Giga Computing has significantly expanded its OCP ORV3 compliant lineup. These solutions, which support both AMD EPYC and Intel Xeon CPUs, encompass compute-centric nodes, high-speed flash storage, and management solutions. A key technical feature across this lineup is the integration of direct liquid cooling (DLC) to ensure stable high-performance operation.
Addressing the industry's shift toward more advanced power and thermal management, the company also featured an 800V CDU from partner TaiSol Electronics. This component is intended to support next-generation, high-density platforms that require sophisticated power delivery and cooling solutions.
Гика
