和硕在OCP 2026展示英伟达Vera Rubin NVL72整柜与液冷能力，仍属产品/能力展示而非新增订单

Pegatron showcased its end-to-end AI infrastructure capabilities at the Open Compute Project (OCP) Global Summit 2026, highlighting Nvidia-based products designed to move from engineering development to large-scale cluster deployment. The Taiwanese electronics manufacturer presented rack-scale AI systems, liquid-cooling technologies, and integrated deployment capabilities as demand grows for infrastructure that can support increasingly complex AI workloads.

The company said successful AI infrastructure deployment depends on more than computing performance. System engineering, liquid cooling, power architecture, open standards, manufacturing readiness, and pre-deployment validation have become essential as enterprises expand from individual servers to large-scale clusters. Pegatron positioned its design, manufacturing, and verification capabilities as tools for accelerating the transition from product development to production deployment.
A central exhibit was the Nvidia Vera Rubin NVL72 platform. Pegatron's RA4803-72N3 rack-scale system integrates 72 Rubin GPUs, 36 Vera CPUs, and sixth-generation Nvidia NVLink in a fully liquid-cooled rack design. The system was presented as an example of rack-level integration and manufacturing capability, as well as the company's ability to advance products from initial design through cluster deployment.
Pegatron also displayed a liquid-cooled Nvidia Vera CPU system and Nvidia Groq 3 LPX products aimed at infrastructure requirements associated with AI agents. The Vera CPU system is designed to optimize operations between model-computation steps, including tool calls, sandboxed code execution, and reinforcement-learning loops. Additional exhibits included a fully liquid-cooled Nvidia HGX Rubin NVL8 system and AI-ready compute and storage infrastructure built around Advanced Micro Devices (AMD) EPYC 9006 series processors.
The manufacturer said its approach combines system design expertise, vertically integrated production, OCP-aligned architectures, and rack-level validation before deployment. Digital-twin simulation, thermal validation, automated manufacturing, and rack-level testing are used together to help customers scale AI servers, clusters, and data center infrastructure. The focus reflects the operational requirements of moving AI systems from engineering development into repeatable, large-scale production deployment.