PEGATRON Unveils AI-Optimized Server Innovations at GTC Paris 2025

Paris, France – June 11, 2025 – PEGATRON, a globally recognized Design, Manufacturing, and Service (DMS) provider, is pleased to announce its participation in GTC Paris 2025, where it will introduce a comprehensive lineup of next-generation AI server solutions. Built on NVIDIA Blackwell architecture, PEGATRON’s latest systems are tailored to meet the evolving demands of AI training, reasoning, and enterprise-scale deployments.

NVIDIA GB300 NVL72
Leading the showcase is the RA4802-72N2, built on the NVIDIA GB300 NVL72 rack system. This high-performance platform incorporates 72 NVIDIA Blackwell Ultra GPUs and 36 NVIDIA Grace CPUs, delivering up to a 50X increase in AI factory output with advanced inference capabilities. To address the demands of high-performance workloads, the rack features PEGATRON’s in-house developed Coolant Distribution Unit (CDU), supporting 310 kW of cooling capacity and equipped with redundant hot-swappable pumps to ensure reliability and operational efficiency in mission-critical data center environments.

NVIDIA HGX B300 Liquid-cooled and Air-cooled Solutions
Also featured is the AS208-2A1, a 2U liquid-cooled AI GPU server equipped with an NVIDIA HGX B300 system with dual AMD EPYC™ 9005 processors in a compact form factor. Scalable within a 48U NVIDIA MGX rack, it supports configurations of up to 128 GPUs and 32 CPUs, operating under a high-efficiency direct liquid cooling design. Optimized for the AI reasoning era, the system leverages the Blackwell Ultra architecture to deliver breakthrough performance for complex workloads such as agentic systems, real-time reasoning, and video generation, making it an ideal platform for next-generation AI data centers. For data centers seeking robust compute capabilities without a liquid cooling infrastructure, PEGATRON introduces an 8U air-cooled GPU server featuring an NVIDIA HGX B300 systems. This system supports either dual Intel® Xeon® 6 processors (AS801-2T1) or AMD EPYC™ 9005 CPUs (AS801-2A1), providing deployment flexibility while maintaining high-performance compute for AI model training, scientific research, and advanced analytics.

NVIDIA RTX PRO Server
Rounding out the portfolio is the AS400-2A1, a 4U RTX PRO server that integrates 8 NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs and dual AMD EPYC™ 9005 processors. This platform offers near-universal acceleration for a wide range of enterprise AI workloads—from multimodal AI inference and physical AI to design, scientific computing, graphics, and video applications. It supports the NVIDIA MGX PCIe Switch Board with the ConnectX-8 SuperNIC, enabling 800 Gbps ultra-high-speed networking in a single device, high throughput and improved energy and cost efficiency, making it an ideal foundation for modern AI infrastructure.

“As AI continues to reshape industries, the need for powerful, scalable, and energy-efficient infrastructure is more critical than ever,” said May Wang, Vice President of PEGATRON. “Our new platforms—built on NVIDIA Blackwell Ultra and RTX PRO architectures—are purpose-built for the next generation of data center and enterprise AI factories, enabling customers to unlock new levels of performance, reliability, and productivity.”

PEGATRON welcomes technology leaders, innovators, and data center professionals to visit Booth G10 at Paris Expo Porte de Versailles on June 11–12, 2025. Connect with PEGATRON’s expert team to discover how our AI-optimized platforms are shaping the future of intelligent infrastructure.