PEGATRON Unveils Cutting-Edge AI-Optimized Server Solutions at GTC Paris 2025
PEGATRON Unveils Cutting-Edge AI-Optimized Server Solutions at GTC Paris 2025
On June 12, 2025, PEGATRON, a globally recognized provider of design, manufacturing, and service (DMS), made headlines by showcasing its latest AI-optimized server innovations at the prestigious GTC Paris 2025 event. This release highlights PEGATRON's commitment to leveraging cutting-edge technology in artificial intelligence to offer exceptional solutions for enterprise needs.
The centerpiece of PEGATRON's presentation was an impressive lineup of servers designed to maximize performance in AI training, inference, and enterprise applications. These systems are engineered on the NVIDIA Blackwell architecture, marking a significant advancement in server technology aimed at fulfilling the demanding requirements of modern computing environments.
Key Innovations
Among the highlighted products was the Ultra High-Density Liquid-Cooled GPU Rack, specifically the RA4802-72N2 model. Built on the NVIDIA GB300 NVL72 rack system, this unit is outfitted with 72 NVIDIA Blackwell Ultra-GPUs and 36 NVIDIA Grace CPUs. This combination is tailored for AI-driven factories, providing up to a 50x increase in performance, making it a game changer for industrial applications.
Additionally, PEGATRON's Coolant Distribution Unit (CDU) is notable for its capacity to deliver 310 kW of cooling power, incorporating redundant, hot-swappable pumps that ensure optimal performance and reliability for critical workloads.
Another remarkable product is the NVIDIA HGX B300, a liquid-cooled 2U GPU server featuring dual AMD EPYC™ 9005 CPUs. This server is designed for scalability within an NVIDIA MGX-48U rack, supporting configurations of up to 128 GPUs and 32 CPUs, all operated via a high-efficiency direct liquid cooling system. Its architecture brings outstanding performance, particularly for complex workloads involving agent-based systems, real-time reasoning, and even video generation – ideal for next-gen AI data centers.
For businesses preferring air-cooled systems, PEGATRON showcases an air-cooled 8U GPU server utilizing the NVIDIA HGX B300 system. This configuration supports dual Intel® Xeon® 6 processors or AMD EPYC™ 9005 GPUs, providing the necessary flexibility combined with substantial computing power for training AI models, scientific research, and advanced analytics applications.
Also on display was the RTX PRO Server, the AS400-2A1 model, featuring 8 NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs alongside two AMD EPYC™ 9005 CPUs. This server is versatile enough to handle a wide array of AI workloads, ranging from multimodal inference to physical AI applications, graphics design, and scientific computing. It integrates the NVIDIA MGX PCIe Switch Board with the ConnectX-8 SuperNIC, facilitating ultra-high-speed networking at 800 Gbit/s while enhancing throughput, energy, and cost efficiency.
Commitment to AI-Driven Infrastructure
May Wang, Vice President at PEGATRON, emphasized the urgency for robust, scalable, and energy-efficient infrastructure in an era where AI continues to revolutionize industries. “Our new platforms, built on the NVIDIA Blackwell Ultra and RTX PRO architectures, empower enterprises to achieve a level of performance, reliability, and productivity that's crucial in today's landscape,” she stated during the presentation.
Visitors at GTC Paris 2025 were encouraged to experience these innovations firsthand at PEGATRON's booth G10, located at the Paris Expo Porte de Versailles on June 11-12, 2025.
For more information, follow PEGATRON on social media platforms such as LinkedIn and YouTube, or visit the PEGATRON SVR website.
About PEGATRON
Founded on January 1, 2008, the PEGATRON Corporation is dedicated to providing innovative design, systematic production, and comprehensive manufacturing services that meet customer needs efficiently and effectively. With a wealth of experience in server development and manufacturing, PEGATRON is focused on creating a diverse range of high-performance servers that cater to the current and future demands of cloud service providers and enterprise data centers alike.