At the OCP Global Summit 2025 in San Jose, Pegatron showcased a comprehensive lineup focused on next-generation AI, HPC, and professional visual computing workloads. The company also collaborated with key partners such as Nvidia and AMD to present a series of server platforms.
Mei-Hui Wang, Pegatron's general manager of the Server Business Unit, stated that AI is transforming every industry, and demand for advanced infrastructure is rapidly growing. Pegatron provides platforms designed for next-generation performance and is validated through real-world deployment, helping enterprises seamlessly scale from server-level performance to cluster-scale supercomputing.
The lineup displayed by Pegatron included server platforms such as Nvidia's GB300 NVL72, HGX B300, and RTX PRO, as well as AMD's Instinct MI355X.
Among them, the Nvidia GB300 NVL72 platform, equipped with 72 Nvidia Blackwell Ultra GPUs and 36 Nvidia Grace CPUs, delivers extreme performance to accelerate large-scale AI training and inference applications. Systems based on this platform, such as the RA4802-72N2, have already begun shipping.
Pegatron's AI inference solutions are centered around the Nvidia HGX B300 platform, which features Nvidia Blackwell Ultra GPUs. Compared with the previous-generation Nvidia Hopper, AI computing performance has increased sevenfold, supporting next-generation AI inference and the most demanding data center workloads.
Additionally, Pegatron's product lineup has been expanded with two new server platforms featuring Nvidia RTX PRO 6000 Blackwell Server Edition GPUs. These systems integrate Nvidia Blackwell GPUs to deliver cutting-edge AI computing, photorealistic rendering, generative AI, digital twin simulation, and complex industrial application workflows.
Apart from Nvidia, Pegatron is also expanding its partnership with AMD. The AMD Instinct MI355X platform achieves breakthroughs in AI supercomputing with ultra-high density. This liquid-cooled cabinet system is equipped with 128 GPUs and 32 CPUs, providing a powerful foundation for AI training, generative AI, HPC, and scientific computing.
Article translated by Emily Kuo and edited by Jack Wu
