PEGATRON, a globally recognized total server solutions provider, today unveiled its next-generation AI infrastructure portfolio and AI factory validation framework at COMPUTEX 2026. Powered by NVIDIA Vera Rubin platform , NVIDIA HGX Rubin NVL8, and NVIDIA RTX PRO Servers, and aligned with the NVIDIA DSX AI factory reference design, Pegatron is advancing how AI factories are designed, validated, and deployed - from digital twin simulation to production-ready infrastructure.
As an NVIDIA DSX ecosystem partner, Pegatron is extending AI infrastructure validation beyond traditional manufacturing by integrating SimReady digital twins and AI factory workflows into its server development and deployment process. Through the NVIDIA Omniverse DSX Blueprint Pegatron enables customers to simulate, validate, and optimize rack-scale AI infrastructure before deployment.
"AI factories require not only high-performance computing, but also end-to-end infrastructure intelligence,"said Gary Cheng, President & CEO of Pegatron."By leveraging NVIDIA DSX architecture, SimReady digital twins, and AI factory validation capabilities, Pegatron has strengthened both its manufacturing operations and AI infrastructure integration expertise. Together with NVIDIA, we aim to help customers simulate, validate, and optimize AI deployments, improving the scalability, reliability, and efficiency of next-generation AI data centers."
NVIDIA Vera Rubin NVL72 Rack-Scale AI Supercomputer
The centerpiece of showcase is the NVIDIA Vera Rubin NVL72 Pegatron's RA4803-72N3, a fully liquid-cooled supercomputer built on the third-generation NVIDIA MGX rack design for seamless drop-in deployment into existing Blackwell data centers.
Unified Fabric: Integrates 72 NVIDIA Rubin GPUs, 36 NVIDIA Vera CPUs, NVIDIA ConnectX-9 SuperNICs, and NVIDIA BlueField-4 DPUs. It scales up via NVIDIA NVLink 6 and scales out via NVIDIA Quantum-X800 InfiniBand or NVIDIA Spectrum-X Ethernet. Paired with NVIDIA Groq 3 LPX racks, it delivers ultra-fast trillion-parameter inference.
TCO Optimization: Trains mixture-of-experts (MoE) models using 1/4th the GPUs, slashes token costs by 10X, and delivers 10X more tokens per megawatt compared to previous Blackwell architectures.
NVIDIA Vera CPU: Purpose-built for agentic AI and reinforcement learning, completing workloads 50% faster with twice the efficiency of traditional CPUs.
Scaling AI Factories with NVIDIA HGX Rubin NVL8
Pegatron also introduced the AS210-2T1-8H3, a 2U fully liquid-cooled server based on NVIDIA HGX Rubin NVL8 with dual NVIDIA Vera CPUs or Intel Xeon 6 processors.
The platform leverages NVIDIA Rubin GPUs, NVIDIA NVLink 6, NVIDIA ConnectX-9 SuperNICs, NVIDIA BlueField-4 DPUs, and NVIDIA Spectrum-X Ethernet to accelerate enterprise AI and HPC workloads.
To support rack-scale deployments, Pegatron expanded the architecture into: RA4800-64H3 — supporting eight 2U systems per rack, up to 64 GPUs and 16 CPUs. RA4800-72H3 — supporting nine systems, up to 72 GPUs and 18 CPUs within a 48U liquid-cooled rack.
Factory-Native NVIDIA DSX Integration and SimReady Digital Twins
Beyond hardware innovation, Pegatron showcased its NVIDIA DSX-aligned AI factory validation workflow, integrating digital twin simulation, thermal analysis, and full-load burn-in testing within the same manufacturing facility.
Pegatron's L11 thermal lab combines Ansys Fluent CFD simulation, NVIDIA Omniverse digital twins, liquid cooling analysis, and power modeling to optimize AI factory infrastructure before deployment, while the L12 burn-in room validates every system at 100% TDP against DSX digital twin predictions prior to shipment.
Using the NVIDIA AI Factory Digital Twin Pipeline Samples, Pegatron converts engineering data into SimReady OpenUSD assets for NVIDIA Omniverse DSX Blueprint environments, enabling customers to validate cooling, power, and rack-scale AI deployments before installation.
Pegatron also highlighted PEGAVERSE, its AI-native manufacturing intelligence platform that combines simulation, factory telemetry, and AI-driven operational analysis to continuously optimize AI factory performance and efficiency.
Experience the Future of AI Infrastructure at COMPUTEX 2026
Pegatron invites customers, partners, and media representatives to explore its AI infrastructure and DSX-integrated digital twin ecosystem at COMPUTEX 2026.Date: June 2–5, 2026,Location: Taipei Nangang Exhibition Center Hall 1, 4F,Booth: L0104.