Back
{{newsDetailData.type}}
{{newsDetailData.date}}
{{productlineTag.title}} ,
MSI, a global leader in high-performance computing, will showcase its latest AI infrastructure solutions at the OCP Global Summit 2025, held from October 13-16, 2025 at the San Jose Convention Center, California (Booth #A55). The company will debut its NVIDIA MGX-based high-density AI servers and the new EdgeXpert Desktop AI Supercomputer, highlighting a strategy that spans large-scale data centers and edge computing environments to meet the surging demand for generative AI and accelerated computing.
GPU Density with NVIDIA MGX
MSI’s GPU servers are built on the NVIDIA MGX modular architecture, accelerating AI workloads across training, inference, and simulation with support for the latest NVIDIA Hopper GPUs and NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.
The CG481-S6053 (4U) integrates dual AMD EPYC 9005 CPUs, 8 FHFL PCIe 6.0 GPU slots, 24 DDR5 DIMM slots, and 8×400G Ethernet networking via NVIDIA ConnectX-8 SuperNICs, ideal for large-scale AI training clusters requiring maximum GPU density and bandwidth.
The CG290-S3063 (2U) features a single Intel Xeon 6 CPU, 4 FHFL PCIe 5.0 GPU slots, and 16 DDR5 DIMM slots, providing a compact, efficient system optimized for AI inference and fine-tuning in space-sensitive environments.
Danny Hsu, General Manager of MSI’s Enterprise Platform Solutions, said:“As AI models continue to scale, data centers require solutions that maximize GPU density and bandwidth. MSI’s servers based on NVIDIA MGX are purpose-built to help customers achieve optimal performance and efficiency for both training and inference workloads.”
EdgeXpert Desktop AI Supercomputer
MSI unveiled the EdgeXpert Desktop AI Supercomputer, built on the NVIDIA DGX™ Spark platform and accelerated by the NVIDIA GB10 Grace Blackwell Superchip, which integrates a 20-core NVIDIA Grace™ CPU and NVIDIA Blackwell GPU connected by NVIDIA NVLink™ C2C technology to deliver 1000 AI TOPS (FP4) Tensor performance.
The EdgeXpert features a coherent unified memory architecture with 128GB LPDDR5x, enabling it to run models with up to 200B parameters on a single unit, and scale up to 405B parameters with two systems linked together. With a compact 1.2-liter volume and 1.2 kg weight, it combines high performance with portability. Augmenting cloud AI instances and GPU server infrastructure, EdgeXpert provides powerful on-premise performance, improving productivity while avoiding the risks of sensitive data leakage—driving the adoption of generative AI across education and research, financial technology, and the medical industry.
David Wu, General Manager of the Customized Product Solutions BU at MSI, stated: “The EdgeXpert Desktop AI Supercomputer brings AI development beyond the limits of the cloud or large data centers. Researchers, financial analysts, and medical professionals can now access supercomputer-level performance right at their desks, accelerating the widespread adoption of generative AI.”
Event Details
Date: October 14-16, 2025
Location: San Jose Convention Center, California
Booth: #A55
Location: San Jose Convention Center, California
Booth: #A55
MSI AIoT: https://www.msi.com/to/aiot
MSI AIoT Facebook: https://www.facebook.com/MSIAIoT
MSI AIoT LinkedIn: https://www.linkedin.com/showcase/msi-aiot
MSI Global YouTube: https://www.youtube.com/user/MSI
MSI AIoT Facebook: https://www.facebook.com/MSIAIoT
MSI AIoT LinkedIn: https://www.linkedin.com/showcase/msi-aiot
MSI Global YouTube: https://www.youtube.com/user/MSI